You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@iceberg.apache.org by gi...@apache.org on 2025/11/13 10:24:36 UTC
(iceberg-rust) branch gh-pages updated: deploy: 12c4c2110174b76eb82e584cb7933efdc9afcecc
This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a commit to branch gh-pages
in repository https://gitbox.apache.org/repos/asf/iceberg-rust.git
The following commit(s) were added to refs/heads/gh-pages by this push:
new 095984ab deploy: 12c4c2110174b76eb82e584cb7933efdc9afcecc
095984ab is described below
commit 095984ab7cbe885a2004566122f8b79052fe518e
Author: liurenjie1024 <li...@users.noreply.github.com>
AuthorDate: Thu Nov 13 10:24:30 2025 +0000
deploy: 12c4c2110174b76eb82e584cb7933efdc9afcecc
---
.../iceberg/arrow/record_batch_transformer.rs.html | 2895 ++++++++++----------
1 file changed, 1494 insertions(+), 1401 deletions(-)
diff --git a/api/src/iceberg/arrow/record_batch_transformer.rs.html b/api/src/iceberg/arrow/record_batch_transformer.rs.html
index 475019b1..84383319 100644
--- a/api/src/iceberg/arrow/record_batch_transformer.rs.html
+++ b/api/src/iceberg/arrow/record_batch_transformer.rs.html
@@ -21,1425 +21,1518 @@
<a href=#21 id=21 data-nosnippet>21</a><span class="kw">use </span>arrow_array::{
<a href=#22 id=22 data-nosnippet>22</a> Array <span class="kw">as </span>ArrowArray, ArrayRef, BinaryArray, BooleanArray, Date32Array, Float32Array,
<a href=#23 id=23 data-nosnippet>23</a> Float64Array, Int32Array, Int64Array, NullArray, RecordBatch, RecordBatchOptions, StringArray,
-<a href=#24 id=24 data-nosnippet>24</a>};
-<a href=#25 id=25 data-nosnippet>25</a><span class="kw">use </span>arrow_cast::cast;
-<a href=#26 id=26 data-nosnippet>26</a><span class="kw">use </span>arrow_schema::{
-<a href=#27 id=27 data-nosnippet>27</a> DataType, FieldRef, Schema <span class="kw">as </span>ArrowSchema, SchemaRef <span class="kw">as </span>ArrowSchemaRef, SchemaRef,
-<a href=#28 id=28 data-nosnippet>28</a>};
-<a href=#29 id=29 data-nosnippet>29</a><span class="kw">use </span>parquet::arrow::PARQUET_FIELD_ID_META_KEY;
-<a href=#30 id=30 data-nosnippet>30</a>
-<a href=#31 id=31 data-nosnippet>31</a><span class="kw">use </span><span class="kw">crate</span>::arrow::schema_to_arrow_schema;
-<a href=#32 id=32 data-nosnippet>32</a><span class="kw">use </span><span class="kw">crate</span>::spec::{
-<a href=#33 id=33 data-nosnippet>33</a> Literal, PartitionSpec, PrimitiveLiteral, Schema <span class="kw">as </span>IcebergSchema, Struct, Transform,
-<a href=#34 id=34 data-nosnippet>34</a>};
-<a href=#35 id=35 data-nosnippet>35</a><span class="kw">use crate</span>::{Error, ErrorKind, <span class="prelude-ty">Result</span>};
-<a href=#36 id=36 data-nosnippet>36</a>
-<a href=#37 id=37 data-nosnippet>37</a><span class="doccomment">/// Build a map of field ID to constant value for identity-partitioned fields.
-<a href=#38 id=38 data-nosnippet>38</a>///
-<a href=#39 id=39 data-nosnippet>39</a>/// Implements Iceberg spec "Column Projection" rule #1: use partition metadata constants
-<a href=#40 id=40 data-nosnippet>40</a>/// only for identity-transformed fields. Non-identity transforms (bucket, truncate, year, etc.)
-<a href=#41 id=41 data-nosnippet>41</a>/// store derived values in partition metadata, so source columns must be read from data files.
-<a href=#42 id=42 data-nosnippet>42</a>///
-<a href=#43 id=43 data-nosnippet>43</a>/// Example: For `bucket(4, id)`, partition metadata has `id_bucket = 2` (bucket number),
-<a href=#44 id=44 data-nosnippet>44</a>/// but the actual `id` values (100, 200, 300) are only in the data file.
-<a href=#45 id=45 data-nosnippet>45</a>///
-<a href=#46 id=46 data-nosnippet>46</a>/// Matches Java's `PartitionUtil.constantsMap()` which filters `if (field.transform().isIdentity())`.
+<a href=#24 id=24 data-nosnippet>24</a> StructArray,
+<a href=#25 id=25 data-nosnippet>25</a>};
+<a href=#26 id=26 data-nosnippet>26</a><span class="kw">use </span>arrow_buffer::NullBuffer;
+<a href=#27 id=27 data-nosnippet>27</a><span class="kw">use </span>arrow_cast::cast;
+<a href=#28 id=28 data-nosnippet>28</a><span class="kw">use </span>arrow_schema::{
+<a href=#29 id=29 data-nosnippet>29</a> DataType, FieldRef, Schema <span class="kw">as </span>ArrowSchema, SchemaRef <span class="kw">as </span>ArrowSchemaRef, SchemaRef,
+<a href=#30 id=30 data-nosnippet>30</a>};
+<a href=#31 id=31 data-nosnippet>31</a><span class="kw">use </span>parquet::arrow::PARQUET_FIELD_ID_META_KEY;
+<a href=#32 id=32 data-nosnippet>32</a>
+<a href=#33 id=33 data-nosnippet>33</a><span class="kw">use </span><span class="kw">crate</span>::arrow::schema_to_arrow_schema;
+<a href=#34 id=34 data-nosnippet>34</a><span class="kw">use </span><span class="kw">crate</span>::spec::{
+<a href=#35 id=35 data-nosnippet>35</a> Literal, PartitionSpec, PrimitiveLiteral, Schema <span class="kw">as </span>IcebergSchema, Struct, Transform,
+<a href=#36 id=36 data-nosnippet>36</a>};
+<a href=#37 id=37 data-nosnippet>37</a><span class="kw">use crate</span>::{Error, ErrorKind, <span class="prelude-ty">Result</span>};
+<a href=#38 id=38 data-nosnippet>38</a>
+<a href=#39 id=39 data-nosnippet>39</a><span class="doccomment">/// Build a map of field ID to constant value for identity-partitioned fields.
+<a href=#40 id=40 data-nosnippet>40</a>///
+<a href=#41 id=41 data-nosnippet>41</a>/// Implements Iceberg spec "Column Projection" rule #1: use partition metadata constants
+<a href=#42 id=42 data-nosnippet>42</a>/// only for identity-transformed fields. Non-identity transforms (bucket, truncate, year, etc.)
+<a href=#43 id=43 data-nosnippet>43</a>/// store derived values in partition metadata, so source columns must be read from data files.
+<a href=#44 id=44 data-nosnippet>44</a>///
+<a href=#45 id=45 data-nosnippet>45</a>/// Example: For `bucket(4, id)`, partition metadata has `id_bucket = 2` (bucket number),
+<a href=#46 id=46 data-nosnippet>46</a>/// but the actual `id` values (100, 200, 300) are only in the data file.
<a href=#47 id=47 data-nosnippet>47</a>///
-<a href=#48 id=48 data-nosnippet>48</a>/// # References
-<a href=#49 id=49 data-nosnippet>49</a>/// - Spec: https://iceberg.apache.org/spec/#column-projection
-<a href=#50 id=50 data-nosnippet>50</a>/// - Java: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java:constantsMap()
-<a href=#51 id=51 data-nosnippet>51</a></span><span class="kw">fn </span>constants_map(
-<a href=#52 id=52 data-nosnippet>52</a> partition_spec: <span class="kw-2">&</span>PartitionSpec,
-<a href=#53 id=53 data-nosnippet>53</a> partition_data: <span class="kw-2">&</span>Struct,
-<a href=#54 id=54 data-nosnippet>54</a>) -> HashMap<i32, PrimitiveLiteral> {
-<a href=#55 id=55 data-nosnippet>55</a> <span class="kw">let </span><span class="kw-2">mut </span>constants = HashMap::new();
-<a href=#56 id=56 data-nosnippet>56</a>
-<a href=#57 id=57 data-nosnippet>57</a> <span class="kw">for </span>(pos, field) <span class="kw">in </span>partition_spec.fields().iter().enumerate() {
-<a href=#58 id=58 data-nosnippet>58</a> <span class="comment">// Only identity transforms should use constant values from partition metadata
-<a href=#59 id=59 data-nosnippet>59</a> </span><span class="kw">if </span><span class="macro">matches!</span>(field.transform, Transform::Identity) {
-<a href=#60 id=60 data-nosnippet>60</a> <span class="comment">// Get the partition value for this field
-<a href=#61 id=61 data-nosnippet>61</a> </span><span class="kw">if let </span><span class="prelude-val">Some</span>(Literal::Primitive(value)) = <span class="kw-2">&</span>partition_data[pos] {
-<a href=#62 id=62 data-nosnippet>62</a> constants.insert(field.source_id, value.clone());
-<a href=#63 id=63 data-nosnippet>63</a> }
-<a href=#64 id=64 data-nosnippet>64</a> }
-<a href=#65 id=65 data-nosnippet>65</a> }
-<a href=#66 id=66 data-nosnippet>66</a>
-<a href=#67 id=67 data-nosnippet>67</a> constants
-<a href=#68 id=68 data-nosnippet>68</a>}
-<a href=#69 id=69 data-nosnippet>69</a>
-<a href=#70 id=70 data-nosnippet>70</a><span class="doccomment">/// Indicates how a particular column in a processed RecordBatch should
-<a href=#71 id=71 data-nosnippet>71</a>/// be sourced.
-<a href=#72 id=72 data-nosnippet>72</a></span><span class="attr">#[derive(Debug)]
-<a href=#73 id=73 data-nosnippet>73</a></span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">enum </span>ColumnSource {
-<a href=#74 id=74 data-nosnippet>74</a> <span class="comment">// signifies that a column should be passed through unmodified
-<a href=#75 id=75 data-nosnippet>75</a> // from the file's RecordBatch
-<a href=#76 id=76 data-nosnippet>76</a> </span>PassThrough {
-<a href=#77 id=77 data-nosnippet>77</a> source_index: usize,
-<a href=#78 id=78 data-nosnippet>78</a> },
-<a href=#79 id=79 data-nosnippet>79</a>
-<a href=#80 id=80 data-nosnippet>80</a> <span class="comment">// signifies that a column from the file's RecordBatch has undergone
-<a href=#81 id=81 data-nosnippet>81</a> // type promotion so the source column with the given index needs
-<a href=#82 id=82 data-nosnippet>82</a> // to be promoted to the specified type
-<a href=#83 id=83 data-nosnippet>83</a> </span>Promote {
-<a href=#84 id=84 data-nosnippet>84</a> target_type: DataType,
-<a href=#85 id=85 data-nosnippet>85</a> source_index: usize,
-<a href=#86 id=86 data-nosnippet>86</a> },
-<a href=#87 id=87 data-nosnippet>87</a>
-<a href=#88 id=88 data-nosnippet>88</a> <span class="comment">// Signifies that a new column has been inserted before the column
-<a href=#89 id=89 data-nosnippet>89</a> // with index `index`. (we choose "before" rather than "after" so
-<a href=#90 id=90 data-nosnippet>90</a> // that we can use usize; if we insert after, then we need to
-<a href=#91 id=91 data-nosnippet>91</a> // be able to store -1 here to signify that a new
-<a href=#92 id=92 data-nosnippet>92</a> // column is to be added at the front of the column list).
-<a href=#93 id=93 data-nosnippet>93</a> // If multiple columns need to be inserted at a given
-<a href=#94 id=94 data-nosnippet>94</a> // location, they should all be given the same index, as the index
-<a href=#95 id=95 data-nosnippet>95</a> // here refers to the original RecordBatch, not the interim state after
-<a href=#96 id=96 data-nosnippet>96</a> // a preceding operation.
-<a href=#97 id=97 data-nosnippet>97</a> </span>Add {
-<a href=#98 id=98 data-nosnippet>98</a> target_type: DataType,
-<a href=#99 id=99 data-nosnippet>99</a> value: <span class="prelude-ty">Option</span><PrimitiveLiteral>,
-<a href=#100 id=100 data-nosnippet>100</a> },
-<a href=#101 id=101 data-nosnippet>101</a> <span class="comment">// The iceberg spec refers to other permissible schema evolution actions
-<a href=#102 id=102 data-nosnippet>102</a> // (see https://iceberg.apache.org/spec/#schema-evolution):
-<a href=#103 id=103 data-nosnippet>103</a> // renaming fields, deleting fields and reordering fields.
-<a href=#104 id=104 data-nosnippet>104</a> // Renames only affect the schema of the RecordBatch rather than the
-<a href=#105 id=105 data-nosnippet>105</a> // columns themselves, so a single updated cached schema can
-<a href=#106 id=106 data-nosnippet>106</a> // be re-used and no per-column actions are required.
-<a href=#107 id=107 data-nosnippet>107</a> // Deletion and Reorder can be achieved without needing this
-<a href=#108 id=108 data-nosnippet>108</a> // post-processing step by using the projection mask.
-<a href=#109 id=109 data-nosnippet>109</a></span>}
-<a href=#110 id=110 data-nosnippet>110</a>
-<a href=#111 id=111 data-nosnippet>111</a><span class="attr">#[derive(Debug)]
-<a href=#112 id=112 data-nosnippet>112</a></span><span class="kw">enum </span>BatchTransform {
-<a href=#113 id=113 data-nosnippet>113</a> <span class="comment">// Indicates that no changes need to be performed to the RecordBatches
-<a href=#114 id=114 data-nosnippet>114</a> // coming in from the stream and that they can be passed through
-<a href=#115 id=115 data-nosnippet>115</a> // unmodified
-<a href=#116 id=116 data-nosnippet>116</a> </span>PassThrough,
-<a href=#117 id=117 data-nosnippet>117</a>
-<a href=#118 id=118 data-nosnippet>118</a> Modify {
-<a href=#119 id=119 data-nosnippet>119</a> <span class="comment">// Every transformed RecordBatch will have the same schema. We create the
-<a href=#120 id=120 data-nosnippet>120</a> // target just once and cache it here. Helpfully, Arc<Schema> is needed in
-<a href=#121 id=121 data-nosnippet>121</a> // the constructor for RecordBatch, so we don't need an expensive copy
-<a href=#122 id=122 data-nosnippet>122</a> // each time we build a new RecordBatch
-<a href=#123 id=123 data-nosnippet>123</a> </span>target_schema: Arc<ArrowSchema>,
-<a href=#124 id=124 data-nosnippet>124</a>
-<a href=#125 id=125 data-nosnippet>125</a> <span class="comment">// Indicates how each column in the target schema is derived.
-<a href=#126 id=126 data-nosnippet>126</a> </span>operations: Vec<ColumnSource>,
-<a href=#127 id=127 data-nosnippet>127</a> },
-<a href=#128 id=128 data-nosnippet>128</a>
-<a href=#129 id=129 data-nosnippet>129</a> <span class="comment">// Sometimes only the schema will need modifying, for example when
-<a href=#130 id=130 data-nosnippet>130</a> // the column names have changed vs the file, but not the column types.
-<a href=#131 id=131 data-nosnippet>131</a> // we can avoid a heap allocation per RecordBach in this case by retaining
-<a href=#132 id=132 data-nosnippet>132</a> // the existing column Vec.
-<a href=#133 id=133 data-nosnippet>133</a> </span>ModifySchema {
-<a href=#134 id=134 data-nosnippet>134</a> target_schema: Arc<ArrowSchema>,
-<a href=#135 id=135 data-nosnippet>135</a> },
-<a href=#136 id=136 data-nosnippet>136</a>}
-<a href=#137 id=137 data-nosnippet>137</a>
-<a href=#138 id=138 data-nosnippet>138</a><span class="attr">#[derive(Debug)]
-<a href=#139 id=139 data-nosnippet>139</a></span><span class="kw">enum </span>SchemaComparison {
-<a href=#140 id=140 data-nosnippet>140</a> Equivalent,
-<a href=#141 id=141 data-nosnippet>141</a> NameChangesOnly,
-<a href=#142 id=142 data-nosnippet>142</a> Different,
-<a href=#143 id=143 data-nosnippet>143</a>}
-<a href=#144 id=144 data-nosnippet>144</a>
-<a href=#145 id=145 data-nosnippet>145</a><span class="doccomment">/// Builder for RecordBatchTransformer to improve ergonomics when constructing with optional parameters.
-<a href=#146 id=146 data-nosnippet>146</a>///
-<a href=#147 id=147 data-nosnippet>147</a>/// See [`RecordBatchTransformer`] for details on partition spec and partition data.
-<a href=#148 id=148 data-nosnippet>148</a></span><span class="attr">#[derive(Debug)]
-<a href=#149 id=149 data-nosnippet>149</a></span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">struct </span>RecordBatchTransformerBuilder {
-<a href=#150 id=150 data-nosnippet>150</a> snapshot_schema: Arc<IcebergSchema>,
-<a href=#151 id=151 data-nosnippet>151</a> projected_iceberg_field_ids: Vec<i32>,
-<a href=#152 id=152 data-nosnippet>152</a> partition_spec: <span class="prelude-ty">Option</span><Arc<PartitionSpec>>,
-<a href=#153 id=153 data-nosnippet>153</a> partition_data: <span class="prelude-ty">Option</span><Struct>,
-<a href=#154 id=154 data-nosnippet>154</a>}
-<a href=#155 id=155 data-nosnippet>155</a>
-<a href=#156 id=156 data-nosnippet>156</a><span class="kw">impl </span>RecordBatchTransformerBuilder {
-<a href=#157 id=157 data-nosnippet>157</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>new(
-<a href=#158 id=158 data-nosnippet>158</a> snapshot_schema: Arc<IcebergSchema>,
-<a href=#159 id=159 data-nosnippet>159</a> projected_iceberg_field_ids: <span class="kw-2">&</span>[i32],
-<a href=#160 id=160 data-nosnippet>160</a> ) -> <span class="self">Self </span>{
-<a href=#161 id=161 data-nosnippet>161</a> <span class="self">Self </span>{
-<a href=#162 id=162 data-nosnippet>162</a> snapshot_schema,
-<a href=#163 id=163 data-nosnippet>163</a> projected_iceberg_field_ids: projected_iceberg_field_ids.to_vec(),
-<a href=#164 id=164 data-nosnippet>164</a> partition_spec: <span class="prelude-val">None</span>,
-<a href=#165 id=165 data-nosnippet>165</a> partition_data: <span class="prelude-val">None</span>,
-<a href=#166 id=166 data-nosnippet>166</a> }
-<a href=#167 id=167 data-nosnippet>167</a> }
-<a href=#168 id=168 data-nosnippet>168</a>
-<a href=#169 id=169 data-nosnippet>169</a> <span class="doccomment">/// Set partition spec and data together for identifying identity-transformed partition columns.
-<a href=#170 id=170 data-nosnippet>170</a> ///
-<a href=#171 id=171 data-nosnippet>171</a> /// Both partition_spec and partition_data must be provided together since the spec defines
-<a href=#172 id=172 data-nosnippet>172</a> /// which fields are identity-partitioned, and the data provides their constant values.
-<a href=#173 id=173 data-nosnippet>173</a> /// One without the other cannot produce a valid constants map.
-<a href=#174 id=174 data-nosnippet>174</a> </span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>with_partition(
-<a href=#175 id=175 data-nosnippet>175</a> <span class="kw-2">mut </span><span class="self">self</span>,
-<a href=#176 id=176 data-nosnippet>176</a> partition_spec: Arc<PartitionSpec>,
-<a href=#177 id=177 data-nosnippet>177</a> partition_data: Struct,
-<a href=#178 id=178 data-nosnippet>178</a> ) -> <span class="self">Self </span>{
-<a href=#179 id=179 data-nosnippet>179</a> <span class="self">self</span>.partition_spec = <span class="prelude-val">Some</span>(partition_spec);
-<a href=#180 id=180 data-nosnippet>180</a> <span class="self">self</span>.partition_data = <span class="prelude-val">Some</span>(partition_data);
-<a href=#181 id=181 data-nosnippet>181</a> <span class="self">self
-<a href=#182 id=182 data-nosnippet>182</a> </span>}
-<a href=#183 id=183 data-nosnippet>183</a>
-<a href=#184 id=184 data-nosnippet>184</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>build(<span class="self">self</span>) -> RecordBatchTransformer {
-<a href=#185 id=185 data-nosnippet>185</a> RecordBatchTransformer {
-<a href=#186 id=186 data-nosnippet>186</a> snapshot_schema: <span class="self">self</span>.snapshot_schema,
-<a href=#187 id=187 data-nosnippet>187</a> projected_iceberg_field_ids: <span class="self">self</span>.projected_iceberg_field_ids,
-<a href=#188 id=188 data-nosnippet>188</a> partition_spec: <span class="self">self</span>.partition_spec,
-<a href=#189 id=189 data-nosnippet>189</a> partition_data: <span class="self">self</span>.partition_data,
-<a href=#190 id=190 data-nosnippet>190</a> batch_transform: <span class="prelude-val">None</span>,
-<a href=#191 id=191 data-nosnippet>191</a> }
-<a href=#192 id=192 data-nosnippet>192</a> }
-<a href=#193 id=193 data-nosnippet>193</a>}
-<a href=#194 id=194 data-nosnippet>194</a>
-<a href=#195 id=195 data-nosnippet>195</a><span class="doccomment">/// Transforms RecordBatches from Parquet files to match the Iceberg table schema.
-<a href=#196 id=196 data-nosnippet>196</a>///
-<a href=#197 id=197 data-nosnippet>197</a>/// Handles schema evolution, column reordering, type promotion, and implements the Iceberg spec's
-<a href=#198 id=198 data-nosnippet>198</a>/// "Column Projection" rules for resolving field IDs "not present" in data files:
-<a href=#199 id=199 data-nosnippet>199</a>/// 1. Return the value from partition metadata if an Identity Transform exists
-<a href=#200 id=200 data-nosnippet>200</a>/// 2. Use schema.name-mapping.default metadata to map field id to columns without field id (applied in ArrowReader)
-<a href=#201 id=201 data-nosnippet>201</a>/// 3. Return the default value if it has a defined initial-default
-<a href=#202 id=202 data-nosnippet>202</a>/// 4. Return null in all other cases
-<a href=#203 id=203 data-nosnippet>203</a>///
-<a href=#204 id=204 data-nosnippet>204</a>/// # Field ID Resolution
+<a href=#48 id=48 data-nosnippet>48</a>/// Matches Java's `PartitionUtil.constantsMap()` which filters `if (field.transform().isIdentity())`.
+<a href=#49 id=49 data-nosnippet>49</a>///
+<a href=#50 id=50 data-nosnippet>50</a>/// # References
+<a href=#51 id=51 data-nosnippet>51</a>/// - Spec: https://iceberg.apache.org/spec/#column-projection
+<a href=#52 id=52 data-nosnippet>52</a>/// - Java: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java:constantsMap()
+<a href=#53 id=53 data-nosnippet>53</a></span><span class="kw">fn </span>constants_map(
+<a href=#54 id=54 data-nosnippet>54</a> partition_spec: <span class="kw-2">&</span>PartitionSpec,
+<a href=#55 id=55 data-nosnippet>55</a> partition_data: <span class="kw-2">&</span>Struct,
+<a href=#56 id=56 data-nosnippet>56</a>) -> HashMap<i32, PrimitiveLiteral> {
+<a href=#57 id=57 data-nosnippet>57</a> <span class="kw">let </span><span class="kw-2">mut </span>constants = HashMap::new();
+<a href=#58 id=58 data-nosnippet>58</a>
+<a href=#59 id=59 data-nosnippet>59</a> <span class="kw">for </span>(pos, field) <span class="kw">in </span>partition_spec.fields().iter().enumerate() {
+<a href=#60 id=60 data-nosnippet>60</a> <span class="comment">// Only identity transforms should use constant values from partition metadata
+<a href=#61 id=61 data-nosnippet>61</a> </span><span class="kw">if </span><span class="macro">matches!</span>(field.transform, Transform::Identity) {
+<a href=#62 id=62 data-nosnippet>62</a> <span class="comment">// Get the partition value for this field
+<a href=#63 id=63 data-nosnippet>63</a> </span><span class="kw">if let </span><span class="prelude-val">Some</span>(Literal::Primitive(value)) = <span class="kw-2">&</span>partition_data[pos] {
+<a href=#64 id=64 data-nosnippet>64</a> constants.insert(field.source_id, value.clone());
+<a href=#65 id=65 data-nosnippet>65</a> }
+<a href=#66 id=66 data-nosnippet>66</a> }
+<a href=#67 id=67 data-nosnippet>67</a> }
+<a href=#68 id=68 data-nosnippet>68</a>
+<a href=#69 id=69 data-nosnippet>69</a> constants
+<a href=#70 id=70 data-nosnippet>70</a>}
+<a href=#71 id=71 data-nosnippet>71</a>
+<a href=#72 id=72 data-nosnippet>72</a><span class="doccomment">/// Indicates how a particular column in a processed RecordBatch should
+<a href=#73 id=73 data-nosnippet>73</a>/// be sourced.
+<a href=#74 id=74 data-nosnippet>74</a></span><span class="attr">#[derive(Debug)]
+<a href=#75 id=75 data-nosnippet>75</a></span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">enum </span>ColumnSource {
+<a href=#76 id=76 data-nosnippet>76</a> <span class="comment">// signifies that a column should be passed through unmodified
+<a href=#77 id=77 data-nosnippet>77</a> // from the file's RecordBatch
+<a href=#78 id=78 data-nosnippet>78</a> </span>PassThrough {
+<a href=#79 id=79 data-nosnippet>79</a> source_index: usize,
+<a href=#80 id=80 data-nosnippet>80</a> },
+<a href=#81 id=81 data-nosnippet>81</a>
+<a href=#82 id=82 data-nosnippet>82</a> <span class="comment">// signifies that a column from the file's RecordBatch has undergone
+<a href=#83 id=83 data-nosnippet>83</a> // type promotion so the source column with the given index needs
+<a href=#84 id=84 data-nosnippet>84</a> // to be promoted to the specified type
+<a href=#85 id=85 data-nosnippet>85</a> </span>Promote {
+<a href=#86 id=86 data-nosnippet>86</a> target_type: DataType,
+<a href=#87 id=87 data-nosnippet>87</a> source_index: usize,
+<a href=#88 id=88 data-nosnippet>88</a> },
+<a href=#89 id=89 data-nosnippet>89</a>
+<a href=#90 id=90 data-nosnippet>90</a> <span class="comment">// Signifies that a new column has been inserted before the column
+<a href=#91 id=91 data-nosnippet>91</a> // with index `index`. (we choose "before" rather than "after" so
+<a href=#92 id=92 data-nosnippet>92</a> // that we can use usize; if we insert after, then we need to
+<a href=#93 id=93 data-nosnippet>93</a> // be able to store -1 here to signify that a new
+<a href=#94 id=94 data-nosnippet>94</a> // column is to be added at the front of the column list).
+<a href=#95 id=95 data-nosnippet>95</a> // If multiple columns need to be inserted at a given
+<a href=#96 id=96 data-nosnippet>96</a> // location, they should all be given the same index, as the index
+<a href=#97 id=97 data-nosnippet>97</a> // here refers to the original RecordBatch, not the interim state after
+<a href=#98 id=98 data-nosnippet>98</a> // a preceding operation.
+<a href=#99 id=99 data-nosnippet>99</a> </span>Add {
+<a href=#100 id=100 data-nosnippet>100</a> target_type: DataType,
+<a href=#101 id=101 data-nosnippet>101</a> value: <span class="prelude-ty">Option</span><PrimitiveLiteral>,
+<a href=#102 id=102 data-nosnippet>102</a> },
+<a href=#103 id=103 data-nosnippet>103</a> <span class="comment">// The iceberg spec refers to other permissible schema evolution actions
+<a href=#104 id=104 data-nosnippet>104</a> // (see https://iceberg.apache.org/spec/#schema-evolution):
+<a href=#105 id=105 data-nosnippet>105</a> // renaming fields, deleting fields and reordering fields.
+<a href=#106 id=106 data-nosnippet>106</a> // Renames only affect the schema of the RecordBatch rather than the
+<a href=#107 id=107 data-nosnippet>107</a> // columns themselves, so a single updated cached schema can
+<a href=#108 id=108 data-nosnippet>108</a> // be re-used and no per-column actions are required.
+<a href=#109 id=109 data-nosnippet>109</a> // Deletion and Reorder can be achieved without needing this
+<a href=#110 id=110 data-nosnippet>110</a> // post-processing step by using the projection mask.
+<a href=#111 id=111 data-nosnippet>111</a></span>}
+<a href=#112 id=112 data-nosnippet>112</a>
+<a href=#113 id=113 data-nosnippet>113</a><span class="attr">#[derive(Debug)]
+<a href=#114 id=114 data-nosnippet>114</a></span><span class="kw">enum </span>BatchTransform {
+<a href=#115 id=115 data-nosnippet>115</a> <span class="comment">// Indicates that no changes need to be performed to the RecordBatches
+<a href=#116 id=116 data-nosnippet>116</a> // coming in from the stream and that they can be passed through
+<a href=#117 id=117 data-nosnippet>117</a> // unmodified
+<a href=#118 id=118 data-nosnippet>118</a> </span>PassThrough,
+<a href=#119 id=119 data-nosnippet>119</a>
+<a href=#120 id=120 data-nosnippet>120</a> Modify {
+<a href=#121 id=121 data-nosnippet>121</a> <span class="comment">// Every transformed RecordBatch will have the same schema. We create the
+<a href=#122 id=122 data-nosnippet>122</a> // target just once and cache it here. Helpfully, Arc<Schema> is needed in
+<a href=#123 id=123 data-nosnippet>123</a> // the constructor for RecordBatch, so we don't need an expensive copy
+<a href=#124 id=124 data-nosnippet>124</a> // each time we build a new RecordBatch
+<a href=#125 id=125 data-nosnippet>125</a> </span>target_schema: Arc<ArrowSchema>,
+<a href=#126 id=126 data-nosnippet>126</a>
+<a href=#127 id=127 data-nosnippet>127</a> <span class="comment">// Indicates how each column in the target schema is derived.
+<a href=#128 id=128 data-nosnippet>128</a> </span>operations: Vec<ColumnSource>,
+<a href=#129 id=129 data-nosnippet>129</a> },
+<a href=#130 id=130 data-nosnippet>130</a>
+<a href=#131 id=131 data-nosnippet>131</a> <span class="comment">// Sometimes only the schema will need modifying, for example when
+<a href=#132 id=132 data-nosnippet>132</a> // the column names have changed vs the file, but not the column types.
+<a href=#133 id=133 data-nosnippet>133</a> // we can avoid a heap allocation per RecordBach in this case by retaining
+<a href=#134 id=134 data-nosnippet>134</a> // the existing column Vec.
+<a href=#135 id=135 data-nosnippet>135</a> </span>ModifySchema {
+<a href=#136 id=136 data-nosnippet>136</a> target_schema: Arc<ArrowSchema>,
+<a href=#137 id=137 data-nosnippet>137</a> },
+<a href=#138 id=138 data-nosnippet>138</a>}
+<a href=#139 id=139 data-nosnippet>139</a>
+<a href=#140 id=140 data-nosnippet>140</a><span class="attr">#[derive(Debug)]
+<a href=#141 id=141 data-nosnippet>141</a></span><span class="kw">enum </span>SchemaComparison {
+<a href=#142 id=142 data-nosnippet>142</a> Equivalent,
+<a href=#143 id=143 data-nosnippet>143</a> NameChangesOnly,
+<a href=#144 id=144 data-nosnippet>144</a> Different,
+<a href=#145 id=145 data-nosnippet>145</a>}
+<a href=#146 id=146 data-nosnippet>146</a>
+<a href=#147 id=147 data-nosnippet>147</a><span class="doccomment">/// Builder for RecordBatchTransformer to improve ergonomics when constructing with optional parameters.
+<a href=#148 id=148 data-nosnippet>148</a>///
+<a href=#149 id=149 data-nosnippet>149</a>/// See [`RecordBatchTransformer`] for details on partition spec and partition data.
+<a href=#150 id=150 data-nosnippet>150</a></span><span class="attr">#[derive(Debug)]
+<a href=#151 id=151 data-nosnippet>151</a></span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">struct </span>RecordBatchTransformerBuilder {
+<a href=#152 id=152 data-nosnippet>152</a> snapshot_schema: Arc<IcebergSchema>,
+<a href=#153 id=153 data-nosnippet>153</a> projected_iceberg_field_ids: Vec<i32>,
+<a href=#154 id=154 data-nosnippet>154</a> partition_spec: <span class="prelude-ty">Option</span><Arc<PartitionSpec>>,
+<a href=#155 id=155 data-nosnippet>155</a> partition_data: <span class="prelude-ty">Option</span><Struct>,
+<a href=#156 id=156 data-nosnippet>156</a>}
+<a href=#157 id=157 data-nosnippet>157</a>
+<a href=#158 id=158 data-nosnippet>158</a><span class="kw">impl </span>RecordBatchTransformerBuilder {
+<a href=#159 id=159 data-nosnippet>159</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>new(
+<a href=#160 id=160 data-nosnippet>160</a> snapshot_schema: Arc<IcebergSchema>,
+<a href=#161 id=161 data-nosnippet>161</a> projected_iceberg_field_ids: <span class="kw-2">&</span>[i32],
+<a href=#162 id=162 data-nosnippet>162</a> ) -> <span class="self">Self </span>{
+<a href=#163 id=163 data-nosnippet>163</a> <span class="self">Self </span>{
+<a href=#164 id=164 data-nosnippet>164</a> snapshot_schema,
+<a href=#165 id=165 data-nosnippet>165</a> projected_iceberg_field_ids: projected_iceberg_field_ids.to_vec(),
+<a href=#166 id=166 data-nosnippet>166</a> partition_spec: <span class="prelude-val">None</span>,
+<a href=#167 id=167 data-nosnippet>167</a> partition_data: <span class="prelude-val">None</span>,
+<a href=#168 id=168 data-nosnippet>168</a> }
+<a href=#169 id=169 data-nosnippet>169</a> }
+<a href=#170 id=170 data-nosnippet>170</a>
+<a href=#171 id=171 data-nosnippet>171</a> <span class="doccomment">/// Set partition spec and data together for identifying identity-transformed partition columns.
+<a href=#172 id=172 data-nosnippet>172</a> ///
+<a href=#173 id=173 data-nosnippet>173</a> /// Both partition_spec and partition_data must be provided together since the spec defines
+<a href=#174 id=174 data-nosnippet>174</a> /// which fields are identity-partitioned, and the data provides their constant values.
+<a href=#175 id=175 data-nosnippet>175</a> /// One without the other cannot produce a valid constants map.
+<a href=#176 id=176 data-nosnippet>176</a> </span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>with_partition(
+<a href=#177 id=177 data-nosnippet>177</a> <span class="kw-2">mut </span><span class="self">self</span>,
+<a href=#178 id=178 data-nosnippet>178</a> partition_spec: Arc<PartitionSpec>,
+<a href=#179 id=179 data-nosnippet>179</a> partition_data: Struct,
+<a href=#180 id=180 data-nosnippet>180</a> ) -> <span class="self">Self </span>{
+<a href=#181 id=181 data-nosnippet>181</a> <span class="self">self</span>.partition_spec = <span class="prelude-val">Some</span>(partition_spec);
+<a href=#182 id=182 data-nosnippet>182</a> <span class="self">self</span>.partition_data = <span class="prelude-val">Some</span>(partition_data);
+<a href=#183 id=183 data-nosnippet>183</a> <span class="self">self
+<a href=#184 id=184 data-nosnippet>184</a> </span>}
+<a href=#185 id=185 data-nosnippet>185</a>
+<a href=#186 id=186 data-nosnippet>186</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>build(<span class="self">self</span>) -> RecordBatchTransformer {
+<a href=#187 id=187 data-nosnippet>187</a> RecordBatchTransformer {
+<a href=#188 id=188 data-nosnippet>188</a> snapshot_schema: <span class="self">self</span>.snapshot_schema,
+<a href=#189 id=189 data-nosnippet>189</a> projected_iceberg_field_ids: <span class="self">self</span>.projected_iceberg_field_ids,
+<a href=#190 id=190 data-nosnippet>190</a> partition_spec: <span class="self">self</span>.partition_spec,
+<a href=#191 id=191 data-nosnippet>191</a> partition_data: <span class="self">self</span>.partition_data,
+<a href=#192 id=192 data-nosnippet>192</a> batch_transform: <span class="prelude-val">None</span>,
+<a href=#193 id=193 data-nosnippet>193</a> }
+<a href=#194 id=194 data-nosnippet>194</a> }
+<a href=#195 id=195 data-nosnippet>195</a>}
+<a href=#196 id=196 data-nosnippet>196</a>
+<a href=#197 id=197 data-nosnippet>197</a><span class="doccomment">/// Transforms RecordBatches from Parquet files to match the Iceberg table schema.
+<a href=#198 id=198 data-nosnippet>198</a>///
+<a href=#199 id=199 data-nosnippet>199</a>/// Handles schema evolution, column reordering, type promotion, and implements the Iceberg spec's
+<a href=#200 id=200 data-nosnippet>200</a>/// "Column Projection" rules for resolving field IDs "not present" in data files:
+<a href=#201 id=201 data-nosnippet>201</a>/// 1. Return the value from partition metadata if an Identity Transform exists
+<a href=#202 id=202 data-nosnippet>202</a>/// 2. Use schema.name-mapping.default metadata to map field id to columns without field id (applied in ArrowReader)
+<a href=#203 id=203 data-nosnippet>203</a>/// 3. Return the default value if it has a defined initial-default
+<a href=#204 id=204 data-nosnippet>204</a>/// 4. Return null in all other cases
<a href=#205 id=205 data-nosnippet>205</a>///
-<a href=#206 id=206 data-nosnippet>206</a>/// Field ID resolution happens in ArrowReader before data is read (matching Java's ReadConf):
-<a href=#207 id=207 data-nosnippet>207</a>/// - If file has embedded field IDs: trust them (ParquetSchemaUtil.hasIds() = true)
-<a href=#208 id=208 data-nosnippet>208</a>/// - If file lacks IDs and name_mapping exists: apply name mapping (ParquetSchemaUtil.applyNameMapping())
-<a href=#209 id=209 data-nosnippet>209</a>/// - If file lacks IDs and no name_mapping: use position-based fallback (ParquetSchemaUtil.addFallbackIds())
-<a href=#210 id=210 data-nosnippet>210</a>///
-<a href=#211 id=211 data-nosnippet>211</a>/// By the time RecordBatchTransformer processes data, all field IDs are trustworthy.
-<a href=#212 id=212 data-nosnippet>212</a>/// This transformer only handles remaining projection rules (#1, #3, #4) for fields still "not present".
-<a href=#213 id=213 data-nosnippet>213</a>///
-<a href=#214 id=214 data-nosnippet>214</a>/// # Partition Spec and Data
+<a href=#206 id=206 data-nosnippet>206</a>/// # Field ID Resolution
+<a href=#207 id=207 data-nosnippet>207</a>///
+<a href=#208 id=208 data-nosnippet>208</a>/// Field ID resolution happens in ArrowReader before data is read (matching Java's ReadConf):
+<a href=#209 id=209 data-nosnippet>209</a>/// - If file has embedded field IDs: trust them (ParquetSchemaUtil.hasIds() = true)
+<a href=#210 id=210 data-nosnippet>210</a>/// - If file lacks IDs and name_mapping exists: apply name mapping (ParquetSchemaUtil.applyNameMapping())
+<a href=#211 id=211 data-nosnippet>211</a>/// - If file lacks IDs and no name_mapping: use position-based fallback (ParquetSchemaUtil.addFallbackIds())
+<a href=#212 id=212 data-nosnippet>212</a>///
+<a href=#213 id=213 data-nosnippet>213</a>/// By the time RecordBatchTransformer processes data, all field IDs are trustworthy.
+<a href=#214 id=214 data-nosnippet>214</a>/// This transformer only handles remaining projection rules (#1, #3, #4) for fields still "not present".
<a href=#215 id=215 data-nosnippet>215</a>///
-<a href=#216 id=216 data-nosnippet>216</a>/// **Bucket partitioning**: Distinguish identity transforms (use partition metadata constants)
-<a href=#217 id=217 data-nosnippet>217</a>/// from non-identity transforms like bucket (read from data file) to enable runtime filtering on
-<a href=#218 id=218 data-nosnippet>218</a>/// bucket-partitioned columns. For example, `bucket(4, id)` stores only the bucket number in
-<a href=#219 id=219 data-nosnippet>219</a>/// partition metadata, so actual `id` values must be read from the data file.
-<a href=#220 id=220 data-nosnippet>220</a>///
-<a href=#221 id=221 data-nosnippet>221</a>/// # References
-<a href=#222 id=222 data-nosnippet>222</a>/// - Spec: https://iceberg.apache.org/spec/#column-projection
-<a href=#223 id=223 data-nosnippet>223</a>/// - Java: parquet/src/main/java/org/apache/iceberg/parquet/ReadConf.java (field ID resolution)
-<a href=#224 id=224 data-nosnippet>224</a>/// - Java: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java (partition constants)
-<a href=#225 id=225 data-nosnippet>225</a></span><span class="attr">#[derive(Debug)]
-<a href=#226 id=226 data-nosnippet>226</a></span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">struct </span>RecordBatchTransformer {
-<a href=#227 id=227 data-nosnippet>227</a> snapshot_schema: Arc<IcebergSchema>,
-<a href=#228 id=228 data-nosnippet>228</a> projected_iceberg_field_ids: Vec<i32>,
-<a href=#229 id=229 data-nosnippet>229</a>
-<a href=#230 id=230 data-nosnippet>230</a> <span class="doccomment">/// Partition spec for identifying identity-transformed partition columns (spec rule #1).
-<a href=#231 id=231 data-nosnippet>231</a> /// Only fields with identity transforms use partition data constants; non-identity transforms
-<a href=#232 id=232 data-nosnippet>232</a> /// (bucket, truncate, etc.) must read source columns from data files.
-<a href=#233 id=233 data-nosnippet>233</a> </span>partition_spec: <span class="prelude-ty">Option</span><Arc<PartitionSpec>>,
-<a href=#234 id=234 data-nosnippet>234</a>
-<a href=#235 id=235 data-nosnippet>235</a> <span class="doccomment">/// Partition data providing constant values for identity-transformed partition columns (spec rule #1).
-<a href=#236 id=236 data-nosnippet>236</a> /// For example, in a file at path `dept=engineering/file.parquet`, this would contain
-<a href=#237 id=237 data-nosnippet>237</a> /// the value "engineering" for the dept field.
-<a href=#238 id=238 data-nosnippet>238</a> </span>partition_data: <span class="prelude-ty">Option</span><Struct>,
-<a href=#239 id=239 data-nosnippet>239</a>
-<a href=#240 id=240 data-nosnippet>240</a> <span class="comment">// BatchTransform gets lazily constructed based on the schema of
-<a href=#241 id=241 data-nosnippet>241</a> // the first RecordBatch we receive from the file
-<a href=#242 id=242 data-nosnippet>242</a> </span>batch_transform: <span class="prelude-ty">Option</span><BatchTransform>,
-<a href=#243 id=243 data-nosnippet>243</a>}
-<a href=#244 id=244 data-nosnippet>244</a>
-<a href=#245 id=245 data-nosnippet>245</a><span class="kw">impl </span>RecordBatchTransformer {
-<a href=#246 id=246 data-nosnippet>246</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>process_record_batch(
-<a href=#247 id=247 data-nosnippet>247</a> <span class="kw-2">&mut </span><span class="self">self</span>,
-<a href=#248 id=248 data-nosnippet>248</a> record_batch: RecordBatch,
-<a href=#249 id=249 data-nosnippet>249</a> ) -> <span class="prelude-ty">Result</span><RecordBatch> {
-<a href=#250 id=250 data-nosnippet>250</a> <span class="prelude-val">Ok</span>(<span class="kw">match </span><span class="kw-2">&</span><span class="self">self</span>.batch_transform {
-<a href=#251 id=251 data-nosnippet>251</a> <span class="prelude-val">Some</span>(BatchTransform::PassThrough) => record_batch,
-<a href=#252 id=252 data-nosnippet>252</a> <span class="prelude-val">Some</span>(BatchTransform::Modify {
-<a href=#253 id=253 data-nosnippet>253</a> target_schema,
-<a href=#254 id=254 data-nosnippet>254</a> operations,
-<a href=#255 id=255 data-nosnippet>255</a> }) => {
-<a href=#256 id=256 data-nosnippet>256</a> <span class="kw">let </span>options = RecordBatchOptions::default()
-<a href=#257 id=257 data-nosnippet>257</a> .with_match_field_names(<span class="bool-val">false</span>)
-<a href=#258 id=258 data-nosnippet>258</a> .with_row_count(<span class="prelude-val">Some</span>(record_batch.num_rows()));
-<a href=#259 id=259 data-nosnippet>259</a> RecordBatch::try_new_with_options(
-<a href=#260 id=260 data-nosnippet>260</a> Arc::clone(target_schema),
-<a href=#261 id=261 data-nosnippet>261</a> <span class="self">self</span>.transform_columns(record_batch.columns(), operations)<span class="question-mark">?</span>,
-<a href=#262 id=262 data-nosnippet>262</a> <span class="kw-2">&</span>options,
-<a href=#263 id=263 data-nosnippet>263</a> )<span class="question-mark">?
-<a href=#264 id=264 data-nosnippet>264</a> </span>}
-<a href=#265 id=265 data-nosnippet>265</a> <span class="prelude-val">Some</span>(BatchTransform::ModifySchema { target_schema }) => {
-<a href=#266 id=266 data-nosnippet>266</a> <span class="kw">let </span>options = RecordBatchOptions::default()
-<a href=#267 id=267 data-nosnippet>267</a> .with_match_field_names(<span class="bool-val">false</span>)
-<a href=#268 id=268 data-nosnippet>268</a> .with_row_count(<span class="prelude-val">Some</span>(record_batch.num_rows()));
-<a href=#269 id=269 data-nosnippet>269</a> RecordBatch::try_new_with_options(
-<a href=#270 id=270 data-nosnippet>270</a> Arc::clone(target_schema),
-<a href=#271 id=271 data-nosnippet>271</a> record_batch.columns().to_vec(),
-<a href=#272 id=272 data-nosnippet>272</a> <span class="kw-2">&</span>options,
-<a href=#273 id=273 data-nosnippet>273</a> )<span class="question-mark">?
-<a href=#274 id=274 data-nosnippet>274</a> </span>}
-<a href=#275 id=275 data-nosnippet>275</a> <span class="prelude-val">None </span>=> {
-<a href=#276 id=276 data-nosnippet>276</a> <span class="self">self</span>.batch_transform = <span class="prelude-val">Some</span>(<span class="self">Self</span>::generate_batch_transform(
-<a href=#277 id=277 data-nosnippet>277</a> record_batch.schema_ref(),
-<a href=#278 id=278 data-nosnippet>278</a> <span class="self">self</span>.snapshot_schema.as_ref(),
-<a href=#279 id=279 data-nosnippet>279</a> <span class="kw-2">&</span><span class="self">self</span>.projected_iceberg_field_ids,
-<a href=#280 id=280 data-nosnippet>280</a> <span class="self">self</span>.partition_spec.as_ref().map(|s| s.as_ref()),
-<a href=#281 id=281 data-nosnippet>281</a> <span class="self">self</span>.partition_data.as_ref(),
-<a href=#282 id=282 data-nosnippet>282</a> )<span class="question-mark">?</span>);
-<a href=#283 id=283 data-nosnippet>283</a>
-<a href=#284 id=284 data-nosnippet>284</a> <span class="self">self</span>.process_record_batch(record_batch)<span class="question-mark">?
-<a href=#285 id=285 data-nosnippet>285</a> </span>}
-<a href=#286 id=286 data-nosnippet>286</a> })
-<a href=#287 id=287 data-nosnippet>287</a> }
-<a href=#288 id=288 data-nosnippet>288</a>
-<a href=#289 id=289 data-nosnippet>289</a> <span class="comment">// Compare the schema of the incoming RecordBatches to the schema of
-<a href=#290 id=290 data-nosnippet>290</a> // the Iceberg snapshot to determine what, if any, transformation
-<a href=#291 id=291 data-nosnippet>291</a> // needs to be applied. If the schemas match, we return BatchTransform::PassThrough
-<a href=#292 id=292 data-nosnippet>292</a> // to indicate that no changes need to be made. Otherwise, we return a
-<a href=#293 id=293 data-nosnippet>293</a> // BatchTransform::Modify containing the target RecordBatch schema and
-<a href=#294 id=294 data-nosnippet>294</a> // the list of `ColumnSource`s that indicate how to source each column in
-<a href=#295 id=295 data-nosnippet>295</a> // the resulting RecordBatches.
-<a href=#296 id=296 data-nosnippet>296</a> </span><span class="kw">fn </span>generate_batch_transform(
-<a href=#297 id=297 data-nosnippet>297</a> source_schema: <span class="kw-2">&</span>ArrowSchemaRef,
-<a href=#298 id=298 data-nosnippet>298</a> snapshot_schema: <span class="kw-2">&</span>IcebergSchema,
-<a href=#299 id=299 data-nosnippet>299</a> projected_iceberg_field_ids: <span class="kw-2">&</span>[i32],
-<a href=#300 id=300 data-nosnippet>300</a> partition_spec: <span class="prelude-ty">Option</span><<span class="kw-2">&</span>PartitionSpec>,
-<a href=#301 id=301 data-nosnippet>301</a> partition_data: <span class="prelude-ty">Option</span><<span class="kw-2">&</span>Struct>,
-<a href=#302 id=302 data-nosnippet>302</a> ) -> <span class="prelude-ty">Result</span><BatchTransform> {
-<a href=#303 id=303 data-nosnippet>303</a> <span class="kw">let </span>mapped_unprojected_arrow_schema = Arc::new(schema_to_arrow_schema(snapshot_schema)<span class="question-mark">?</span>);
-<a href=#304 id=304 data-nosnippet>304</a> <span class="kw">let </span>field_id_to_mapped_schema_map =
-<a href=#305 id=305 data-nosnippet>305</a> <span class="self">Self</span>::build_field_id_to_arrow_schema_map(<span class="kw-2">&</span>mapped_unprojected_arrow_schema)<span class="question-mark">?</span>;
-<a href=#306 id=306 data-nosnippet>306</a>
-<a href=#307 id=307 data-nosnippet>307</a> <span class="comment">// Create a new arrow schema by selecting fields from mapped_unprojected,
-<a href=#308 id=308 data-nosnippet>308</a> // in the order of the field ids in projected_iceberg_field_ids
-<a href=#309 id=309 data-nosnippet>309</a> </span><span class="kw">let </span>fields: <span class="prelude-ty">Result</span><Vec<<span class="kw">_</span>>> = projected_iceberg_field_ids
-<a href=#310 id=310 data-nosnippet>310</a> .iter()
-<a href=#311 id=311 data-nosnippet>311</a> .map(|field_id| {
-<a href=#312 id=312 data-nosnippet>312</a> <span class="prelude-val">Ok</span>(field_id_to_mapped_schema_map
-<a href=#313 id=313 data-nosnippet>313</a> .get(field_id)
-<a href=#314 id=314 data-nosnippet>314</a> .ok_or(Error::new(ErrorKind::Unexpected, <span class="string">"field not found"</span>))<span class="question-mark">?
-<a href=#315 id=315 data-nosnippet>315</a> </span>.<span class="number">0
-<a href=#316 id=316 data-nosnippet>316</a> </span>.clone())
-<a href=#317 id=317 data-nosnippet>317</a> })
-<a href=#318 id=318 data-nosnippet>318</a> .collect();
-<a href=#319 id=319 data-nosnippet>319</a>
-<a href=#320 id=320 data-nosnippet>320</a> <span class="kw">let </span>target_schema = Arc::new(ArrowSchema::new(fields<span class="question-mark">?</span>));
+<a href=#216 id=216 data-nosnippet>216</a>/// # Partition Spec and Data
+<a href=#217 id=217 data-nosnippet>217</a>///
+<a href=#218 id=218 data-nosnippet>218</a>/// **Bucket partitioning**: Distinguish identity transforms (use partition metadata constants)
+<a href=#219 id=219 data-nosnippet>219</a>/// from non-identity transforms like bucket (read from data file) to enable runtime filtering on
+<a href=#220 id=220 data-nosnippet>220</a>/// bucket-partitioned columns. For example, `bucket(4, id)` stores only the bucket number in
+<a href=#221 id=221 data-nosnippet>221</a>/// partition metadata, so actual `id` values must be read from the data file.
+<a href=#222 id=222 data-nosnippet>222</a>///
+<a href=#223 id=223 data-nosnippet>223</a>/// # References
+<a href=#224 id=224 data-nosnippet>224</a>/// - Spec: https://iceberg.apache.org/spec/#column-projection
+<a href=#225 id=225 data-nosnippet>225</a>/// - Java: parquet/src/main/java/org/apache/iceberg/parquet/ReadConf.java (field ID resolution)
+<a href=#226 id=226 data-nosnippet>226</a>/// - Java: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java (partition constants)
+<a href=#227 id=227 data-nosnippet>227</a></span><span class="attr">#[derive(Debug)]
+<a href=#228 id=228 data-nosnippet>228</a></span><span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">struct </span>RecordBatchTransformer {
+<a href=#229 id=229 data-nosnippet>229</a> snapshot_schema: Arc<IcebergSchema>,
+<a href=#230 id=230 data-nosnippet>230</a> projected_iceberg_field_ids: Vec<i32>,
+<a href=#231 id=231 data-nosnippet>231</a>
+<a href=#232 id=232 data-nosnippet>232</a> <span class="doccomment">/// Partition spec for identifying identity-transformed partition columns (spec rule #1).
+<a href=#233 id=233 data-nosnippet>233</a> /// Only fields with identity transforms use partition data constants; non-identity transforms
+<a href=#234 id=234 data-nosnippet>234</a> /// (bucket, truncate, etc.) must read source columns from data files.
+<a href=#235 id=235 data-nosnippet>235</a> </span>partition_spec: <span class="prelude-ty">Option</span><Arc<PartitionSpec>>,
+<a href=#236 id=236 data-nosnippet>236</a>
+<a href=#237 id=237 data-nosnippet>237</a> <span class="doccomment">/// Partition data providing constant values for identity-transformed partition columns (spec rule #1).
+<a href=#238 id=238 data-nosnippet>238</a> /// For example, in a file at path `dept=engineering/file.parquet`, this would contain
+<a href=#239 id=239 data-nosnippet>239</a> /// the value "engineering" for the dept field.
+<a href=#240 id=240 data-nosnippet>240</a> </span>partition_data: <span class="prelude-ty">Option</span><Struct>,
+<a href=#241 id=241 data-nosnippet>241</a>
+<a href=#242 id=242 data-nosnippet>242</a> <span class="comment">// BatchTransform gets lazily constructed based on the schema of
+<a href=#243 id=243 data-nosnippet>243</a> // the first RecordBatch we receive from the file
+<a href=#244 id=244 data-nosnippet>244</a> </span>batch_transform: <span class="prelude-ty">Option</span><BatchTransform>,
+<a href=#245 id=245 data-nosnippet>245</a>}
+<a href=#246 id=246 data-nosnippet>246</a>
+<a href=#247 id=247 data-nosnippet>247</a><span class="kw">impl </span>RecordBatchTransformer {
+<a href=#248 id=248 data-nosnippet>248</a> <span class="kw">pub</span>(<span class="kw">crate</span>) <span class="kw">fn </span>process_record_batch(
+<a href=#249 id=249 data-nosnippet>249</a> <span class="kw-2">&mut </span><span class="self">self</span>,
+<a href=#250 id=250 data-nosnippet>250</a> record_batch: RecordBatch,
+<a href=#251 id=251 data-nosnippet>251</a> ) -> <span class="prelude-ty">Result</span><RecordBatch> {
+<a href=#252 id=252 data-nosnippet>252</a> <span class="prelude-val">Ok</span>(<span class="kw">match </span><span class="kw-2">&</span><span class="self">self</span>.batch_transform {
+<a href=#253 id=253 data-nosnippet>253</a> <span class="prelude-val">Some</span>(BatchTransform::PassThrough) => record_batch,
+<a href=#254 id=254 data-nosnippet>254</a> <span class="prelude-val">Some</span>(BatchTransform::Modify {
+<a href=#255 id=255 data-nosnippet>255</a> target_schema,
+<a href=#256 id=256 data-nosnippet>256</a> operations,
+<a href=#257 id=257 data-nosnippet>257</a> }) => {
+<a href=#258 id=258 data-nosnippet>258</a> <span class="kw">let </span>options = RecordBatchOptions::default()
+<a href=#259 id=259 data-nosnippet>259</a> .with_match_field_names(<span class="bool-val">false</span>)
+<a href=#260 id=260 data-nosnippet>260</a> .with_row_count(<span class="prelude-val">Some</span>(record_batch.num_rows()));
+<a href=#261 id=261 data-nosnippet>261</a> RecordBatch::try_new_with_options(
+<a href=#262 id=262 data-nosnippet>262</a> Arc::clone(target_schema),
+<a href=#263 id=263 data-nosnippet>263</a> <span class="self">self</span>.transform_columns(record_batch.columns(), operations)<span class="question-mark">?</span>,
+<a href=#264 id=264 data-nosnippet>264</a> <span class="kw-2">&</span>options,
+<a href=#265 id=265 data-nosnippet>265</a> )<span class="question-mark">?
+<a href=#266 id=266 data-nosnippet>266</a> </span>}
+<a href=#267 id=267 data-nosnippet>267</a> <span class="prelude-val">Some</span>(BatchTransform::ModifySchema { target_schema }) => {
+<a href=#268 id=268 data-nosnippet>268</a> <span class="kw">let </span>options = RecordBatchOptions::default()
+<a href=#269 id=269 data-nosnippet>269</a> .with_match_field_names(<span class="bool-val">false</span>)
+<a href=#270 id=270 data-nosnippet>270</a> .with_row_count(<span class="prelude-val">Some</span>(record_batch.num_rows()));
+<a href=#271 id=271 data-nosnippet>271</a> RecordBatch::try_new_with_options(
+<a href=#272 id=272 data-nosnippet>272</a> Arc::clone(target_schema),
+<a href=#273 id=273 data-nosnippet>273</a> record_batch.columns().to_vec(),
+<a href=#274 id=274 data-nosnippet>274</a> <span class="kw-2">&</span>options,
+<a href=#275 id=275 data-nosnippet>275</a> )<span class="question-mark">?
+<a href=#276 id=276 data-nosnippet>276</a> </span>}
+<a href=#277 id=277 data-nosnippet>277</a> <span class="prelude-val">None </span>=> {
+<a href=#278 id=278 data-nosnippet>278</a> <span class="self">self</span>.batch_transform = <span class="prelude-val">Some</span>(<span class="self">Self</span>::generate_batch_transform(
+<a href=#279 id=279 data-nosnippet>279</a> record_batch.schema_ref(),
+<a href=#280 id=280 data-nosnippet>280</a> <span class="self">self</span>.snapshot_schema.as_ref(),
+<a href=#281 id=281 data-nosnippet>281</a> <span class="kw-2">&</span><span class="self">self</span>.projected_iceberg_field_ids,
+<a href=#282 id=282 data-nosnippet>282</a> <span class="self">self</span>.partition_spec.as_ref().map(|s| s.as_ref()),
+<a href=#283 id=283 data-nosnippet>283</a> <span class="self">self</span>.partition_data.as_ref(),
+<a href=#284 id=284 data-nosnippet>284</a> )<span class="question-mark">?</span>);
+<a href=#285 id=285 data-nosnippet>285</a>
+<a href=#286 id=286 data-nosnippet>286</a> <span class="self">self</span>.process_record_batch(record_batch)<span class="question-mark">?
+<a href=#287 id=287 data-nosnippet>287</a> </span>}
+<a href=#288 id=288 data-nosnippet>288</a> })
+<a href=#289 id=289 data-nosnippet>289</a> }
+<a href=#290 id=290 data-nosnippet>290</a>
+<a href=#291 id=291 data-nosnippet>291</a> <span class="comment">// Compare the schema of the incoming RecordBatches to the schema of
+<a href=#292 id=292 data-nosnippet>292</a> // the Iceberg snapshot to determine what, if any, transformation
+<a href=#293 id=293 data-nosnippet>293</a> // needs to be applied. If the schemas match, we return BatchTransform::PassThrough
+<a href=#294 id=294 data-nosnippet>294</a> // to indicate that no changes need to be made. Otherwise, we return a
+<a href=#295 id=295 data-nosnippet>295</a> // BatchTransform::Modify containing the target RecordBatch schema and
+<a href=#296 id=296 data-nosnippet>296</a> // the list of `ColumnSource`s that indicate how to source each column in
+<a href=#297 id=297 data-nosnippet>297</a> // the resulting RecordBatches.
+<a href=#298 id=298 data-nosnippet>298</a> </span><span class="kw">fn </span>generate_batch_transform(
+<a href=#299 id=299 data-nosnippet>299</a> source_schema: <span class="kw-2">&</span>ArrowSchemaRef,
+<a href=#300 id=300 data-nosnippet>300</a> snapshot_schema: <span class="kw-2">&</span>IcebergSchema,
+<a href=#301 id=301 data-nosnippet>301</a> projected_iceberg_field_ids: <span class="kw-2">&</span>[i32],
+<a href=#302 id=302 data-nosnippet>302</a> partition_spec: <span class="prelude-ty">Option</span><<span class="kw-2">&</span>PartitionSpec>,
+<a href=#303 id=303 data-nosnippet>303</a> partition_data: <span class="prelude-ty">Option</span><<span class="kw-2">&</span>Struct>,
+<a href=#304 id=304 data-nosnippet>304</a> ) -> <span class="prelude-ty">Result</span><BatchTransform> {
+<a href=#305 id=305 data-nosnippet>305</a> <span class="kw">let </span>mapped_unprojected_arrow_schema = Arc::new(schema_to_arrow_schema(snapshot_schema)<span class="question-mark">?</span>);
+<a href=#306 id=306 data-nosnippet>306</a> <span class="kw">let </span>field_id_to_mapped_schema_map =
+<a href=#307 id=307 data-nosnippet>307</a> <span class="self">Self</span>::build_field_id_to_arrow_schema_map(<span class="kw-2">&</span>mapped_unprojected_arrow_schema)<span class="question-mark">?</span>;
+<a href=#308 id=308 data-nosnippet>308</a>
+<a href=#309 id=309 data-nosnippet>309</a> <span class="comment">// Create a new arrow schema by selecting fields from mapped_unprojected,
+<a href=#310 id=310 data-nosnippet>310</a> // in the order of the field ids in projected_iceberg_field_ids
+<a href=#311 id=311 data-nosnippet>311</a> </span><span class="kw">let </span>fields: <span class="prelude-ty">Result</span><Vec<<span class="kw">_</span>>> = projected_iceberg_field_ids
+<a href=#312 id=312 data-nosnippet>312</a> .iter()
+<a href=#313 id=313 data-nosnippet>313</a> .map(|field_id| {
+<a href=#314 id=314 data-nosnippet>314</a> <span class="prelude-val">Ok</span>(field_id_to_mapped_schema_map
+<a href=#315 id=315 data-nosnippet>315</a> .get(field_id)
+<a href=#316 id=316 data-nosnippet>316</a> .ok_or(Error::new(ErrorKind::Unexpected, <span class="string">"field not found"</span>))<span class="question-mark">?
+<a href=#317 id=317 data-nosnippet>317</a> </span>.<span class="number">0
+<a href=#318 id=318 data-nosnippet>318</a> </span>.clone())
+<a href=#319 id=319 data-nosnippet>319</a> })
+<a href=#320 id=320 data-nosnippet>320</a> .collect();
<a href=#321 id=321 data-nosnippet>321</a>
-<a href=#322 id=322 data-nosnippet>322</a> <span class="kw">let </span>constants_map = <span class="kw">if let </span>(<span class="prelude-val">Some</span>(spec), <span class="prelude-val">Some</span>(data)) = (partition_spec, partition_data) {
-<a href=#323 id=323 data-nosnippet>323</a> constants_map(spec, data)
-<a href=#324 id=324 data-nosnippet>324</a> } <span class="kw">else </span>{
-<a href=#325 id=325 data-nosnippet>325</a> HashMap::new()
-<a href=#326 id=326 data-nosnippet>326</a> };
-<a href=#327 id=327 data-nosnippet>327</a>
-<a href=#328 id=328 data-nosnippet>328</a> <span class="kw">match </span><span class="self">Self</span>::compare_schemas(source_schema, <span class="kw-2">&</span>target_schema) {
-<a href=#329 id=329 data-nosnippet>329</a> SchemaComparison::Equivalent => <span class="prelude-val">Ok</span>(BatchTransform::PassThrough),
-<a href=#330 id=330 data-nosnippet>330</a> SchemaComparison::NameChangesOnly => <span class="prelude-val">Ok</span>(BatchTransform::ModifySchema { target_schema }),
-<a href=#331 id=331 data-nosnippet>331</a> SchemaComparison::Different => <span class="prelude-val">Ok</span>(BatchTransform::Modify {
-<a href=#332 id=332 data-nosnippet>332</a> operations: <span class="self">Self</span>::generate_transform_operations(
-<a href=#333 id=333 data-nosnippet>333</a> source_schema,
-<a href=#334 id=334 data-nosnippet>334</a> snapshot_schema,
-<a href=#335 id=335 data-nosnippet>335</a> projected_iceberg_field_ids,
-<a href=#336 id=336 data-nosnippet>336</a> field_id_to_mapped_schema_map,
-<a href=#337 id=337 data-nosnippet>337</a> constants_map,
-<a href=#338 id=338 data-nosnippet>338</a> partition_spec,
-<a href=#339 id=339 data-nosnippet>339</a> )<span class="question-mark">?</span>,
-<a href=#340 id=340 data-nosnippet>340</a> target_schema,
-<a href=#341 id=341 data-nosnippet>341</a> }),
-<a href=#342 id=342 data-nosnippet>342</a> }
-<a href=#343 id=343 data-nosnippet>343</a> }
-<a href=#344 id=344 data-nosnippet>344</a>
-<a href=#345 id=345 data-nosnippet>345</a> <span class="doccomment">/// Compares the source and target schemas
-<a href=#346 id=346 data-nosnippet>346</a> /// Determines if they have changed in any meaningful way:
-<a href=#347 id=347 data-nosnippet>347</a> /// * If they have different numbers of fields, then we need to modify
-<a href=#348 id=348 data-nosnippet>348</a> /// the incoming RecordBatch schema AND columns
-<a href=#349 id=349 data-nosnippet>349</a> /// * If they have the same number of fields, but some of them differ in
-<a href=#350 id=350 data-nosnippet>350</a> /// either data type or nullability, then we need to modify the
-<a href=#351 id=351 data-nosnippet>351</a> /// incoming RecordBatch schema AND columns
-<a href=#352 id=352 data-nosnippet>352</a> /// * If the schemas differ only in the column names, then we need
-<a href=#353 id=353 data-nosnippet>353</a> /// to modify the RecordBatch schema BUT we can keep the
-<a href=#354 id=354 data-nosnippet>354</a> /// original column data unmodified
-<a href=#355 id=355 data-nosnippet>355</a> /// * If the schemas are identical (or differ only in inconsequential
-<a href=#356 id=356 data-nosnippet>356</a> /// ways) then we can pass through the original RecordBatch unmodified
-<a href=#357 id=357 data-nosnippet>357</a> </span><span class="kw">fn </span>compare_schemas(
-<a href=#358 id=358 data-nosnippet>358</a> source_schema: <span class="kw-2">&</span>ArrowSchemaRef,
-<a href=#359 id=359 data-nosnippet>359</a> target_schema: <span class="kw-2">&</span>ArrowSchemaRef,
-<a href=#360 id=360 data-nosnippet>360</a> ) -> SchemaComparison {
-<a href=#361 id=361 data-nosnippet>361</a> <span class="kw">if </span>source_schema.fields().len() != target_schema.fields().len() {
-<a href=#362 id=362 data-nosnippet>362</a> <span class="kw">return </span>SchemaComparison::Different;
-<a href=#363 id=363 data-nosnippet>363</a> }
-<a href=#364 id=364 data-nosnippet>364</a>
-<a href=#365 id=365 data-nosnippet>365</a> <span class="kw">let </span><span class="kw-2">mut </span>names_changed = <span class="bool-val">false</span>;
+<a href=#322 id=322 data-nosnippet>322</a> <span class="kw">let </span>target_schema = Arc::new(ArrowSchema::new(fields<span class="question-mark">?</span>));
+<a href=#323 id=323 data-nosnippet>323</a>
+<a href=#324 id=324 data-nosnippet>324</a> <span class="kw">let </span>constants_map = <span class="kw">if let </span>(<span class="prelude-val">Some</span>(spec), <span class="prelude-val">Some</span>(data)) = (partition_spec, partition_data) {
+<a href=#325 id=325 data-nosnippet>325</a> constants_map(spec, data)
+<a href=#326 id=326 data-nosnippet>326</a> } <span class="kw">else </span>{
+<a href=#327 id=327 data-nosnippet>327</a> HashMap::new()
+<a href=#328 id=328 data-nosnippet>328</a> };
+<a href=#329 id=329 data-nosnippet>329</a>
+<a href=#330 id=330 data-nosnippet>330</a> <span class="kw">match </span><span class="self">Self</span>::compare_schemas(source_schema, <span class="kw-2">&</span>target_schema) {
+<a href=#331 id=331 data-nosnippet>331</a> SchemaComparison::Equivalent => <span class="prelude-val">Ok</span>(BatchTransform::PassThrough),
+<a href=#332 id=332 data-nosnippet>332</a> SchemaComparison::NameChangesOnly => <span class="prelude-val">Ok</span>(BatchTransform::ModifySchema { target_schema }),
+<a href=#333 id=333 data-nosnippet>333</a> SchemaComparison::Different => <span class="prelude-val">Ok</span>(BatchTransform::Modify {
+<a href=#334 id=334 data-nosnippet>334</a> operations: <span class="self">Self</span>::generate_transform_operations(
+<a href=#335 id=335 data-nosnippet>335</a> source_schema,
+<a href=#336 id=336 data-nosnippet>336</a> snapshot_schema,
+<a href=#337 id=337 data-nosnippet>337</a> projected_iceberg_field_ids,
+<a href=#338 id=338 data-nosnippet>338</a> field_id_to_mapped_schema_map,
+<a href=#339 id=339 data-nosnippet>339</a> constants_map,
+<a href=#340 id=340 data-nosnippet>340</a> partition_spec,
+<a href=#341 id=341 data-nosnippet>341</a> )<span class="question-mark">?</span>,
+<a href=#342 id=342 data-nosnippet>342</a> target_schema,
+<a href=#343 id=343 data-nosnippet>343</a> }),
+<a href=#344 id=344 data-nosnippet>344</a> }
+<a href=#345 id=345 data-nosnippet>345</a> }
+<a href=#346 id=346 data-nosnippet>346</a>
+<a href=#347 id=347 data-nosnippet>347</a> <span class="doccomment">/// Compares the source and target schemas
+<a href=#348 id=348 data-nosnippet>348</a> /// Determines if they have changed in any meaningful way:
+<a href=#349 id=349 data-nosnippet>349</a> /// * If they have different numbers of fields, then we need to modify
+<a href=#350 id=350 data-nosnippet>350</a> /// the incoming RecordBatch schema AND columns
+<a href=#351 id=351 data-nosnippet>351</a> /// * If they have the same number of fields, but some of them differ in
+<a href=#352 id=352 data-nosnippet>352</a> /// either data type or nullability, then we need to modify the
+<a href=#353 id=353 data-nosnippet>353</a> /// incoming RecordBatch schema AND columns
+<a href=#354 id=354 data-nosnippet>354</a> /// * If the schemas differ only in the column names, then we need
+<a href=#355 id=355 data-nosnippet>355</a> /// to modify the RecordBatch schema BUT we can keep the
+<a href=#356 id=356 data-nosnippet>356</a> /// original column data unmodified
+<a href=#357 id=357 data-nosnippet>357</a> /// * If the schemas are identical (or differ only in inconsequential
+<a href=#358 id=358 data-nosnippet>358</a> /// ways) then we can pass through the original RecordBatch unmodified
+<a href=#359 id=359 data-nosnippet>359</a> </span><span class="kw">fn </span>compare_schemas(
+<a href=#360 id=360 data-nosnippet>360</a> source_schema: <span class="kw-2">&</span>ArrowSchemaRef,
+<a href=#361 id=361 data-nosnippet>361</a> target_schema: <span class="kw-2">&</span>ArrowSchemaRef,
+<a href=#362 id=362 data-nosnippet>362</a> ) -> SchemaComparison {
+<a href=#363 id=363 data-nosnippet>363</a> <span class="kw">if </span>source_schema.fields().len() != target_schema.fields().len() {
+<a href=#364 id=364 data-nosnippet>364</a> <span class="kw">return </span>SchemaComparison::Different;
+<a href=#365 id=365 data-nosnippet>365</a> }
<a href=#366 id=366 data-nosnippet>366</a>
-<a href=#367 id=367 data-nosnippet>367</a> <span class="kw">for </span>(source_field, target_field) <span class="kw">in </span>source_schema
-<a href=#368 id=368 data-nosnippet>368</a> .fields()
-<a href=#369 id=369 data-nosnippet>369</a> .iter()
-<a href=#370 id=370 data-nosnippet>370</a> .zip(target_schema.fields().iter())
-<a href=#371 id=371 data-nosnippet>371</a> {
-<a href=#372 id=372 data-nosnippet>372</a> <span class="kw">if </span>source_field.data_type() != target_field.data_type()
-<a href=#373 id=373 data-nosnippet>373</a> || source_field.is_nullable() != target_field.is_nullable()
-<a href=#374 id=374 data-nosnippet>374</a> {
-<a href=#375 id=375 data-nosnippet>375</a> <span class="kw">return </span>SchemaComparison::Different;
-<a href=#376 id=376 data-nosnippet>376</a> }
-<a href=#377 id=377 data-nosnippet>377</a>
-<a href=#378 id=378 data-nosnippet>378</a> <span class="kw">if </span>source_field.name() != target_field.name() {
-<a href=#379 id=379 data-nosnippet>379</a> names_changed = <span class="bool-val">true</span>;
-<a href=#380 id=380 data-nosnippet>380</a> }
-<a href=#381 id=381 data-nosnippet>381</a> }
-<a href=#382 id=382 data-nosnippet>382</a>
-<a href=#383 id=383 data-nosnippet>383</a> <span class="kw">if </span>names_changed {
-<a href=#384 id=384 data-nosnippet>384</a> SchemaComparison::NameChangesOnly
-<a href=#385 id=385 data-nosnippet>385</a> } <span class="kw">else </span>{
-<a href=#386 id=386 data-nosnippet>386</a> SchemaComparison::Equivalent
-<a href=#387 id=387 data-nosnippet>387</a> }
-<a href=#388 id=388 data-nosnippet>388</a> }
-<a href=#389 id=389 data-nosnippet>389</a>
-<a href=#390 id=390 data-nosnippet>390</a> <span class="kw">fn </span>generate_transform_operations(
-<a href=#391 id=391 data-nosnippet>391</a> source_schema: <span class="kw-2">&</span>ArrowSchemaRef,
-<a href=#392 id=392 data-nosnippet>392</a> snapshot_schema: <span class="kw-2">&</span>IcebergSchema,
-<a href=#393 id=393 data-nosnippet>393</a> projected_iceberg_field_ids: <span class="kw-2">&</span>[i32],
-<a href=#394 id=394 data-nosnippet>394</a> field_id_to_mapped_schema_map: HashMap<i32, (FieldRef, usize)>,
-<a href=#395 id=395 data-nosnippet>395</a> constants_map: HashMap<i32, PrimitiveLiteral>,
-<a href=#396 id=396 data-nosnippet>396</a> _partition_spec: <span class="prelude-ty">Option</span><<span class="kw-2">&</span>PartitionSpec>,
-<a href=#397 id=397 data-nosnippet>397</a> ) -> <span class="prelude-ty">Result</span><Vec<ColumnSource>> {
-<a href=#398 id=398 data-nosnippet>398</a> <span class="kw">let </span>field_id_to_source_schema_map =
-<a href=#399 id=399 data-nosnippet>399</a> <span class="self">Self</span>::build_field_id_to_arrow_schema_map(source_schema)<span class="question-mark">?</span>;
-<a href=#400 id=400 data-nosnippet>400</a>
-<a href=#401 id=401 data-nosnippet>401</a> projected_iceberg_field_ids
-<a href=#402 id=402 data-nosnippet>402</a> .iter()
-<a href=#403 id=403 data-nosnippet>403</a> .map(|field_id| {
-<a href=#404 id=404 data-nosnippet>404</a> <span class="kw">let </span>(target_field, <span class="kw">_</span>) =
-<a href=#405 id=405 data-nosnippet>405</a> field_id_to_mapped_schema_map
-<a href=#406 id=406 data-nosnippet>406</a> .get(field_id)
-<a href=#407 id=407 data-nosnippet>407</a> .ok_or(Error::new(
-<a href=#408 id=408 data-nosnippet>408</a> ErrorKind::Unexpected,
-<a href=#409 id=409 data-nosnippet>409</a> <span class="string">"could not find field in schema"</span>,
-<a href=#410 id=410 data-nosnippet>410</a> ))<span class="question-mark">?</span>;
-<a href=#411 id=411 data-nosnippet>411</a> <span class="kw">let </span>target_type = target_field.data_type();
-<a href=#412 id=412 data-nosnippet>412</a>
-<a href=#413 id=413 data-nosnippet>413</a> <span class="kw">let </span>iceberg_field = snapshot_schema.field_by_id(<span class="kw-2">*</span>field_id).ok_or(Error::new(
-<a href=#414 id=414 data-nosnippet>414</a> ErrorKind::Unexpected,
-<a href=#415 id=415 data-nosnippet>415</a> <span class="string">"Field not found in snapshot schema"</span>,
-<a href=#416 id=416 data-nosnippet>416</a> ))<span class="question-mark">?</span>;
-<a href=#417 id=417 data-nosnippet>417</a>
-<a href=#418 id=418 data-nosnippet>418</a> <span class="comment">// Iceberg spec's "Column Projection" rules (https://iceberg.apache.org/spec/#column-projection).
-<a href=#419 id=419 data-nosnippet>419</a> // For fields "not present" in data files:
-<a href=#420 id=420 data-nosnippet>420</a> // 1. Use partition metadata (identity transforms only)
-<a href=#421 id=421 data-nosnippet>421</a> // 2. Use name mapping
-<a href=#422 id=422 data-nosnippet>422</a> // 3. Use initial_default
-<a href=#423 id=423 data-nosnippet>423</a> // 4. Return null
-<a href=#424 id=424 data-nosnippet>424</a> //
-<a href=#425 id=425 data-nosnippet>425</a> // Why check partition constants before Parquet field IDs (Java: BaseParquetReaders.java:299):
-<a href=#426 id=426 data-nosnippet>426</a> // In add_files scenarios, partition columns may exist in BOTH Parquet AND partition metadata.
-<a href=#427 id=427 data-nosnippet>427</a> // Partition metadata is authoritative - it defines which partition this file belongs to.
-<a href=#428 id=428 data-nosnippet>428</a>
-<a href=#429 id=429 data-nosnippet>429</a> // Field ID resolution now happens in ArrowReader via:
-<a href=#430 id=430 data-nosnippet>430</a> // 1. Embedded field IDs (ParquetSchemaUtil.hasIds() = true) - trust them
-<a href=#431 id=431 data-nosnippet>431</a> // 2. Name mapping (ParquetSchemaUtil.applyNameMapping()) - applied upfront
-<a href=#432 id=432 data-nosnippet>432</a> // 3. Position-based fallback (ParquetSchemaUtil.addFallbackIds()) - applied upfront
-<a href=#433 id=433 data-nosnippet>433</a> //
-<a href=#434 id=434 data-nosnippet>434</a> // At this point, all field IDs in the source schema are trustworthy.
-<a href=#435 id=435 data-nosnippet>435</a> // No conflict detection needed - schema resolution happened in reader.rs.
-<a href=#436 id=436 data-nosnippet>436</a> </span><span class="kw">let </span>field_by_id = field_id_to_source_schema_map.get(field_id).map(
-<a href=#437 id=437 data-nosnippet>437</a> |(source_field, source_index)| {
-<a href=#438 id=438 data-nosnippet>438</a> <span class="kw">if </span>source_field.data_type().equals_datatype(target_type) {
-<a href=#439 id=439 data-nosnippet>439</a> ColumnSource::PassThrough {
-<a href=#440 id=440 data-nosnippet>440</a> source_index: <span class="kw-2">*</span>source_index,
-<a href=#441 id=441 data-nosnippet>441</a> }
-<a href=#442 id=442 data-nosnippet>442</a> } <span class="kw">else </span>{
-<a href=#443 id=443 data-nosnippet>443</a> ColumnSource::Promote {
-<a href=#444 id=444 data-nosnippet>444</a> target_type: target_type.clone(),
-<a href=#445 id=445 data-nosnippet>445</a> source_index: <span class="kw-2">*</span>source_index,
-<a href=#446 id=446 data-nosnippet>446</a> }
-<a href=#447 id=447 data-nosnippet>447</a> }
-<a href=#448 id=448 data-nosnippet>448</a> },
-<a href=#449 id=449 data-nosnippet>449</a> );
-<a href=#450 id=450 data-nosnippet>450</a>
-<a href=#451 id=451 data-nosnippet>451</a> <span class="comment">// Apply spec's fallback steps for "not present" fields.
-<a href=#452 id=452 data-nosnippet>452</a> </span><span class="kw">let </span>column_source = <span class="kw">if let </span><span class="prelude-val">Some</span>(constant_value) = constants_map.get(field_id) {
-<a href=#453 id=453 data-nosnippet>453</a> <span class="comment">// Rule #1: Identity partition constant
-<a href=#454 id=454 data-nosnippet>454</a> </span>ColumnSource::Add {
-<a href=#455 id=455 data-nosnippet>455</a> value: <span class="prelude-val">Some</span>(constant_value.clone()),
-<a href=#456 id=456 data-nosnippet>456</a> target_type: target_type.clone(),
-<a href=#457 id=457 data-nosnippet>457</a> }
-<a href=#458 id=458 data-nosnippet>458</a> } <span class="kw">else if let </span><span class="prelude-val">Some</span>(source) = field_by_id {
-<a href=#459 id=459 data-nosnippet>459</a> source
-<a href=#460 id=460 data-nosnippet>460</a> } <span class="kw">else </span>{
-<a href=#461 id=461 data-nosnippet>461</a> <span class="comment">// Rules #2, #3 and #4:
-<a href=#462 id=462 data-nosnippet>462</a> // Rule #2 (name mapping) was already applied in reader.rs if needed.
-<a href=#463 id=463 data-nosnippet>463</a> // If field_id is still not found, the column doesn't exist in the Parquet file.
-<a href=#464 id=464 data-nosnippet>464</a> // Fall through to rule #3 (initial_default) or rule #4 (null).
-<a href=#465 id=465 data-nosnippet>465</a> </span><span class="kw">let </span>default_value = iceberg_field.initial_default.as_ref().and_then(|lit| {
-<a href=#466 id=466 data-nosnippet>466</a> <span class="kw">if let </span>Literal::Primitive(prim) = lit {
-<a href=#467 id=467 data-nosnippet>467</a> <span class="prelude-val">Some</span>(prim.clone())
-<a href=#468 id=468 data-nosnippet>468</a> } <span class="kw">else </span>{
-<a href=#469 id=469 data-nosnippet>469</a> <span class="prelude-val">None
-<a href=#470 id=470 data-nosnippet>470</a> </span>}
-<a href=#471 id=471 data-nosnippet>471</a> });
-<a href=#472 id=472 data-nosnippet>472</a> ColumnSource::Add {
-<a href=#473 id=473 data-nosnippet>473</a> value: default_value,
-<a href=#474 id=474 data-nosnippet>474</a> target_type: target_type.clone(),
-<a href=#475 id=475 data-nosnippet>475</a> }
-<a href=#476 id=476 data-nosnippet>476</a> };
-<a href=#477 id=477 data-nosnippet>477</a>
-<a href=#478 id=478 data-nosnippet>478</a> <span class="prelude-val">Ok</span>(column_source)
-<a href=#479 id=479 data-nosnippet>479</a> })
-<a href=#480 id=480 data-nosnippet>480</a> .collect()
-<a href=#481 id=481 data-nosnippet>481</a> }
-<a href=#482 id=482 data-nosnippet>482</a>
-<a href=#483 id=483 data-nosnippet>483</a> <span class="kw">fn </span>build_field_id_to_arrow_schema_map(
-<a href=#484 id=484 data-nosnippet>484</a> source_schema: <span class="kw-2">&</span>SchemaRef,
-<a href=#485 id=485 data-nosnippet>485</a> ) -> <span class="prelude-ty">Result</span><HashMap<i32, (FieldRef, usize)>> {
-<a href=#486 id=486 data-nosnippet>486</a> <span class="kw">let </span><span class="kw-2">mut </span>field_id_to_source_schema = HashMap::new();
-<a href=#487 id=487 data-nosnippet>487</a> <span class="kw">for </span>(source_field_idx, source_field) <span class="kw">in </span>source_schema.fields.iter().enumerate() {
-<a href=#488 id=488 data-nosnippet>488</a> <span class="comment">// Check if field has a field ID in metadata
-<a href=#489 id=489 data-nosnippet>489</a> </span><span class="kw">if let </span><span class="prelude-val">Some</span>(field_id_str) = source_field.metadata().get(PARQUET_FIELD_ID_META_KEY) {
-<a href=#490 id=490 data-nosnippet>490</a> <span class="kw">let </span>this_field_id = field_id_str.parse().map_err(|e| {
-<a href=#491 id=491 data-nosnippet>491</a> Error::new(
-<a href=#492 id=492 data-nosnippet>492</a> ErrorKind::DataInvalid,
-<a href=#493 id=493 data-nosnippet>493</a> <span class="macro">format!</span>(<span class="string">"field id not parseable as an i32: {}"</span>, e),
-<a href=#494 id=494 data-nosnippet>494</a> )
-<a href=#495 id=495 data-nosnippet>495</a> })<span class="question-mark">?</span>;
-<a href=#496 id=496 data-nosnippet>496</a>
-<a href=#497 id=497 data-nosnippet>497</a> field_id_to_source_schema
-<a href=#498 id=498 data-nosnippet>498</a> .insert(this_field_id, (source_field.clone(), source_field_idx));
-<a href=#499 id=499 data-nosnippet>499</a> }
-<a href=#500 id=500 data-nosnippet>500</a> <span class="comment">// If field doesn't have a field ID, skip it - name mapping will handle it
-<a href=#501 id=501 data-nosnippet>501</a> </span>}
-<a href=#502 id=502 data-nosnippet>502</a>
-<a href=#503 id=503 data-nosnippet>503</a> <span class="prelude-val">Ok</span>(field_id_to_source_schema)
-<a href=#504 id=504 data-nosnippet>504</a> }
-<a href=#505 id=505 data-nosnippet>505</a>
-<a href=#506 id=506 data-nosnippet>506</a> <span class="kw">fn </span>transform_columns(
-<a href=#507 id=507 data-nosnippet>507</a> <span class="kw-2">&</span><span class="self">self</span>,
-<a href=#508 id=508 data-nosnippet>508</a> columns: <span class="kw-2">&</span>[Arc<<span class="kw">dyn </span>ArrowArray>],
-<a href=#509 id=509 data-nosnippet>509</a> operations: <span class="kw-2">&</span>[ColumnSource],
-<a href=#510 id=510 data-nosnippet>510</a> ) -> <span class="prelude-ty">Result</span><Vec<Arc<<span class="kw">dyn </span>ArrowArray>>> {
-<a href=#511 id=511 data-nosnippet>511</a> <span class="kw">if </span>columns.is_empty() {
-<a href=#512 id=512 data-nosnippet>512</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(columns.to_vec());
-<a href=#513 id=513 data-nosnippet>513</a> }
-<a href=#514 id=514 data-nosnippet>514</a> <span class="kw">let </span>num_rows = columns[<span class="number">0</span>].len();
-<a href=#515 id=515 data-nosnippet>515</a>
-<a href=#516 id=516 data-nosnippet>516</a> operations
-<a href=#517 id=517 data-nosnippet>517</a> .iter()
-<a href=#518 id=518 data-nosnippet>518</a> .map(|op| {
-<a href=#519 id=519 data-nosnippet>519</a> <span class="prelude-val">Ok</span>(<span class="kw">match </span>op {
-<a href=#520 id=520 data-nosnippet>520</a> ColumnSource::PassThrough { source_index } => columns[<span class="kw-2">*</span>source_index].clone(),
-<a href=#521 id=521 data-nosnippet>521</a>
-<a href=#522 id=522 data-nosnippet>522</a> ColumnSource::Promote {
-<a href=#523 id=523 data-nosnippet>523</a> target_type,
-<a href=#524 id=524 data-nosnippet>524</a> source_index,
-<a href=#525 id=525 data-nosnippet>525</a> } => cast(<span class="kw-2">&*</span>columns[<span class="kw-2">*</span>source_index], target_type)<span class="question-mark">?</span>,
-<a href=#526 id=526 data-nosnippet>526</a>
-<a href=#527 id=527 data-nosnippet>527</a> ColumnSource::Add { target_type, value } => {
-<a href=#528 id=528 data-nosnippet>528</a> <span class="self">Self</span>::create_column(target_type, value, num_rows)<span class="question-mark">?
-<a href=#529 id=529 data-nosnippet>529</a> </span>}
-<a href=#530 id=530 data-nosnippet>530</a> })
-<a href=#531 id=531 data-nosnippet>531</a> })
-<a href=#532 id=532 data-nosnippet>532</a> .collect()
-<a href=#533 id=533 data-nosnippet>533</a> }
-<a href=#534 id=534 data-nosnippet>534</a>
-<a href=#535 id=535 data-nosnippet>535</a> <span class="kw">fn </span>create_column(
-<a href=#536 id=536 data-nosnippet>536</a> target_type: <span class="kw-2">&</span>DataType,
-<a href=#537 id=537 data-nosnippet>537</a> prim_lit: <span class="kw-2">&</span><span class="prelude-ty">Option</span><PrimitiveLiteral>,
-<a href=#538 id=538 data-nosnippet>538</a> num_rows: usize,
-<a href=#539 id=539 data-nosnippet>539</a> ) -> <span class="prelude-ty">Result</span><ArrayRef> {
-<a href=#540 id=540 data-nosnippet>540</a> <span class="prelude-val">Ok</span>(<span class="kw">match </span>(target_type, prim_lit) {
-<a href=#541 id=541 data-nosnippet>541</a> (DataType::Boolean, <span class="prelude-val">Some</span>(PrimitiveLiteral::Boolean(value))) => {
-<a href=#542 id=542 data-nosnippet>542</a> Arc::new(BooleanArray::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
-<a href=#543 id=543 data-nosnippet>543</a> }
-<a href=#544 id=544 data-nosnippet>544</a> (DataType::Boolean, <span class="prelude-val">None</span>) => {
-<a href=#545 id=545 data-nosnippet>545</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><bool>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#546 id=546 data-nosnippet>546</a> Arc::new(BooleanArray::from(vals))
-<a href=#547 id=547 data-nosnippet>547</a> }
-<a href=#548 id=548 data-nosnippet>548</a> (DataType::Int32, <span class="prelude-val">Some</span>(PrimitiveLiteral::Int(value))) => {
-<a href=#549 id=549 data-nosnippet>549</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
-<a href=#550 id=550 data-nosnippet>550</a> }
-<a href=#551 id=551 data-nosnippet>551</a> (DataType::Int32, <span class="prelude-val">None</span>) => {
-<a href=#552 id=552 data-nosnippet>552</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><i32>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#553 id=553 data-nosnippet>553</a> Arc::new(Int32Array::from(vals))
-<a href=#554 id=554 data-nosnippet>554</a> }
-<a href=#555 id=555 data-nosnippet>555</a> (DataType::Date32, <span class="prelude-val">Some</span>(PrimitiveLiteral::Int(value))) => {
-<a href=#556 id=556 data-nosnippet>556</a> Arc::new(Date32Array::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
-<a href=#557 id=557 data-nosnippet>557</a> }
-<a href=#558 id=558 data-nosnippet>558</a> (DataType::Date32, <span class="prelude-val">None</span>) => {
-<a href=#559 id=559 data-nosnippet>559</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><i32>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#560 id=560 data-nosnippet>560</a> Arc::new(Date32Array::from(vals))
-<a href=#561 id=561 data-nosnippet>561</a> }
-<a href=#562 id=562 data-nosnippet>562</a> (DataType::Int64, <span class="prelude-val">Some</span>(PrimitiveLiteral::Long(value))) => {
-<a href=#563 id=563 data-nosnippet>563</a> Arc::new(Int64Array::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
-<a href=#564 id=564 data-nosnippet>564</a> }
-<a href=#565 id=565 data-nosnippet>565</a> (DataType::Int64, <span class="prelude-val">None</span>) => {
-<a href=#566 id=566 data-nosnippet>566</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><i64>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#567 id=567 data-nosnippet>567</a> Arc::new(Int64Array::from(vals))
-<a href=#568 id=568 data-nosnippet>568</a> }
-<a href=#569 id=569 data-nosnippet>569</a> (DataType::Float32, <span class="prelude-val">Some</span>(PrimitiveLiteral::Float(value))) => {
-<a href=#570 id=570 data-nosnippet>570</a> Arc::new(Float32Array::from(<span class="macro">vec!</span>[value.<span class="number">0</span>; num_rows]))
-<a href=#571 id=571 data-nosnippet>571</a> }
-<a href=#572 id=572 data-nosnippet>572</a> (DataType::Float32, <span class="prelude-val">None</span>) => {
-<a href=#573 id=573 data-nosnippet>573</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><f32>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#574 id=574 data-nosnippet>574</a> Arc::new(Float32Array::from(vals))
-<a href=#575 id=575 data-nosnippet>575</a> }
-<a href=#576 id=576 data-nosnippet>576</a> (DataType::Float64, <span class="prelude-val">Some</span>(PrimitiveLiteral::Double(value))) => {
-<a href=#577 id=577 data-nosnippet>577</a> Arc::new(Float64Array::from(<span class="macro">vec!</span>[value.<span class="number">0</span>; num_rows]))
-<a href=#578 id=578 data-nosnippet>578</a> }
-<a href=#579 id=579 data-nosnippet>579</a> (DataType::Float64, <span class="prelude-val">None</span>) => {
-<a href=#580 id=580 data-nosnippet>580</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><f64>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#581 id=581 data-nosnippet>581</a> Arc::new(Float64Array::from(vals))
-<a href=#582 id=582 data-nosnippet>582</a> }
-<a href=#583 id=583 data-nosnippet>583</a> (DataType::Utf8, <span class="prelude-val">Some</span>(PrimitiveLiteral::String(value))) => {
-<a href=#584 id=584 data-nosnippet>584</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[value.clone(); num_rows]))
-<a href=#585 id=585 data-nosnippet>585</a> }
-<a href=#586 id=586 data-nosnippet>586</a> (DataType::Utf8, <span class="prelude-val">None</span>) => {
-<a href=#587 id=587 data-nosnippet>587</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><String>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#588 id=588 data-nosnippet>588</a> Arc::new(StringArray::from(vals))
-<a href=#589 id=589 data-nosnippet>589</a> }
-<a href=#590 id=590 data-nosnippet>590</a> (DataType::Binary, <span class="prelude-val">Some</span>(PrimitiveLiteral::Binary(value))) => {
-<a href=#591 id=591 data-nosnippet>591</a> Arc::new(BinaryArray::from_vec(<span class="macro">vec!</span>[value; num_rows]))
-<a href=#592 id=592 data-nosnippet>592</a> }
-<a href=#593 id=593 data-nosnippet>593</a> (DataType::Binary, <span class="prelude-val">None</span>) => {
-<a href=#594 id=594 data-nosnippet>594</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><<span class="kw-2">&</span>[u8]>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
-<a href=#595 id=595 data-nosnippet>595</a> Arc::new(BinaryArray::from_opt_vec(vals))
-<a href=#596 id=596 data-nosnippet>596</a> }
-<a href=#597 id=597 data-nosnippet>597</a> (DataType::Null, <span class="kw">_</span>) => Arc::new(NullArray::new(num_rows)),
-<a href=#598 id=598 data-nosnippet>598</a> (dt, <span class="kw">_</span>) => {
-<a href=#599 id=599 data-nosnippet>599</a> <span class="kw">return </span><span class="prelude-val">Err</span>(Error::new(
-<a href=#600 id=600 data-nosnippet>600</a> ErrorKind::Unexpected,
-<a href=#601 id=601 data-nosnippet>601</a> <span class="macro">format!</span>(<span class="string">"unexpected target column type {}"</span>, dt),
-<a href=#602 id=602 data-nosnippet>602</a> ));
-<a href=#603 id=603 data-nosnippet>603</a> }
-<a href=#604 id=604 data-nosnippet>604</a> })
-<a href=#605 id=605 data-nosnippet>605</a> }
-<a href=#606 id=606 data-nosnippet>606</a>}
+<a href=#367 id=367 data-nosnippet>367</a> <span class="kw">let </span><span class="kw-2">mut </span>names_changed = <span class="bool-val">false</span>;
+<a href=#368 id=368 data-nosnippet>368</a>
+<a href=#369 id=369 data-nosnippet>369</a> <span class="kw">for </span>(source_field, target_field) <span class="kw">in </span>source_schema
+<a href=#370 id=370 data-nosnippet>370</a> .fields()
+<a href=#371 id=371 data-nosnippet>371</a> .iter()
+<a href=#372 id=372 data-nosnippet>372</a> .zip(target_schema.fields().iter())
+<a href=#373 id=373 data-nosnippet>373</a> {
+<a href=#374 id=374 data-nosnippet>374</a> <span class="kw">if </span>source_field.data_type() != target_field.data_type()
+<a href=#375 id=375 data-nosnippet>375</a> || source_field.is_nullable() != target_field.is_nullable()
+<a href=#376 id=376 data-nosnippet>376</a> {
+<a href=#377 id=377 data-nosnippet>377</a> <span class="kw">return </span>SchemaComparison::Different;
+<a href=#378 id=378 data-nosnippet>378</a> }
+<a href=#379 id=379 data-nosnippet>379</a>
+<a href=#380 id=380 data-nosnippet>380</a> <span class="kw">if </span>source_field.name() != target_field.name() {
+<a href=#381 id=381 data-nosnippet>381</a> names_changed = <span class="bool-val">true</span>;
+<a href=#382 id=382 data-nosnippet>382</a> }
+<a href=#383 id=383 data-nosnippet>383</a> }
+<a href=#384 id=384 data-nosnippet>384</a>
+<a href=#385 id=385 data-nosnippet>385</a> <span class="kw">if </span>names_changed {
+<a href=#386 id=386 data-nosnippet>386</a> SchemaComparison::NameChangesOnly
+<a href=#387 id=387 data-nosnippet>387</a> } <span class="kw">else </span>{
+<a href=#388 id=388 data-nosnippet>388</a> SchemaComparison::Equivalent
+<a href=#389 id=389 data-nosnippet>389</a> }
+<a href=#390 id=390 data-nosnippet>390</a> }
+<a href=#391 id=391 data-nosnippet>391</a>
+<a href=#392 id=392 data-nosnippet>392</a> <span class="kw">fn </span>generate_transform_operations(
+<a href=#393 id=393 data-nosnippet>393</a> source_schema: <span class="kw-2">&</span>ArrowSchemaRef,
+<a href=#394 id=394 data-nosnippet>394</a> snapshot_schema: <span class="kw-2">&</span>IcebergSchema,
+<a href=#395 id=395 data-nosnippet>395</a> projected_iceberg_field_ids: <span class="kw-2">&</span>[i32],
+<a href=#396 id=396 data-nosnippet>396</a> field_id_to_mapped_schema_map: HashMap<i32, (FieldRef, usize)>,
+<a href=#397 id=397 data-nosnippet>397</a> constants_map: HashMap<i32, PrimitiveLiteral>,
+<a href=#398 id=398 data-nosnippet>398</a> _partition_spec: <span class="prelude-ty">Option</span><<span class="kw-2">&</span>PartitionSpec>,
+<a href=#399 id=399 data-nosnippet>399</a> ) -> <span class="prelude-ty">Result</span><Vec<ColumnSource>> {
+<a href=#400 id=400 data-nosnippet>400</a> <span class="kw">let </span>field_id_to_source_schema_map =
+<a href=#401 id=401 data-nosnippet>401</a> <span class="self">Self</span>::build_field_id_to_arrow_schema_map(source_schema)<span class="question-mark">?</span>;
+<a href=#402 id=402 data-nosnippet>402</a>
+<a href=#403 id=403 data-nosnippet>403</a> projected_iceberg_field_ids
+<a href=#404 id=404 data-nosnippet>404</a> .iter()
+<a href=#405 id=405 data-nosnippet>405</a> .map(|field_id| {
+<a href=#406 id=406 data-nosnippet>406</a> <span class="kw">let </span>(target_field, <span class="kw">_</span>) =
+<a href=#407 id=407 data-nosnippet>407</a> field_id_to_mapped_schema_map
+<a href=#408 id=408 data-nosnippet>408</a> .get(field_id)
+<a href=#409 id=409 data-nosnippet>409</a> .ok_or(Error::new(
+<a href=#410 id=410 data-nosnippet>410</a> ErrorKind::Unexpected,
+<a href=#411 id=411 data-nosnippet>411</a> <span class="string">"could not find field in schema"</span>,
+<a href=#412 id=412 data-nosnippet>412</a> ))<span class="question-mark">?</span>;
+<a href=#413 id=413 data-nosnippet>413</a> <span class="kw">let </span>target_type = target_field.data_type();
+<a href=#414 id=414 data-nosnippet>414</a>
+<a href=#415 id=415 data-nosnippet>415</a> <span class="kw">let </span>iceberg_field = snapshot_schema.field_by_id(<span class="kw-2">*</span>field_id).ok_or(Error::new(
+<a href=#416 id=416 data-nosnippet>416</a> ErrorKind::Unexpected,
+<a href=#417 id=417 data-nosnippet>417</a> <span class="string">"Field not found in snapshot schema"</span>,
+<a href=#418 id=418 data-nosnippet>418</a> ))<span class="question-mark">?</span>;
+<a href=#419 id=419 data-nosnippet>419</a>
+<a href=#420 id=420 data-nosnippet>420</a> <span class="comment">// Iceberg spec's "Column Projection" rules (https://iceberg.apache.org/spec/#column-projection).
+<a href=#421 id=421 data-nosnippet>421</a> // For fields "not present" in data files:
+<a href=#422 id=422 data-nosnippet>422</a> // 1. Use partition metadata (identity transforms only)
+<a href=#423 id=423 data-nosnippet>423</a> // 2. Use name mapping
+<a href=#424 id=424 data-nosnippet>424</a> // 3. Use initial_default
+<a href=#425 id=425 data-nosnippet>425</a> // 4. Return null
+<a href=#426 id=426 data-nosnippet>426</a> //
+<a href=#427 id=427 data-nosnippet>427</a> // Why check partition constants before Parquet field IDs (Java: BaseParquetReaders.java:299):
+<a href=#428 id=428 data-nosnippet>428</a> // In add_files scenarios, partition columns may exist in BOTH Parquet AND partition metadata.
+<a href=#429 id=429 data-nosnippet>429</a> // Partition metadata is authoritative - it defines which partition this file belongs to.
+<a href=#430 id=430 data-nosnippet>430</a>
+<a href=#431 id=431 data-nosnippet>431</a> // Field ID resolution now happens in ArrowReader via:
+<a href=#432 id=432 data-nosnippet>432</a> // 1. Embedded field IDs (ParquetSchemaUtil.hasIds() = true) - trust them
+<a href=#433 id=433 data-nosnippet>433</a> // 2. Name mapping (ParquetSchemaUtil.applyNameMapping()) - applied upfront
+<a href=#434 id=434 data-nosnippet>434</a> // 3. Position-based fallback (ParquetSchemaUtil.addFallbackIds()) - applied upfront
+<a href=#435 id=435 data-nosnippet>435</a> //
+<a href=#436 id=436 data-nosnippet>436</a> // At this point, all field IDs in the source schema are trustworthy.
+<a href=#437 id=437 data-nosnippet>437</a> // No conflict detection needed - schema resolution happened in reader.rs.
+<a href=#438 id=438 data-nosnippet>438</a> </span><span class="kw">let </span>field_by_id = field_id_to_source_schema_map.get(field_id).map(
+<a href=#439 id=439 data-nosnippet>439</a> |(source_field, source_index)| {
+<a href=#440 id=440 data-nosnippet>440</a> <span class="kw">if </span>source_field.data_type().equals_datatype(target_type) {
+<a href=#441 id=441 data-nosnippet>441</a> ColumnSource::PassThrough {
+<a href=#442 id=442 data-nosnippet>442</a> source_index: <span class="kw-2">*</span>source_index,
+<a href=#443 id=443 data-nosnippet>443</a> }
+<a href=#444 id=444 data-nosnippet>444</a> } <span class="kw">else </span>{
+<a href=#445 id=445 data-nosnippet>445</a> ColumnSource::Promote {
+<a href=#446 id=446 data-nosnippet>446</a> target_type: target_type.clone(),
+<a href=#447 id=447 data-nosnippet>447</a> source_index: <span class="kw-2">*</span>source_index,
+<a href=#448 id=448 data-nosnippet>448</a> }
+<a href=#449 id=449 data-nosnippet>449</a> }
+<a href=#450 id=450 data-nosnippet>450</a> },
+<a href=#451 id=451 data-nosnippet>451</a> );
+<a href=#452 id=452 data-nosnippet>452</a>
+<a href=#453 id=453 data-nosnippet>453</a> <span class="comment">// Apply spec's fallback steps for "not present" fields.
+<a href=#454 id=454 data-nosnippet>454</a> </span><span class="kw">let </span>column_source = <span class="kw">if let </span><span class="prelude-val">Some</span>(constant_value) = constants_map.get(field_id) {
+<a href=#455 id=455 data-nosnippet>455</a> <span class="comment">// Rule #1: Identity partition constant
+<a href=#456 id=456 data-nosnippet>456</a> </span>ColumnSource::Add {
+<a href=#457 id=457 data-nosnippet>457</a> value: <span class="prelude-val">Some</span>(constant_value.clone()),
+<a href=#458 id=458 data-nosnippet>458</a> target_type: target_type.clone(),
+<a href=#459 id=459 data-nosnippet>459</a> }
+<a href=#460 id=460 data-nosnippet>460</a> } <span class="kw">else if let </span><span class="prelude-val">Some</span>(source) = field_by_id {
+<a href=#461 id=461 data-nosnippet>461</a> source
+<a href=#462 id=462 data-nosnippet>462</a> } <span class="kw">else </span>{
+<a href=#463 id=463 data-nosnippet>463</a> <span class="comment">// Rules #2, #3 and #4:
+<a href=#464 id=464 data-nosnippet>464</a> // Rule #2 (name mapping) was already applied in reader.rs if needed.
+<a href=#465 id=465 data-nosnippet>465</a> // If field_id is still not found, the column doesn't exist in the Parquet file.
+<a href=#466 id=466 data-nosnippet>466</a> // Fall through to rule #3 (initial_default) or rule #4 (null).
+<a href=#467 id=467 data-nosnippet>467</a> </span><span class="kw">let </span>default_value = iceberg_field.initial_default.as_ref().and_then(|lit| {
+<a href=#468 id=468 data-nosnippet>468</a> <span class="kw">if let </span>Literal::Primitive(prim) = lit {
+<a href=#469 id=469 data-nosnippet>469</a> <span class="prelude-val">Some</span>(prim.clone())
+<a href=#470 id=470 data-nosnippet>470</a> } <span class="kw">else </span>{
+<a href=#471 id=471 data-nosnippet>471</a> <span class="prelude-val">None
+<a href=#472 id=472 data-nosnippet>472</a> </span>}
+<a href=#473 id=473 data-nosnippet>473</a> });
+<a href=#474 id=474 data-nosnippet>474</a> ColumnSource::Add {
+<a href=#475 id=475 data-nosnippet>475</a> value: default_value,
+<a href=#476 id=476 data-nosnippet>476</a> target_type: target_type.clone(),
+<a href=#477 id=477 data-nosnippet>477</a> }
+<a href=#478 id=478 data-nosnippet>478</a> };
+<a href=#479 id=479 data-nosnippet>479</a>
+<a href=#480 id=480 data-nosnippet>480</a> <span class="prelude-val">Ok</span>(column_source)
+<a href=#481 id=481 data-nosnippet>481</a> })
+<a href=#482 id=482 data-nosnippet>482</a> .collect()
+<a href=#483 id=483 data-nosnippet>483</a> }
+<a href=#484 id=484 data-nosnippet>484</a>
+<a href=#485 id=485 data-nosnippet>485</a> <span class="kw">fn </span>build_field_id_to_arrow_schema_map(
+<a href=#486 id=486 data-nosnippet>486</a> source_schema: <span class="kw-2">&</span>SchemaRef,
+<a href=#487 id=487 data-nosnippet>487</a> ) -> <span class="prelude-ty">Result</span><HashMap<i32, (FieldRef, usize)>> {
+<a href=#488 id=488 data-nosnippet>488</a> <span class="kw">let </span><span class="kw-2">mut </span>field_id_to_source_schema = HashMap::new();
+<a href=#489 id=489 data-nosnippet>489</a> <span class="kw">for </span>(source_field_idx, source_field) <span class="kw">in </span>source_schema.fields.iter().enumerate() {
+<a href=#490 id=490 data-nosnippet>490</a> <span class="comment">// Check if field has a field ID in metadata
+<a href=#491 id=491 data-nosnippet>491</a> </span><span class="kw">if let </span><span class="prelude-val">Some</span>(field_id_str) = source_field.metadata().get(PARQUET_FIELD_ID_META_KEY) {
+<a href=#492 id=492 data-nosnippet>492</a> <span class="kw">let </span>this_field_id = field_id_str.parse().map_err(|e| {
+<a href=#493 id=493 data-nosnippet>493</a> Error::new(
+<a href=#494 id=494 data-nosnippet>494</a> ErrorKind::DataInvalid,
+<a href=#495 id=495 data-nosnippet>495</a> <span class="macro">format!</span>(<span class="string">"field id not parseable as an i32: {}"</span>, e),
+<a href=#496 id=496 data-nosnippet>496</a> )
+<a href=#497 id=497 data-nosnippet>497</a> })<span class="question-mark">?</span>;
+<a href=#498 id=498 data-nosnippet>498</a>
+<a href=#499 id=499 data-nosnippet>499</a> field_id_to_source_schema
+<a href=#500 id=500 data-nosnippet>500</a> .insert(this_field_id, (source_field.clone(), source_field_idx));
+<a href=#501 id=501 data-nosnippet>501</a> }
+<a href=#502 id=502 data-nosnippet>502</a> <span class="comment">// If field doesn't have a field ID, skip it - name mapping will handle it
+<a href=#503 id=503 data-nosnippet>503</a> </span>}
+<a href=#504 id=504 data-nosnippet>504</a>
+<a href=#505 id=505 data-nosnippet>505</a> <span class="prelude-val">Ok</span>(field_id_to_source_schema)
+<a href=#506 id=506 data-nosnippet>506</a> }
+<a href=#507 id=507 data-nosnippet>507</a>
+<a href=#508 id=508 data-nosnippet>508</a> <span class="kw">fn </span>transform_columns(
+<a href=#509 id=509 data-nosnippet>509</a> <span class="kw-2">&</span><span class="self">self</span>,
+<a href=#510 id=510 data-nosnippet>510</a> columns: <span class="kw-2">&</span>[Arc<<span class="kw">dyn </span>ArrowArray>],
+<a href=#511 id=511 data-nosnippet>511</a> operations: <span class="kw-2">&</span>[ColumnSource],
+<a href=#512 id=512 data-nosnippet>512</a> ) -> <span class="prelude-ty">Result</span><Vec<Arc<<span class="kw">dyn </span>ArrowArray>>> {
+<a href=#513 id=513 data-nosnippet>513</a> <span class="kw">if </span>columns.is_empty() {
+<a href=#514 id=514 data-nosnippet>514</a> <span class="kw">return </span><span class="prelude-val">Ok</span>(columns.to_vec());
+<a href=#515 id=515 data-nosnippet>515</a> }
+<a href=#516 id=516 data-nosnippet>516</a> <span class="kw">let </span>num_rows = columns[<span class="number">0</span>].len();
+<a href=#517 id=517 data-nosnippet>517</a>
+<a href=#518 id=518 data-nosnippet>518</a> operations
+<a href=#519 id=519 data-nosnippet>519</a> .iter()
+<a href=#520 id=520 data-nosnippet>520</a> .map(|op| {
+<a href=#521 id=521 data-nosnippet>521</a> <span class="prelude-val">Ok</span>(<span class="kw">match </span>op {
+<a href=#522 id=522 data-nosnippet>522</a> ColumnSource::PassThrough { source_index } => columns[<span class="kw-2">*</span>source_index].clone(),
+<a href=#523 id=523 data-nosnippet>523</a>
+<a href=#524 id=524 data-nosnippet>524</a> ColumnSource::Promote {
+<a href=#525 id=525 data-nosnippet>525</a> target_type,
+<a href=#526 id=526 data-nosnippet>526</a> source_index,
+<a href=#527 id=527 data-nosnippet>527</a> } => cast(<span class="kw-2">&*</span>columns[<span class="kw-2">*</span>source_index], target_type)<span class="question-mark">?</span>,
+<a href=#528 id=528 data-nosnippet>528</a>
+<a href=#529 id=529 data-nosnippet>529</a> ColumnSource::Add { target_type, value } => {
+<a href=#530 id=530 data-nosnippet>530</a> <span class="self">Self</span>::create_column(target_type, value, num_rows)<span class="question-mark">?
+<a href=#531 id=531 data-nosnippet>531</a> </span>}
+<a href=#532 id=532 data-nosnippet>532</a> })
+<a href=#533 id=533 data-nosnippet>533</a> })
+<a href=#534 id=534 data-nosnippet>534</a> .collect()
+<a href=#535 id=535 data-nosnippet>535</a> }
+<a href=#536 id=536 data-nosnippet>536</a>
+<a href=#537 id=537 data-nosnippet>537</a> <span class="kw">fn </span>create_column(
+<a href=#538 id=538 data-nosnippet>538</a> target_type: <span class="kw-2">&</span>DataType,
+<a href=#539 id=539 data-nosnippet>539</a> prim_lit: <span class="kw-2">&</span><span class="prelude-ty">Option</span><PrimitiveLiteral>,
+<a href=#540 id=540 data-nosnippet>540</a> num_rows: usize,
+<a href=#541 id=541 data-nosnippet>541</a> ) -> <span class="prelude-ty">Result</span><ArrayRef> {
+<a href=#542 id=542 data-nosnippet>542</a> <span class="prelude-val">Ok</span>(<span class="kw">match </span>(target_type, prim_lit) {
+<a href=#543 id=543 data-nosnippet>543</a> (DataType::Boolean, <span class="prelude-val">Some</span>(PrimitiveLiteral::Boolean(value))) => {
+<a href=#544 id=544 data-nosnippet>544</a> Arc::new(BooleanArray::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
+<a href=#545 id=545 data-nosnippet>545</a> }
+<a href=#546 id=546 data-nosnippet>546</a> (DataType::Boolean, <span class="prelude-val">None</span>) => {
+<a href=#547 id=547 data-nosnippet>547</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><bool>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#548 id=548 data-nosnippet>548</a> Arc::new(BooleanArray::from(vals))
+<a href=#549 id=549 data-nosnippet>549</a> }
+<a href=#550 id=550 data-nosnippet>550</a> (DataType::Int32, <span class="prelude-val">Some</span>(PrimitiveLiteral::Int(value))) => {
+<a href=#551 id=551 data-nosnippet>551</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
+<a href=#552 id=552 data-nosnippet>552</a> }
+<a href=#553 id=553 data-nosnippet>553</a> (DataType::Int32, <span class="prelude-val">None</span>) => {
+<a href=#554 id=554 data-nosnippet>554</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><i32>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#555 id=555 data-nosnippet>555</a> Arc::new(Int32Array::from(vals))
+<a href=#556 id=556 data-nosnippet>556</a> }
+<a href=#557 id=557 data-nosnippet>557</a> (DataType::Date32, <span class="prelude-val">Some</span>(PrimitiveLiteral::Int(value))) => {
+<a href=#558 id=558 data-nosnippet>558</a> Arc::new(Date32Array::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
+<a href=#559 id=559 data-nosnippet>559</a> }
+<a href=#560 id=560 data-nosnippet>560</a> (DataType::Date32, <span class="prelude-val">None</span>) => {
+<a href=#561 id=561 data-nosnippet>561</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><i32>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#562 id=562 data-nosnippet>562</a> Arc::new(Date32Array::from(vals))
+<a href=#563 id=563 data-nosnippet>563</a> }
+<a href=#564 id=564 data-nosnippet>564</a> (DataType::Int64, <span class="prelude-val">Some</span>(PrimitiveLiteral::Long(value))) => {
+<a href=#565 id=565 data-nosnippet>565</a> Arc::new(Int64Array::from(<span class="macro">vec!</span>[<span class="kw-2">*</span>value; num_rows]))
+<a href=#566 id=566 data-nosnippet>566</a> }
+<a href=#567 id=567 data-nosnippet>567</a> (DataType::Int64, <span class="prelude-val">None</span>) => {
+<a href=#568 id=568 data-nosnippet>568</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><i64>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#569 id=569 data-nosnippet>569</a> Arc::new(Int64Array::from(vals))
+<a href=#570 id=570 data-nosnippet>570</a> }
+<a href=#571 id=571 data-nosnippet>571</a> (DataType::Float32, <span class="prelude-val">Some</span>(PrimitiveLiteral::Float(value))) => {
+<a href=#572 id=572 data-nosnippet>572</a> Arc::new(Float32Array::from(<span class="macro">vec!</span>[value.<span class="number">0</span>; num_rows]))
+<a href=#573 id=573 data-nosnippet>573</a> }
+<a href=#574 id=574 data-nosnippet>574</a> (DataType::Float32, <span class="prelude-val">None</span>) => {
+<a href=#575 id=575 data-nosnippet>575</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><f32>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#576 id=576 data-nosnippet>576</a> Arc::new(Float32Array::from(vals))
+<a href=#577 id=577 data-nosnippet>577</a> }
+<a href=#578 id=578 data-nosnippet>578</a> (DataType::Float64, <span class="prelude-val">Some</span>(PrimitiveLiteral::Double(value))) => {
+<a href=#579 id=579 data-nosnippet>579</a> Arc::new(Float64Array::from(<span class="macro">vec!</span>[value.<span class="number">0</span>; num_rows]))
+<a href=#580 id=580 data-nosnippet>580</a> }
+<a href=#581 id=581 data-nosnippet>581</a> (DataType::Float64, <span class="prelude-val">None</span>) => {
+<a href=#582 id=582 data-nosnippet>582</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><f64>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#583 id=583 data-nosnippet>583</a> Arc::new(Float64Array::from(vals))
+<a href=#584 id=584 data-nosnippet>584</a> }
+<a href=#585 id=585 data-nosnippet>585</a> (DataType::Utf8, <span class="prelude-val">Some</span>(PrimitiveLiteral::String(value))) => {
+<a href=#586 id=586 data-nosnippet>586</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[value.clone(); num_rows]))
+<a href=#587 id=587 data-nosnippet>587</a> }
+<a href=#588 id=588 data-nosnippet>588</a> (DataType::Utf8, <span class="prelude-val">None</span>) => {
+<a href=#589 id=589 data-nosnippet>589</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><String>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#590 id=590 data-nosnippet>590</a> Arc::new(StringArray::from(vals))
+<a href=#591 id=591 data-nosnippet>591</a> }
+<a href=#592 id=592 data-nosnippet>592</a> (DataType::Binary, <span class="prelude-val">Some</span>(PrimitiveLiteral::Binary(value))) => {
+<a href=#593 id=593 data-nosnippet>593</a> Arc::new(BinaryArray::from_vec(<span class="macro">vec!</span>[value; num_rows]))
+<a href=#594 id=594 data-nosnippet>594</a> }
+<a href=#595 id=595 data-nosnippet>595</a> (DataType::Binary, <span class="prelude-val">None</span>) => {
+<a href=#596 id=596 data-nosnippet>596</a> <span class="kw">let </span>vals: Vec<<span class="prelude-ty">Option</span><<span class="kw-2">&</span>[u8]>> = <span class="macro">vec!</span>[<span class="prelude-val">None</span>; num_rows];
+<a href=#597 id=597 data-nosnippet>597</a> Arc::new(BinaryArray::from_opt_vec(vals))
+<a href=#598 id=598 data-nosnippet>598</a> }
+<a href=#599 id=599 data-nosnippet>599</a> (DataType::Struct(fields), <span class="prelude-val">None</span>) => {
+<a href=#600 id=600 data-nosnippet>600</a> <span class="comment">// Create a StructArray filled with nulls. Per Iceberg spec, optional struct fields
+<a href=#601 id=601 data-nosnippet>601</a> // default to null when added to the schema. We defer non-null default struct values
+<a href=#602 id=602 data-nosnippet>602</a> // and leave them as not implemented yet.
+<a href=#603 id=603 data-nosnippet>603</a> </span><span class="kw">let </span>null_arrays: Vec<ArrayRef> = fields
+<a href=#604 id=604 data-nosnippet>604</a> .iter()
+<a href=#605 id=605 data-nosnippet>605</a> .map(|field| <span class="self">Self</span>::create_column(field.data_type(), <span class="kw-2">&</span><span class="prelude-val">None</span>, num_rows))
+<a href=#606 id=606 data-nosnippet>606</a> .collect::<<span class="prelude-ty">Result</span><Vec<<span class="kw">_</span>>>>()<span class="question-mark">?</span>;
<a href=#607 id=607 data-nosnippet>607</a>
-<a href=#608 id=608 data-nosnippet>608</a><span class="attr">#[cfg(test)]
-<a href=#609 id=609 data-nosnippet>609</a></span><span class="kw">mod </span>test {
-<a href=#610 id=610 data-nosnippet>610</a> <span class="kw">use </span>std::collections::HashMap;
-<a href=#611 id=611 data-nosnippet>611</a> <span class="kw">use </span>std::sync::Arc;
-<a href=#612 id=612 data-nosnippet>612</a>
-<a href=#613 id=613 data-nosnippet>613</a> <span class="kw">use </span>arrow_array::{
-<a href=#614 id=614 data-nosnippet>614</a> Array, Date32Array, Float32Array, Float64Array, Int32Array, Int64Array, RecordBatch,
-<a href=#615 id=615 data-nosnippet>615</a> StringArray,
-<a href=#616 id=616 data-nosnippet>616</a> };
-<a href=#617 id=617 data-nosnippet>617</a> <span class="kw">use </span>arrow_schema::{DataType, Field, Schema <span class="kw">as </span>ArrowSchema};
-<a href=#618 id=618 data-nosnippet>618</a> <span class="kw">use </span>parquet::arrow::PARQUET_FIELD_ID_META_KEY;
-<a href=#619 id=619 data-nosnippet>619</a>
-<a href=#620 id=620 data-nosnippet>620</a> <span class="kw">use </span><span class="kw">crate</span>::arrow::record_batch_transformer::{
-<a href=#621 id=621 data-nosnippet>621</a> RecordBatchTransformer, RecordBatchTransformerBuilder,
-<a href=#622 id=622 data-nosnippet>622</a> };
-<a href=#623 id=623 data-nosnippet>623</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Literal, NestedField, PrimitiveType, Schema, Struct, Type};
+<a href=#608 id=608 data-nosnippet>608</a> Arc::new(StructArray::new(
+<a href=#609 id=609 data-nosnippet>609</a> fields.clone(),
+<a href=#610 id=610 data-nosnippet>610</a> null_arrays,
+<a href=#611 id=611 data-nosnippet>611</a> <span class="prelude-val">Some</span>(NullBuffer::new_null(num_rows)),
+<a href=#612 id=612 data-nosnippet>612</a> ))
+<a href=#613 id=613 data-nosnippet>613</a> }
+<a href=#614 id=614 data-nosnippet>614</a> (DataType::Null, <span class="kw">_</span>) => Arc::new(NullArray::new(num_rows)),
+<a href=#615 id=615 data-nosnippet>615</a> (dt, <span class="kw">_</span>) => {
+<a href=#616 id=616 data-nosnippet>616</a> <span class="kw">return </span><span class="prelude-val">Err</span>(Error::new(
+<a href=#617 id=617 data-nosnippet>617</a> ErrorKind::Unexpected,
+<a href=#618 id=618 data-nosnippet>618</a> <span class="macro">format!</span>(<span class="string">"unexpected target column type {}"</span>, dt),
+<a href=#619 id=619 data-nosnippet>619</a> ));
+<a href=#620 id=620 data-nosnippet>620</a> }
+<a href=#621 id=621 data-nosnippet>621</a> })
+<a href=#622 id=622 data-nosnippet>622</a> }
+<a href=#623 id=623 data-nosnippet>623</a>}
<a href=#624 id=624 data-nosnippet>624</a>
-<a href=#625 id=625 data-nosnippet>625</a> <span class="attr">#[test]
-<a href=#626 id=626 data-nosnippet>626</a> </span><span class="kw">fn </span>build_field_id_to_source_schema_map_works() {
-<a href=#627 id=627 data-nosnippet>627</a> <span class="kw">let </span>arrow_schema = arrow_schema_already_same_as_target();
-<a href=#628 id=628 data-nosnippet>628</a>
-<a href=#629 id=629 data-nosnippet>629</a> <span class="kw">let </span>result =
-<a href=#630 id=630 data-nosnippet>630</a> RecordBatchTransformer::build_field_id_to_arrow_schema_map(<span class="kw-2">&</span>arrow_schema).unwrap();
-<a href=#631 id=631 data-nosnippet>631</a>
-<a href=#632 id=632 data-nosnippet>632</a> <span class="kw">let </span>expected = HashMap::from_iter([
-<a href=#633 id=633 data-nosnippet>633</a> (<span class="number">10</span>, (arrow_schema.fields()[<span class="number">0</span>].clone(), <span class="number">0</span>)),
-<a href=#634 id=634 data-nosnippet>634</a> (<span class="number">11</span>, (arrow_schema.fields()[<span class="number">1</span>].clone(), <span class="number">1</span>)),
-<a href=#635 id=635 data-nosnippet>635</a> (<span class="number">12</span>, (arrow_schema.fields()[<span class="number">2</span>].clone(), <span class="number">2</span>)),
-<a href=#636 id=636 data-nosnippet>636</a> (<span class="number">14</span>, (arrow_schema.fields()[<span class="number">3</span>].clone(), <span class="number">3</span>)),
-<a href=#637 id=637 data-nosnippet>637</a> (<span class="number">15</span>, (arrow_schema.fields()[<span class="number">4</span>].clone(), <span class="number">4</span>)),
-<a href=#638 id=638 data-nosnippet>638</a> ]);
-<a href=#639 id=639 data-nosnippet>639</a>
-<a href=#640 id=640 data-nosnippet>640</a> <span class="macro">assert!</span>(result.eq(<span class="kw-2">&</span>expected));
-<a href=#641 id=641 data-nosnippet>641</a> }
-<a href=#642 id=642 data-nosnippet>642</a>
-<a href=#643 id=643 data-nosnippet>643</a> <span class="attr">#[test]
-<a href=#644 id=644 data-nosnippet>644</a> </span><span class="kw">fn </span>processor_returns_properly_shaped_record_batch_when_no_schema_migration_required() {
-<a href=#645 id=645 data-nosnippet>645</a> <span class="kw">let </span>snapshot_schema = Arc::new(iceberg_table_schema());
-<a href=#646 id=646 data-nosnippet>646</a> <span class="kw">let </span>projected_iceberg_field_ids = [<span class="number">13</span>, <span class="number">14</span>];
-<a href=#647 id=647 data-nosnippet>647</a>
-<a href=#648 id=648 data-nosnippet>648</a> <span class="kw">let </span><span class="kw-2">mut </span>inst =
-<a href=#649 id=649 data-nosnippet>649</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_iceberg_field_ids)
-<a href=#650 id=650 data-nosnippet>650</a> .build();
-<a href=#651 id=651 data-nosnippet>651</a>
-<a href=#652 id=652 data-nosnippet>652</a> <span class="kw">let </span>result = inst
-<a href=#653 id=653 data-nosnippet>653</a> .process_record_batch(source_record_batch_no_migration_required())
-<a href=#654 id=654 data-nosnippet>654</a> .unwrap();
-<a href=#655 id=655 data-nosnippet>655</a>
-<a href=#656 id=656 data-nosnippet>656</a> <span class="kw">let </span>expected = source_record_batch_no_migration_required();
-<a href=#657 id=657 data-nosnippet>657</a>
-<a href=#658 id=658 data-nosnippet>658</a> <span class="macro">assert_eq!</span>(result, expected);
-<a href=#659 id=659 data-nosnippet>659</a> }
-<a href=#660 id=660 data-nosnippet>660</a>
-<a href=#661 id=661 data-nosnippet>661</a> <span class="attr">#[test]
-<a href=#662 id=662 data-nosnippet>662</a> </span><span class="kw">fn </span>processor_returns_properly_shaped_record_batch_when_schema_migration_required() {
-<a href=#663 id=663 data-nosnippet>663</a> <span class="kw">let </span>snapshot_schema = Arc::new(iceberg_table_schema());
-<a href=#664 id=664 data-nosnippet>664</a> <span class="kw">let </span>projected_iceberg_field_ids = [<span class="number">10</span>, <span class="number">11</span>, <span class="number">12</span>, <span class="number">14</span>, <span class="number">15</span>]; <span class="comment">// a, b, c, e, f
-<a href=#665 id=665 data-nosnippet>665</a>
-<a href=#666 id=666 data-nosnippet>666</a> </span><span class="kw">let </span><span class="kw-2">mut </span>inst =
-<a href=#667 id=667 data-nosnippet>667</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_iceberg_field_ids)
-<a href=#668 id=668 data-nosnippet>668</a> .build();
-<a href=#669 id=669 data-nosnippet>669</a>
-<a href=#670 id=670 data-nosnippet>670</a> <span class="kw">let </span>result = inst.process_record_batch(source_record_batch()).unwrap();
-<a href=#671 id=671 data-nosnippet>671</a>
-<a href=#672 id=672 data-nosnippet>672</a> <span class="kw">let </span>expected = expected_record_batch_migration_required();
-<a href=#673 id=673 data-nosnippet>673</a>
-<a href=#674 id=674 data-nosnippet>674</a> <span class="macro">assert_eq!</span>(result, expected);
-<a href=#675 id=675 data-nosnippet>675</a> }
-<a href=#676 id=676 data-nosnippet>676</a>
-<a href=#677 id=677 data-nosnippet>677</a> <span class="attr">#[test]
-<a href=#678 id=678 data-nosnippet>678</a> </span><span class="kw">fn </span>schema_evolution_adds_date_column_with_nulls() {
-<a href=#679 id=679 data-nosnippet>679</a> <span class="comment">// Reproduces TestSelect.readAndWriteWithBranchAfterSchemaChange from iceberg-spark.
-<a href=#680 id=680 data-nosnippet>680</a> // When reading old snapshots after adding a DATE column, the transformer must
-<a href=#681 id=681 data-nosnippet>681</a> // populate the new column with NULL values since old files lack this field.
-<a href=#682 id=682 data-nosnippet>682</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
-<a href=#683 id=683 data-nosnippet>683</a> Schema::builder()
-<a href=#684 id=684 data-nosnippet>684</a> .with_schema_id(<span class="number">1</span>)
-<a href=#685 id=685 data-nosnippet>685</a> .with_fields(<span class="macro">vec!</span>[
-<a href=#686 id=686 data-nosnippet>686</a> NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
-<a href=#687 id=687 data-nosnippet>687</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#688 id=688 data-nosnippet>688</a> NestedField::optional(<span class="number">3</span>, <span class="string">"date_col"</span>, Type::Primitive(PrimitiveType::Date))
-<a href=#689 id=689 data-nosnippet>689</a> .into(),
-<a href=#690 id=690 data-nosnippet>690</a> ])
-<a href=#691 id=691 data-nosnippet>691</a> .build()
-<a href=#692 id=692 data-nosnippet>692</a> .unwrap(),
-<a href=#693 id=693 data-nosnippet>693</a> );
-<a href=#694 id=694 data-nosnippet>694</a> <span class="kw">let </span>projected_iceberg_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>];
-<a href=#695 id=695 data-nosnippet>695</a>
-<a href=#696 id=696 data-nosnippet>696</a> <span class="kw">let </span><span class="kw-2">mut </span>transformer =
-<a href=#697 id=697 data-nosnippet>697</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_iceberg_field_ids)
-<a href=#698 id=698 data-nosnippet>698</a> .build();
-<a href=#699 id=699 data-nosnippet>699</a>
-<a href=#700 id=700 data-nosnippet>700</a> <span class="kw">let </span>file_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#701 id=701 data-nosnippet>701</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
-<a href=#702 id=702 data-nosnippet>702</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"2"</span>),
-<a href=#703 id=703 data-nosnippet>703</a> ]));
-<a href=#704 id=704 data-nosnippet>704</a>
-<a href=#705 id=705 data-nosnippet>705</a> <span class="kw">let </span>file_batch = RecordBatch::try_new(file_schema, <span class="macro">vec!</span>[
-<a href=#706 id=706 data-nosnippet>706</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>])),
-<a href=#707 id=707 data-nosnippet>707</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[
-<a href=#708 id=708 data-nosnippet>708</a> <span class="prelude-val">Some</span>(<span class="string">"Alice"</span>),
-<a href=#709 id=709 data-nosnippet>709</a> <span class="prelude-val">Some</span>(<span class="string">"Bob"</span>),
-<a href=#710 id=710 data-nosnippet>710</a> <span class="prelude-val">Some</span>(<span class="string">"Charlie"</span>),
-<a href=#711 id=711 data-nosnippet>711</a> ])),
-<a href=#712 id=712 data-nosnippet>712</a> ])
-<a href=#713 id=713 data-nosnippet>713</a> .unwrap();
-<a href=#714 id=714 data-nosnippet>714</a>
-<a href=#715 id=715 data-nosnippet>715</a> <span class="kw">let </span>result = transformer.process_record_batch(file_batch).unwrap();
+<a href=#625 id=625 data-nosnippet>625</a><span class="attr">#[cfg(test)]
+<a href=#626 id=626 data-nosnippet>626</a></span><span class="kw">mod </span>test {
+<a href=#627 id=627 data-nosnippet>627</a> <span class="kw">use </span>std::collections::HashMap;
+<a href=#628 id=628 data-nosnippet>628</a> <span class="kw">use </span>std::sync::Arc;
+<a href=#629 id=629 data-nosnippet>629</a>
+<a href=#630 id=630 data-nosnippet>630</a> <span class="kw">use </span>arrow_array::{
+<a href=#631 id=631 data-nosnippet>631</a> Array, Date32Array, Float32Array, Float64Array, Int32Array, Int64Array, RecordBatch,
+<a href=#632 id=632 data-nosnippet>632</a> StringArray,
+<a href=#633 id=633 data-nosnippet>633</a> };
+<a href=#634 id=634 data-nosnippet>634</a> <span class="kw">use </span>arrow_schema::{DataType, Field, Schema <span class="kw">as </span>ArrowSchema};
+<a href=#635 id=635 data-nosnippet>635</a> <span class="kw">use </span>parquet::arrow::PARQUET_FIELD_ID_META_KEY;
+<a href=#636 id=636 data-nosnippet>636</a>
+<a href=#637 id=637 data-nosnippet>637</a> <span class="kw">use </span><span class="kw">crate</span>::arrow::record_batch_transformer::{
+<a href=#638 id=638 data-nosnippet>638</a> RecordBatchTransformer, RecordBatchTransformerBuilder,
+<a href=#639 id=639 data-nosnippet>639</a> };
+<a href=#640 id=640 data-nosnippet>640</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Literal, NestedField, PrimitiveType, Schema, Struct, Type};
+<a href=#641 id=641 data-nosnippet>641</a>
+<a href=#642 id=642 data-nosnippet>642</a> <span class="attr">#[test]
+<a href=#643 id=643 data-nosnippet>643</a> </span><span class="kw">fn </span>build_field_id_to_source_schema_map_works() {
+<a href=#644 id=644 data-nosnippet>644</a> <span class="kw">let </span>arrow_schema = arrow_schema_already_same_as_target();
+<a href=#645 id=645 data-nosnippet>645</a>
+<a href=#646 id=646 data-nosnippet>646</a> <span class="kw">let </span>result =
+<a href=#647 id=647 data-nosnippet>647</a> RecordBatchTransformer::build_field_id_to_arrow_schema_map(<span class="kw-2">&</span>arrow_schema).unwrap();
+<a href=#648 id=648 data-nosnippet>648</a>
+<a href=#649 id=649 data-nosnippet>649</a> <span class="kw">let </span>expected = HashMap::from_iter([
+<a href=#650 id=650 data-nosnippet>650</a> (<span class="number">10</span>, (arrow_schema.fields()[<span class="number">0</span>].clone(), <span class="number">0</span>)),
+<a href=#651 id=651 data-nosnippet>651</a> (<span class="number">11</span>, (arrow_schema.fields()[<span class="number">1</span>].clone(), <span class="number">1</span>)),
+<a href=#652 id=652 data-nosnippet>652</a> (<span class="number">12</span>, (arrow_schema.fields()[<span class="number">2</span>].clone(), <span class="number">2</span>)),
+<a href=#653 id=653 data-nosnippet>653</a> (<span class="number">14</span>, (arrow_schema.fields()[<span class="number">3</span>].clone(), <span class="number">3</span>)),
+<a href=#654 id=654 data-nosnippet>654</a> (<span class="number">15</span>, (arrow_schema.fields()[<span class="number">4</span>].clone(), <span class="number">4</span>)),
+<a href=#655 id=655 data-nosnippet>655</a> ]);
+<a href=#656 id=656 data-nosnippet>656</a>
+<a href=#657 id=657 data-nosnippet>657</a> <span class="macro">assert!</span>(result.eq(<span class="kw-2">&</span>expected));
+<a href=#658 id=658 data-nosnippet>658</a> }
+<a href=#659 id=659 data-nosnippet>659</a>
+<a href=#660 id=660 data-nosnippet>660</a> <span class="attr">#[test]
+<a href=#661 id=661 data-nosnippet>661</a> </span><span class="kw">fn </span>processor_returns_properly_shaped_record_batch_when_no_schema_migration_required() {
+<a href=#662 id=662 data-nosnippet>662</a> <span class="kw">let </span>snapshot_schema = Arc::new(iceberg_table_schema());
+<a href=#663 id=663 data-nosnippet>663</a> <span class="kw">let </span>projected_iceberg_field_ids = [<span class="number">13</span>, <span class="number">14</span>];
+<a href=#664 id=664 data-nosnippet>664</a>
+<a href=#665 id=665 data-nosnippet>665</a> <span class="kw">let </span><span class="kw-2">mut </span>inst =
+<a href=#666 id=666 data-nosnippet>666</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_iceberg_field_ids)
+<a href=#667 id=667 data-nosnippet>667</a> .build();
+<a href=#668 id=668 data-nosnippet>668</a>
+<a href=#669 id=669 data-nosnippet>669</a> <span class="kw">let </span>result = inst
+<a href=#670 id=670 data-nosnippet>670</a> .process_record_batch(source_record_batch_no_migration_required())
+<a href=#671 id=671 data-nosnippet>671</a> .unwrap();
+<a href=#672 id=672 data-nosnippet>672</a>
+<a href=#673 id=673 data-nosnippet>673</a> <span class="kw">let </span>expected = source_record_batch_no_migration_required();
+<a href=#674 id=674 data-nosnippet>674</a>
+<a href=#675 id=675 data-nosnippet>675</a> <span class="macro">assert_eq!</span>(result, expected);
+<a href=#676 id=676 data-nosnippet>676</a> }
+<a href=#677 id=677 data-nosnippet>677</a>
+<a href=#678 id=678 data-nosnippet>678</a> <span class="attr">#[test]
+<a href=#679 id=679 data-nosnippet>679</a> </span><span class="kw">fn </span>processor_returns_properly_shaped_record_batch_when_schema_migration_required() {
+<a href=#680 id=680 data-nosnippet>680</a> <span class="kw">let </span>snapshot_schema = Arc::new(iceberg_table_schema());
+<a href=#681 id=681 data-nosnippet>681</a> <span class="kw">let </span>projected_iceberg_field_ids = [<span class="number">10</span>, <span class="number">11</span>, <span class="number">12</span>, <span class="number">14</span>, <span class="number">15</span>]; <span class="comment">// a, b, c, e, f
+<a href=#682 id=682 data-nosnippet>682</a>
+<a href=#683 id=683 data-nosnippet>683</a> </span><span class="kw">let </span><span class="kw-2">mut </span>inst =
+<a href=#684 id=684 data-nosnippet>684</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_iceberg_field_ids)
+<a href=#685 id=685 data-nosnippet>685</a> .build();
+<a href=#686 id=686 data-nosnippet>686</a>
+<a href=#687 id=687 data-nosnippet>687</a> <span class="kw">let </span>result = inst.process_record_batch(source_record_batch()).unwrap();
+<a href=#688 id=688 data-nosnippet>688</a>
+<a href=#689 id=689 data-nosnippet>689</a> <span class="kw">let </span>expected = expected_record_batch_migration_required();
+<a href=#690 id=690 data-nosnippet>690</a>
+<a href=#691 id=691 data-nosnippet>691</a> <span class="macro">assert_eq!</span>(result, expected);
+<a href=#692 id=692 data-nosnippet>692</a> }
+<a href=#693 id=693 data-nosnippet>693</a>
+<a href=#694 id=694 data-nosnippet>694</a> <span class="attr">#[test]
+<a href=#695 id=695 data-nosnippet>695</a> </span><span class="kw">fn </span>schema_evolution_adds_date_column_with_nulls() {
+<a href=#696 id=696 data-nosnippet>696</a> <span class="comment">// Reproduces TestSelect.readAndWriteWithBranchAfterSchemaChange from iceberg-spark.
+<a href=#697 id=697 data-nosnippet>697</a> // When reading old snapshots after adding a DATE column, the transformer must
+<a href=#698 id=698 data-nosnippet>698</a> // populate the new column with NULL values since old files lack this field.
+<a href=#699 id=699 data-nosnippet>699</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
+<a href=#700 id=700 data-nosnippet>700</a> Schema::builder()
+<a href=#701 id=701 data-nosnippet>701</a> .with_schema_id(<span class="number">1</span>)
+<a href=#702 id=702 data-nosnippet>702</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#703 id=703 data-nosnippet>703</a> NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
+<a href=#704 id=704 data-nosnippet>704</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#705 id=705 data-nosnippet>705</a> NestedField::optional(<span class="number">3</span>, <span class="string">"date_col"</span>, Type::Primitive(PrimitiveType::Date))
+<a href=#706 id=706 data-nosnippet>706</a> .into(),
+<a href=#707 id=707 data-nosnippet>707</a> ])
+<a href=#708 id=708 data-nosnippet>708</a> .build()
+<a href=#709 id=709 data-nosnippet>709</a> .unwrap(),
+<a href=#710 id=710 data-nosnippet>710</a> );
+<a href=#711 id=711 data-nosnippet>711</a> <span class="kw">let </span>projected_iceberg_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>];
+<a href=#712 id=712 data-nosnippet>712</a>
+<a href=#713 id=713 data-nosnippet>713</a> <span class="kw">let </span><span class="kw-2">mut </span>transformer =
+<a href=#714 id=714 data-nosnippet>714</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_iceberg_field_ids)
+<a href=#715 id=715 data-nosnippet>715</a> .build();
<a href=#716 id=716 data-nosnippet>716</a>
-<a href=#717 id=717 data-nosnippet>717</a> <span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">3</span>);
-<a href=#718 id=718 data-nosnippet>718</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">3</span>);
-<a href=#719 id=719 data-nosnippet>719</a>
-<a href=#720 id=720 data-nosnippet>720</a> <span class="kw">let </span>id_column = result
-<a href=#721 id=721 data-nosnippet>721</a> .column(<span class="number">0</span>)
-<a href=#722 id=722 data-nosnippet>722</a> .as_any()
-<a href=#723 id=723 data-nosnippet>723</a> .downcast_ref::<Int32Array>()
-<a href=#724 id=724 data-nosnippet>724</a> .unwrap();
-<a href=#725 id=725 data-nosnippet>725</a> <span class="macro">assert_eq!</span>(id_column.values(), <span class="kw-2">&</span>[<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>]);
-<a href=#726 id=726 data-nosnippet>726</a>
-<a href=#727 id=727 data-nosnippet>727</a> <span class="kw">let </span>name_column = result
-<a href=#728 id=728 data-nosnippet>728</a> .column(<span class="number">1</span>)
-<a href=#729 id=729 data-nosnippet>729</a> .as_any()
-<a href=#730 id=730 data-nosnippet>730</a> .downcast_ref::<StringArray>()
-<a href=#731 id=731 data-nosnippet>731</a> .unwrap();
-<a href=#732 id=732 data-nosnippet>732</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
-<a href=#733 id=733 data-nosnippet>733</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
-<a href=#734 id=734 data-nosnippet>734</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">2</span>), <span class="string">"Charlie"</span>);
-<a href=#735 id=735 data-nosnippet>735</a>
-<a href=#736 id=736 data-nosnippet>736</a> <span class="kw">let </span>date_column = result
-<a href=#737 id=737 data-nosnippet>737</a> .column(<span class="number">2</span>)
-<a href=#738 id=738 data-nosnippet>738</a> .as_any()
-<a href=#739 id=739 data-nosnippet>739</a> .downcast_ref::<Date32Array>()
-<a href=#740 id=740 data-nosnippet>740</a> .unwrap();
-<a href=#741 id=741 data-nosnippet>741</a> <span class="macro">assert!</span>(date_column.is_null(<span class="number">0</span>));
-<a href=#742 id=742 data-nosnippet>742</a> <span class="macro">assert!</span>(date_column.is_null(<span class="number">1</span>));
-<a href=#743 id=743 data-nosnippet>743</a> <span class="macro">assert!</span>(date_column.is_null(<span class="number">2</span>));
-<a href=#744 id=744 data-nosnippet>744</a> }
-<a href=#745 id=745 data-nosnippet>745</a>
-<a href=#746 id=746 data-nosnippet>746</a> <span class="kw">pub fn </span>source_record_batch() -> RecordBatch {
-<a href=#747 id=747 data-nosnippet>747</a> RecordBatch::try_new(
-<a href=#748 id=748 data-nosnippet>748</a> arrow_schema_promotion_addition_and_renaming_required(),
-<a href=#749 id=749 data-nosnippet>749</a> <span class="macro">vec!</span>[
-<a href=#750 id=750 data-nosnippet>750</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">1001</span>), <span class="prelude-val">Some</span>(<span class="number">1002</span>), <span class="prelude-val">Some</span>(<span class="number">1003</span>)])), <span class="comment">// b
-<a href=#751 id=751 data-nosnippet>751</a> </span>Arc::new(Float32Array::from(<span class="macro">vec!</span>[
-<a href=#752 id=752 data-nosnippet>752</a> <span class="prelude-val">Some</span>(<span class="number">12.125</span>),
-<a href=#753 id=753 data-nosnippet>753</a> <span class="prelude-val">Some</span>(<span class="number">23.375</span>),
-<a href=#754 id=754 data-nosnippet>754</a> <span class="prelude-val">Some</span>(<span class="number">34.875</span>),
-<a href=#755 id=755 data-nosnippet>755</a> ])), <span class="comment">// c
-<a href=#756 id=756 data-nosnippet>756</a> </span>Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">2001</span>), <span class="prelude-val">Some</span>(<span class="number">2002</span>), <span class="prelude-val">Some</span>(<span class="number">2003</span>)])), <span class="comment">// d
-<a href=#757 id=757 data-nosnippet>757</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
-<a href=#758 id=758 data-nosnippet>758</a> <span class="prelude-val">Some</span>(<span class="string">"Apache"</span>),
-<a href=#759 id=759 data-nosnippet>759</a> <span class="prelude-val">Some</span>(<span class="string">"Iceberg"</span>),
-<a href=#760 id=760 data-nosnippet>760</a> <span class="prelude-val">Some</span>(<span class="string">"Rocks"</span>),
-<a href=#761 id=761 data-nosnippet>761</a> ])), <span class="comment">// e
-<a href=#762 id=762 data-nosnippet>762</a> </span>],
-<a href=#763 id=763 data-nosnippet>763</a> )
-<a href=#764 id=764 data-nosnippet>764</a> .unwrap()
-<a href=#765 id=765 data-nosnippet>765</a> }
-<a href=#766 id=766 data-nosnippet>766</a>
-<a href=#767 id=767 data-nosnippet>767</a> <span class="kw">pub fn </span>source_record_batch_no_migration_required() -> RecordBatch {
-<a href=#768 id=768 data-nosnippet>768</a> RecordBatch::try_new(
-<a href=#769 id=769 data-nosnippet>769</a> arrow_schema_no_promotion_addition_or_renaming_required(),
-<a href=#770 id=770 data-nosnippet>770</a> <span class="macro">vec!</span>[
-<a href=#771 id=771 data-nosnippet>771</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">2001</span>), <span class="prelude-val">Some</span>(<span class="number">2002</span>), <span class="prelude-val">Some</span>(<span class="number">2003</span>)])), <span class="comment">// d
-<a href=#772 id=772 data-nosnippet>772</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
-<a href=#773 id=773 data-nosnippet>773</a> <span class="prelude-val">Some</span>(<span class="string">"Apache"</span>),
-<a href=#774 id=774 data-nosnippet>774</a> <span class="prelude-val">Some</span>(<span class="string">"Iceberg"</span>),
-<a href=#775 id=775 data-nosnippet>775</a> <span class="prelude-val">Some</span>(<span class="string">"Rocks"</span>),
-<a href=#776 id=776 data-nosnippet>776</a> ])), <span class="comment">// e
-<a href=#777 id=777 data-nosnippet>777</a> </span>],
-<a href=#778 id=778 data-nosnippet>778</a> )
-<a href=#779 id=779 data-nosnippet>779</a> .unwrap()
-<a href=#780 id=780 data-nosnippet>780</a> }
-<a href=#781 id=781 data-nosnippet>781</a>
-<a href=#782 id=782 data-nosnippet>782</a> <span class="kw">pub fn </span>expected_record_batch_migration_required() -> RecordBatch {
-<a href=#783 id=783 data-nosnippet>783</a> RecordBatch::try_new(arrow_schema_already_same_as_target(), <span class="macro">vec!</span>[
-<a href=#784 id=784 data-nosnippet>784</a> Arc::new(StringArray::from(Vec::<<span class="prelude-ty">Option</span><String>>::from([
-<a href=#785 id=785 data-nosnippet>785</a> <span class="prelude-val">None</span>, <span class="prelude-val">None</span>, <span class="prelude-val">None</span>,
-<a href=#786 id=786 data-nosnippet>786</a> ]))), <span class="comment">// a
-<a href=#787 id=787 data-nosnippet>787</a> </span>Arc::new(Int64Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">1001</span>), <span class="prelude-val">Some</span>(<span class="number">1002</span>), <span class="prelude-val">Some</span>(<span class="number">1003</span>)])), <span class="comment">// b
-<a href=#788 id=788 data-nosnippet>788</a> </span>Arc::new(Float64Array::from(<span class="macro">vec!</span>[
-<a href=#789 id=789 data-nosnippet>789</a> <span class="prelude-val">Some</span>(<span class="number">12.125</span>),
-<a href=#790 id=790 data-nosnippet>790</a> <span class="prelude-val">Some</span>(<span class="number">23.375</span>),
-<a href=#791 id=791 data-nosnippet>791</a> <span class="prelude-val">Some</span>(<span class="number">34.875</span>),
-<a href=#792 id=792 data-nosnippet>792</a> ])), <span class="comment">// c
-<a href=#793 id=793 data-nosnippet>793</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
-<a href=#794 id=794 data-nosnippet>794</a> <span class="prelude-val">Some</span>(<span class="string">"Apache"</span>),
-<a href=#795 id=795 data-nosnippet>795</a> <span class="prelude-val">Some</span>(<span class="string">"Iceberg"</span>),
-<a href=#796 id=796 data-nosnippet>796</a> <span class="prelude-val">Some</span>(<span class="string">"Rocks"</span>),
-<a href=#797 id=797 data-nosnippet>797</a> ])), <span class="comment">// e (d skipped by projection)
-<a href=#798 id=798 data-nosnippet>798</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
-<a href=#799 id=799 data-nosnippet>799</a> <span class="prelude-val">Some</span>(<span class="string">"(╯°□°)╯"</span>),
-<a href=#800 id=800 data-nosnippet>800</a> <span class="prelude-val">Some</span>(<span class="string">"(╯°□°)╯"</span>),
-<a href=#801 id=801 data-nosnippet>801</a> <span class="prelude-val">Some</span>(<span class="string">"(╯°□°)╯"</span>),
-<a href=#802 id=802 data-nosnippet>802</a> ])), <span class="comment">// f
-<a href=#803 id=803 data-nosnippet>803</a> </span>])
-<a href=#804 id=804 data-nosnippet>804</a> .unwrap()
-<a href=#805 id=805 data-nosnippet>805</a> }
-<a href=#806 id=806 data-nosnippet>806</a>
-<a href=#807 id=807 data-nosnippet>807</a> <span class="kw">pub fn </span>iceberg_table_schema() -> Schema {
-<a href=#808 id=808 data-nosnippet>808</a> Schema::builder()
-<a href=#809 id=809 data-nosnippet>809</a> .with_schema_id(<span class="number">2</span>)
-<a href=#810 id=810 data-nosnippet>810</a> .with_fields(<span class="macro">vec!</span>[
-<a href=#811 id=811 data-nosnippet>811</a> NestedField::optional(<span class="number">10</span>, <span class="string">"a"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#812 id=812 data-nosnippet>812</a> NestedField::required(<span class="number">11</span>, <span class="string">"b"</span>, Type::Primitive(PrimitiveType::Long)).into(),
-<a href=#813 id=813 data-nosnippet>813</a> NestedField::required(<span class="number">12</span>, <span class="string">"c"</span>, Type::Primitive(PrimitiveType::Double)).into(),
-<a href=#814 id=814 data-nosnippet>814</a> NestedField::required(<span class="number">13</span>, <span class="string">"d"</span>, Type::Primitive(PrimitiveType::Int)).into(),
-<a href=#815 id=815 data-nosnippet>815</a> NestedField::optional(<span class="number">14</span>, <span class="string">"e"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#816 id=816 data-nosnippet>816</a> NestedField::required(<span class="number">15</span>, <span class="string">"f"</span>, Type::Primitive(PrimitiveType::String))
-<a href=#817 id=817 data-nosnippet>817</a> .with_initial_default(Literal::string(<span class="string">"(╯°□°)╯"</span>))
-<a href=#818 id=818 data-nosnippet>818</a> .into(),
-<a href=#819 id=819 data-nosnippet>819</a> ])
-<a href=#820 id=820 data-nosnippet>820</a> .build()
-<a href=#821 id=821 data-nosnippet>821</a> .unwrap()
-<a href=#822 id=822 data-nosnippet>822</a> }
-<a href=#823 id=823 data-nosnippet>823</a>
-<a href=#824 id=824 data-nosnippet>824</a> <span class="kw">fn </span>arrow_schema_already_same_as_target() -> Arc<ArrowSchema> {
-<a href=#825 id=825 data-nosnippet>825</a> Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#826 id=826 data-nosnippet>826</a> simple_field(<span class="string">"a"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"10"</span>),
-<a href=#827 id=827 data-nosnippet>827</a> simple_field(<span class="string">"b"</span>, DataType::Int64, <span class="bool-val">false</span>, <span class="string">"11"</span>),
-<a href=#828 id=828 data-nosnippet>828</a> simple_field(<span class="string">"c"</span>, DataType::Float64, <span class="bool-val">false</span>, <span class="string">"12"</span>),
-<a href=#829 id=829 data-nosnippet>829</a> simple_field(<span class="string">"e"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"14"</span>),
-<a href=#830 id=830 data-nosnippet>830</a> simple_field(<span class="string">"f"</span>, DataType::Utf8, <span class="bool-val">false</span>, <span class="string">"15"</span>),
-<a href=#831 id=831 data-nosnippet>831</a> ]))
-<a href=#832 id=832 data-nosnippet>832</a> }
-<a href=#833 id=833 data-nosnippet>833</a>
-<a href=#834 id=834 data-nosnippet>834</a> <span class="kw">fn </span>arrow_schema_promotion_addition_and_renaming_required() -> Arc<ArrowSchema> {
-<a href=#835 id=835 data-nosnippet>835</a> Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#836 id=836 data-nosnippet>836</a> simple_field(<span class="string">"b"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"11"</span>),
-<a href=#837 id=837 data-nosnippet>837</a> simple_field(<span class="string">"c"</span>, DataType::Float32, <span class="bool-val">false</span>, <span class="string">"12"</span>),
-<a href=#838 id=838 data-nosnippet>838</a> simple_field(<span class="string">"d"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"13"</span>),
-<a href=#839 id=839 data-nosnippet>839</a> simple_field(<span class="string">"e_old"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"14"</span>),
-<a href=#840 id=840 data-nosnippet>840</a> ]))
-<a href=#841 id=841 data-nosnippet>841</a> }
-<a href=#842 id=842 data-nosnippet>842</a>
-<a href=#843 id=843 data-nosnippet>843</a> <span class="kw">fn </span>arrow_schema_no_promotion_addition_or_renaming_required() -> Arc<ArrowSchema> {
-<a href=#844 id=844 data-nosnippet>844</a> Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#845 id=845 data-nosnippet>845</a> simple_field(<span class="string">"d"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"13"</span>),
-<a href=#846 id=846 data-nosnippet>846</a> simple_field(<span class="string">"e"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"14"</span>),
-<a href=#847 id=847 data-nosnippet>847</a> ]))
-<a href=#848 id=848 data-nosnippet>848</a> }
-<a href=#849 id=849 data-nosnippet>849</a>
-<a href=#850 id=850 data-nosnippet>850</a> <span class="doccomment">/// Create a simple arrow field with metadata.
-<a href=#851 id=851 data-nosnippet>851</a> </span><span class="kw">fn </span>simple_field(name: <span class="kw-2">&</span>str, ty: DataType, nullable: bool, value: <span class="kw-2">&</span>str) -> Field {
-<a href=#852 id=852 data-nosnippet>852</a> Field::new(name, ty, nullable).with_metadata(HashMap::from([(
-<a href=#853 id=853 data-nosnippet>853</a> PARQUET_FIELD_ID_META_KEY.to_string(),
-<a href=#854 id=854 data-nosnippet>854</a> value.to_string(),
-<a href=#855 id=855 data-nosnippet>855</a> )]))
-<a href=#856 id=856 data-nosnippet>856</a> }
-<a href=#857 id=857 data-nosnippet>857</a>
-<a href=#858 id=858 data-nosnippet>858</a> <span class="doccomment">/// Test for add_files with Parquet files that have NO field IDs (Hive tables).
-<a href=#859 id=859 data-nosnippet>859</a> ///
-<a href=#860 id=860 data-nosnippet>860</a> /// This reproduces the scenario from Iceberg spec where:
-<a href=#861 id=861 data-nosnippet>861</a> /// - Hive-style partitioned Parquet files are imported via add_files procedure
-<a href=#862 id=862 data-nosnippet>862</a> /// - Parquet files originally DO NOT have field IDs (typical for Hive tables)
-<a href=#863 id=863 data-nosnippet>863</a> /// - ArrowReader applies name mapping to assign correct Iceberg field IDs
-<a href=#864 id=864 data-nosnippet>864</a> /// - Iceberg schema assigns field IDs: id (1), name (2), dept (3), subdept (4)
-<a href=#865 id=865 data-nosnippet>865</a> /// - Partition columns (id, dept) have initial_default values
-<a href=#866 id=866 data-nosnippet>866</a> ///
-<a href=#867 id=867 data-nosnippet>867</a> /// Per the Iceberg spec (https://iceberg.apache.org/spec/#column-projection),
-<a href=#868 id=868 data-nosnippet>868</a> /// this scenario requires `schema.name-mapping.default` from table metadata
-<a href=#869 id=869 data-nosnippet>869</a> /// to correctly map Parquet columns by name to Iceberg field IDs.
-<a href=#870 id=870 data-nosnippet>870</a> /// This mapping is now applied in ArrowReader before data is processed.
-<a href=#871 id=871 data-nosnippet>871</a> ///
-<a href=#872 id=872 data-nosnippet>872</a> /// Expected behavior:
-<a href=#873 id=873 data-nosnippet>873</a> /// 1. id=1 (from initial_default) - spec rule #3
-<a href=#874 id=874 data-nosnippet>874</a> /// 2. name="John Doe" (from Parquet with field_id=2 assigned by reader) - found by field ID
-<a href=#875 id=875 data-nosnippet>875</a> /// 3. dept="hr" (from initial_default) - spec rule #3
-<a href=#876 id=876 data-nosnippet>876</a> /// 4. subdept="communications" (from Parquet with field_id=4 assigned by reader) - found by field ID
-<a href=#877 id=877 data-nosnippet>877</a> </span><span class="attr">#[test]
-<a href=#878 id=878 data-nosnippet>878</a> </span><span class="kw">fn </span>add_files_with_name_mapping_applied_in_reader() {
-<a href=#879 id=879 data-nosnippet>879</a> <span class="comment">// Iceberg schema after add_files: id (partition), name, dept (partition), subdept
-<a href=#880 id=880 data-nosnippet>880</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
-<a href=#881 id=881 data-nosnippet>881</a> Schema::builder()
-<a href=#882 id=882 data-nosnippet>882</a> .with_schema_id(<span class="number">0</span>)
-<a href=#883 id=883 data-nosnippet>883</a> .with_fields(<span class="macro">vec!</span>[
-<a href=#884 id=884 data-nosnippet>884</a> NestedField::optional(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int))
-<a href=#885 id=885 data-nosnippet>885</a> .with_initial_default(Literal::int(<span class="number">1</span>))
-<a href=#886 id=886 data-nosnippet>886</a> .into(),
-<a href=#887 id=887 data-nosnippet>887</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#888 id=888 data-nosnippet>888</a> NestedField::optional(<span class="number">3</span>, <span class="string">"dept"</span>, Type::Primitive(PrimitiveType::String))
-<a href=#889 id=889 data-nosnippet>889</a> .with_initial_default(Literal::string(<span class="string">"hr"</span>))
-<a href=#890 id=890 data-nosnippet>890</a> .into(),
-<a href=#891 id=891 data-nosnippet>891</a> NestedField::optional(<span class="number">4</span>, <span class="string">"subdept"</span>, Type::Primitive(PrimitiveType::String))
-<a href=#892 id=892 data-nosnippet>892</a> .into(),
-<a href=#893 id=893 data-nosnippet>893</a> ])
-<a href=#894 id=894 data-nosnippet>894</a> .build()
-<a href=#895 id=895 data-nosnippet>895</a> .unwrap(),
-<a href=#896 id=896 data-nosnippet>896</a> );
-<a href=#897 id=897 data-nosnippet>897</a>
-<a href=#898 id=898 data-nosnippet>898</a> <span class="comment">// Simulate ArrowReader having applied name mapping:
-<a href=#899 id=899 data-nosnippet>899</a> // Original Parquet: name, subdept (NO field IDs)
-<a href=#900 id=900 data-nosnippet>900</a> // After reader.rs applies name mapping: name (field_id=2), subdept (field_id=4)
-<a href=#901 id=901 data-nosnippet>901</a> //
-<a href=#902 id=902 data-nosnippet>902</a> // Note: Partition columns (id, dept) are NOT in the Parquet file - they're in directory paths
-<a href=#903 id=903 data-nosnippet>903</a> </span><span class="kw">use </span>std::collections::HashMap;
-<a href=#904 id=904 data-nosnippet>904</a> <span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#905 id=905 data-nosnippet>905</a> Field::new(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>).with_metadata(HashMap::from([(
-<a href=#906 id=906 data-nosnippet>906</a> <span class="string">"PARQUET:field_id"</span>.to_string(),
-<a href=#907 id=907 data-nosnippet>907</a> <span class="string">"2"</span>.to_string(),
-<a href=#908 id=908 data-nosnippet>908</a> )])),
-<a href=#909 id=909 data-nosnippet>909</a> Field::new(<span class="string">"subdept"</span>, DataType::Utf8, <span class="bool-val">true</span>).with_metadata(HashMap::from([(
-<a href=#910 id=910 data-nosnippet>910</a> <span class="string">"PARQUET:field_id"</span>.to_string(),
-<a href=#911 id=911 data-nosnippet>911</a> <span class="string">"4"</span>.to_string(),
-<a href=#912 id=912 data-nosnippet>912</a> )])),
-<a href=#913 id=913 data-nosnippet>913</a> ]));
-<a href=#914 id=914 data-nosnippet>914</a>
-<a href=#915 id=915 data-nosnippet>915</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>, <span class="number">4</span>]; <span class="comment">// id, name, dept, subdept
+<a href=#717 id=717 data-nosnippet>717</a> <span class="kw">let </span>file_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#718 id=718 data-nosnippet>718</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
+<a href=#719 id=719 data-nosnippet>719</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"2"</span>),
+<a href=#720 id=720 data-nosnippet>720</a> ]));
+<a href=#721 id=721 data-nosnippet>721</a>
+<a href=#722 id=722 data-nosnippet>722</a> <span class="kw">let </span>file_batch = RecordBatch::try_new(file_schema, <span class="macro">vec!</span>[
+<a href=#723 id=723 data-nosnippet>723</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>])),
+<a href=#724 id=724 data-nosnippet>724</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[
+<a href=#725 id=725 data-nosnippet>725</a> <span class="prelude-val">Some</span>(<span class="string">"Alice"</span>),
+<a href=#726 id=726 data-nosnippet>726</a> <span class="prelude-val">Some</span>(<span class="string">"Bob"</span>),
+<a href=#727 id=727 data-nosnippet>727</a> <span class="prelude-val">Some</span>(<span class="string">"Charlie"</span>),
+<a href=#728 id=728 data-nosnippet>728</a> ])),
+<a href=#729 id=729 data-nosnippet>729</a> ])
+<a href=#730 id=730 data-nosnippet>730</a> .unwrap();
+<a href=#731 id=731 data-nosnippet>731</a>
+<a href=#732 id=732 data-nosnippet>732</a> <span class="kw">let </span>result = transformer.process_record_batch(file_batch).unwrap();
+<a href=#733 id=733 data-nosnippet>733</a>
+<a href=#734 id=734 data-nosnippet>734</a> <span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">3</span>);
+<a href=#735 id=735 data-nosnippet>735</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">3</span>);
+<a href=#736 id=736 data-nosnippet>736</a>
+<a href=#737 id=737 data-nosnippet>737</a> <span class="kw">let </span>id_column = result
+<a href=#738 id=738 data-nosnippet>738</a> .column(<span class="number">0</span>)
+<a href=#739 id=739 data-nosnippet>739</a> .as_any()
+<a href=#740 id=740 data-nosnippet>740</a> .downcast_ref::<Int32Array>()
+<a href=#741 id=741 data-nosnippet>741</a> .unwrap();
+<a href=#742 id=742 data-nosnippet>742</a> <span class="macro">assert_eq!</span>(id_column.values(), <span class="kw-2">&</span>[<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>]);
+<a href=#743 id=743 data-nosnippet>743</a>
+<a href=#744 id=744 data-nosnippet>744</a> <span class="kw">let </span>name_column = result
+<a href=#745 id=745 data-nosnippet>745</a> .column(<span class="number">1</span>)
+<a href=#746 id=746 data-nosnippet>746</a> .as_any()
+<a href=#747 id=747 data-nosnippet>747</a> .downcast_ref::<StringArray>()
+<a href=#748 id=748 data-nosnippet>748</a> .unwrap();
+<a href=#749 id=749 data-nosnippet>749</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
+<a href=#750 id=750 data-nosnippet>750</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
+<a href=#751 id=751 data-nosnippet>751</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">2</span>), <span class="string">"Charlie"</span>);
+<a href=#752 id=752 data-nosnippet>752</a>
+<a href=#753 id=753 data-nosnippet>753</a> <span class="kw">let </span>date_column = result
+<a href=#754 id=754 data-nosnippet>754</a> .column(<span class="number">2</span>)
+<a href=#755 id=755 data-nosnippet>755</a> .as_any()
+<a href=#756 id=756 data-nosnippet>756</a> .downcast_ref::<Date32Array>()
+<a href=#757 id=757 data-nosnippet>757</a> .unwrap();
+<a href=#758 id=758 data-nosnippet>758</a> <span class="macro">assert!</span>(date_column.is_null(<span class="number">0</span>));
+<a href=#759 id=759 data-nosnippet>759</a> <span class="macro">assert!</span>(date_column.is_null(<span class="number">1</span>));
+<a href=#760 id=760 data-nosnippet>760</a> <span class="macro">assert!</span>(date_column.is_null(<span class="number">2</span>));
+<a href=#761 id=761 data-nosnippet>761</a> }
+<a href=#762 id=762 data-nosnippet>762</a>
+<a href=#763 id=763 data-nosnippet>763</a> <span class="attr">#[test]
+<a href=#764 id=764 data-nosnippet>764</a> </span><span class="kw">fn </span>schema_evolution_adds_struct_column_with_nulls() {
+<a href=#765 id=765 data-nosnippet>765</a> <span class="comment">// Test that when a struct column is added after data files are written,
+<a href=#766 id=766 data-nosnippet>766</a> // the transformer can materialize the missing struct column with null values.
+<a href=#767 id=767 data-nosnippet>767</a> // This reproduces the scenario from Iceberg 1.10.0 TestSparkReaderDeletes tests
+<a href=#768 id=768 data-nosnippet>768</a> // where binaryData and structData columns were added to the schema.
+<a href=#769 id=769 data-nosnippet>769</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
+<a href=#770 id=770 data-nosnippet>770</a> Schema::builder()
+<a href=#771 id=771 data-nosnippet>771</a> .with_schema_id(<span class="number">1</span>)
+<a href=#772 id=772 data-nosnippet>772</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#773 id=773 data-nosnippet>773</a> NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
+<a href=#774 id=774 data-nosnippet>774</a> NestedField::required(<span class="number">2</span>, <span class="string">"data"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#775 id=775 data-nosnippet>775</a> NestedField::optional(
+<a href=#776 id=776 data-nosnippet>776</a> <span class="number">3</span>,
+<a href=#777 id=777 data-nosnippet>777</a> <span class="string">"struct_col"</span>,
+<a href=#778 id=778 data-nosnippet>778</a> Type::Struct(<span class="kw">crate</span>::spec::StructType::new(<span class="macro">vec!</span>[
+<a href=#779 id=779 data-nosnippet>779</a> NestedField::optional(
+<a href=#780 id=780 data-nosnippet>780</a> <span class="number">100</span>,
+<a href=#781 id=781 data-nosnippet>781</a> <span class="string">"inner_field"</span>,
+<a href=#782 id=782 data-nosnippet>782</a> Type::Primitive(PrimitiveType::String),
+<a href=#783 id=783 data-nosnippet>783</a> )
+<a href=#784 id=784 data-nosnippet>784</a> .into(),
+<a href=#785 id=785 data-nosnippet>785</a> ])),
+<a href=#786 id=786 data-nosnippet>786</a> )
+<a href=#787 id=787 data-nosnippet>787</a> .into(),
+<a href=#788 id=788 data-nosnippet>788</a> ])
+<a href=#789 id=789 data-nosnippet>789</a> .build()
+<a href=#790 id=790 data-nosnippet>790</a> .unwrap(),
+<a href=#791 id=791 data-nosnippet>791</a> );
+<a href=#792 id=792 data-nosnippet>792</a> <span class="kw">let </span>projected_iceberg_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>];
+<a href=#793 id=793 data-nosnippet>793</a>
+<a href=#794 id=794 data-nosnippet>794</a> <span class="kw">let </span><span class="kw-2">mut </span>transformer =
+<a href=#795 id=795 data-nosnippet>795</a> RecordBatchTransformer::build(snapshot_schema, <span class="kw-2">&</span>projected_iceberg_field_ids);
+<a href=#796 id=796 data-nosnippet>796</a>
+<a href=#797 id=797 data-nosnippet>797</a> <span class="kw">let </span>file_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#798 id=798 data-nosnippet>798</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
+<a href=#799 id=799 data-nosnippet>799</a> simple_field(<span class="string">"data"</span>, DataType::Utf8, <span class="bool-val">false</span>, <span class="string">"2"</span>),
+<a href=#800 id=800 data-nosnippet>800</a> ]));
+<a href=#801 id=801 data-nosnippet>801</a>
+<a href=#802 id=802 data-nosnippet>802</a> <span class="kw">let </span>file_batch = RecordBatch::try_new(file_schema, <span class="macro">vec!</span>[
+<a href=#803 id=803 data-nosnippet>803</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>])),
+<a href=#804 id=804 data-nosnippet>804</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"a"</span>, <span class="string">"b"</span>, <span class="string">"c"</span>])),
+<a href=#805 id=805 data-nosnippet>805</a> ])
+<a href=#806 id=806 data-nosnippet>806</a> .unwrap();
+<a href=#807 id=807 data-nosnippet>807</a>
+<a href=#808 id=808 data-nosnippet>808</a> <span class="kw">let </span>result = transformer.process_record_batch(file_batch).unwrap();
+<a href=#809 id=809 data-nosnippet>809</a>
+<a href=#810 id=810 data-nosnippet>810</a> <span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">3</span>);
+<a href=#811 id=811 data-nosnippet>811</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">3</span>);
+<a href=#812 id=812 data-nosnippet>812</a>
+<a href=#813 id=813 data-nosnippet>813</a> <span class="kw">let </span>id_column = result
+<a href=#814 id=814 data-nosnippet>814</a> .column(<span class="number">0</span>)
+<a href=#815 id=815 data-nosnippet>815</a> .as_any()
+<a href=#816 id=816 data-nosnippet>816</a> .downcast_ref::<Int32Array>()
+<a href=#817 id=817 data-nosnippet>817</a> .unwrap();
+<a href=#818 id=818 data-nosnippet>818</a> <span class="macro">assert_eq!</span>(id_column.values(), <span class="kw-2">&</span>[<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>]);
+<a href=#819 id=819 data-nosnippet>819</a>
+<a href=#820 id=820 data-nosnippet>820</a> <span class="kw">let </span>data_column = result
+<a href=#821 id=821 data-nosnippet>821</a> .column(<span class="number">1</span>)
+<a href=#822 id=822 data-nosnippet>822</a> .as_any()
+<a href=#823 id=823 data-nosnippet>823</a> .downcast_ref::<StringArray>()
+<a href=#824 id=824 data-nosnippet>824</a> .unwrap();
+<a href=#825 id=825 data-nosnippet>825</a> <span class="macro">assert_eq!</span>(data_column.value(<span class="number">0</span>), <span class="string">"a"</span>);
+<a href=#826 id=826 data-nosnippet>826</a> <span class="macro">assert_eq!</span>(data_column.value(<span class="number">1</span>), <span class="string">"b"</span>);
+<a href=#827 id=827 data-nosnippet>827</a> <span class="macro">assert_eq!</span>(data_column.value(<span class="number">2</span>), <span class="string">"c"</span>);
+<a href=#828 id=828 data-nosnippet>828</a>
+<a href=#829 id=829 data-nosnippet>829</a> <span class="kw">let </span>struct_column = result
+<a href=#830 id=830 data-nosnippet>830</a> .column(<span class="number">2</span>)
+<a href=#831 id=831 data-nosnippet>831</a> .as_any()
+<a href=#832 id=832 data-nosnippet>832</a> .downcast_ref::<arrow_array::StructArray>()
+<a href=#833 id=833 data-nosnippet>833</a> .unwrap();
+<a href=#834 id=834 data-nosnippet>834</a> <span class="macro">assert!</span>(struct_column.is_null(<span class="number">0</span>));
+<a href=#835 id=835 data-nosnippet>835</a> <span class="macro">assert!</span>(struct_column.is_null(<span class="number">1</span>));
+<a href=#836 id=836 data-nosnippet>836</a> <span class="macro">assert!</span>(struct_column.is_null(<span class="number">2</span>));
+<a href=#837 id=837 data-nosnippet>837</a> }
+<a href=#838 id=838 data-nosnippet>838</a>
+<a href=#839 id=839 data-nosnippet>839</a> <span class="kw">pub fn </span>source_record_batch() -> RecordBatch {
+<a href=#840 id=840 data-nosnippet>840</a> RecordBatch::try_new(
+<a href=#841 id=841 data-nosnippet>841</a> arrow_schema_promotion_addition_and_renaming_required(),
+<a href=#842 id=842 data-nosnippet>842</a> <span class="macro">vec!</span>[
+<a href=#843 id=843 data-nosnippet>843</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">1001</span>), <span class="prelude-val">Some</span>(<span class="number">1002</span>), <span class="prelude-val">Some</span>(<span class="number">1003</span>)])), <span class="comment">// b
+<a href=#844 id=844 data-nosnippet>844</a> </span>Arc::new(Float32Array::from(<span class="macro">vec!</span>[
+<a href=#845 id=845 data-nosnippet>845</a> <span class="prelude-val">Some</span>(<span class="number">12.125</span>),
+<a href=#846 id=846 data-nosnippet>846</a> <span class="prelude-val">Some</span>(<span class="number">23.375</span>),
+<a href=#847 id=847 data-nosnippet>847</a> <span class="prelude-val">Some</span>(<span class="number">34.875</span>),
+<a href=#848 id=848 data-nosnippet>848</a> ])), <span class="comment">// c
+<a href=#849 id=849 data-nosnippet>849</a> </span>Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">2001</span>), <span class="prelude-val">Some</span>(<span class="number">2002</span>), <span class="prelude-val">Some</span>(<span class="number">2003</span>)])), <span class="comment">// d
+<a href=#850 id=850 data-nosnippet>850</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
+<a href=#851 id=851 data-nosnippet>851</a> <span class="prelude-val">Some</span>(<span class="string">"Apache"</span>),
+<a href=#852 id=852 data-nosnippet>852</a> <span class="prelude-val">Some</span>(<span class="string">"Iceberg"</span>),
+<a href=#853 id=853 data-nosnippet>853</a> <span class="prelude-val">Some</span>(<span class="string">"Rocks"</span>),
+<a href=#854 id=854 data-nosnippet>854</a> ])), <span class="comment">// e
+<a href=#855 id=855 data-nosnippet>855</a> </span>],
+<a href=#856 id=856 data-nosnippet>856</a> )
+<a href=#857 id=857 data-nosnippet>857</a> .unwrap()
+<a href=#858 id=858 data-nosnippet>858</a> }
+<a href=#859 id=859 data-nosnippet>859</a>
+<a href=#860 id=860 data-nosnippet>860</a> <span class="kw">pub fn </span>source_record_batch_no_migration_required() -> RecordBatch {
+<a href=#861 id=861 data-nosnippet>861</a> RecordBatch::try_new(
+<a href=#862 id=862 data-nosnippet>862</a> arrow_schema_no_promotion_addition_or_renaming_required(),
+<a href=#863 id=863 data-nosnippet>863</a> <span class="macro">vec!</span>[
+<a href=#864 id=864 data-nosnippet>864</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">2001</span>), <span class="prelude-val">Some</span>(<span class="number">2002</span>), <span class="prelude-val">Some</span>(<span class="number">2003</span>)])), <span class="comment">// d
+<a href=#865 id=865 data-nosnippet>865</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
+<a href=#866 id=866 data-nosnippet>866</a> <span class="prelude-val">Some</span>(<span class="string">"Apache"</span>),
+<a href=#867 id=867 data-nosnippet>867</a> <span class="prelude-val">Some</span>(<span class="string">"Iceberg"</span>),
+<a href=#868 id=868 data-nosnippet>868</a> <span class="prelude-val">Some</span>(<span class="string">"Rocks"</span>),
+<a href=#869 id=869 data-nosnippet>869</a> ])), <span class="comment">// e
+<a href=#870 id=870 data-nosnippet>870</a> </span>],
+<a href=#871 id=871 data-nosnippet>871</a> )
+<a href=#872 id=872 data-nosnippet>872</a> .unwrap()
+<a href=#873 id=873 data-nosnippet>873</a> }
+<a href=#874 id=874 data-nosnippet>874</a>
+<a href=#875 id=875 data-nosnippet>875</a> <span class="kw">pub fn </span>expected_record_batch_migration_required() -> RecordBatch {
+<a href=#876 id=876 data-nosnippet>876</a> RecordBatch::try_new(arrow_schema_already_same_as_target(), <span class="macro">vec!</span>[
+<a href=#877 id=877 data-nosnippet>877</a> Arc::new(StringArray::from(Vec::<<span class="prelude-ty">Option</span><String>>::from([
+<a href=#878 id=878 data-nosnippet>878</a> <span class="prelude-val">None</span>, <span class="prelude-val">None</span>, <span class="prelude-val">None</span>,
+<a href=#879 id=879 data-nosnippet>879</a> ]))), <span class="comment">// a
+<a href=#880 id=880 data-nosnippet>880</a> </span>Arc::new(Int64Array::from(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(<span class="number">1001</span>), <span class="prelude-val">Some</span>(<span class="number">1002</span>), <span class="prelude-val">Some</span>(<span class="number">1003</span>)])), <span class="comment">// b
+<a href=#881 id=881 data-nosnippet>881</a> </span>Arc::new(Float64Array::from(<span class="macro">vec!</span>[
+<a href=#882 id=882 data-nosnippet>882</a> <span class="prelude-val">Some</span>(<span class="number">12.125</span>),
+<a href=#883 id=883 data-nosnippet>883</a> <span class="prelude-val">Some</span>(<span class="number">23.375</span>),
+<a href=#884 id=884 data-nosnippet>884</a> <span class="prelude-val">Some</span>(<span class="number">34.875</span>),
+<a href=#885 id=885 data-nosnippet>885</a> ])), <span class="comment">// c
+<a href=#886 id=886 data-nosnippet>886</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
+<a href=#887 id=887 data-nosnippet>887</a> <span class="prelude-val">Some</span>(<span class="string">"Apache"</span>),
+<a href=#888 id=888 data-nosnippet>888</a> <span class="prelude-val">Some</span>(<span class="string">"Iceberg"</span>),
+<a href=#889 id=889 data-nosnippet>889</a> <span class="prelude-val">Some</span>(<span class="string">"Rocks"</span>),
+<a href=#890 id=890 data-nosnippet>890</a> ])), <span class="comment">// e (d skipped by projection)
+<a href=#891 id=891 data-nosnippet>891</a> </span>Arc::new(StringArray::from(<span class="macro">vec!</span>[
+<a href=#892 id=892 data-nosnippet>892</a> <span class="prelude-val">Some</span>(<span class="string">"(╯°□°)╯"</span>),
+<a href=#893 id=893 data-nosnippet>893</a> <span class="prelude-val">Some</span>(<span class="string">"(╯°□°)╯"</span>),
+<a href=#894 id=894 data-nosnippet>894</a> <span class="prelude-val">Some</span>(<span class="string">"(╯°□°)╯"</span>),
+<a href=#895 id=895 data-nosnippet>895</a> ])), <span class="comment">// f
+<a href=#896 id=896 data-nosnippet>896</a> </span>])
+<a href=#897 id=897 data-nosnippet>897</a> .unwrap()
+<a href=#898 id=898 data-nosnippet>898</a> }
+<a href=#899 id=899 data-nosnippet>899</a>
+<a href=#900 id=900 data-nosnippet>900</a> <span class="kw">pub fn </span>iceberg_table_schema() -> Schema {
+<a href=#901 id=901 data-nosnippet>901</a> Schema::builder()
+<a href=#902 id=902 data-nosnippet>902</a> .with_schema_id(<span class="number">2</span>)
+<a href=#903 id=903 data-nosnippet>903</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#904 id=904 data-nosnippet>904</a> NestedField::optional(<span class="number">10</span>, <span class="string">"a"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#905 id=905 data-nosnippet>905</a> NestedField::required(<span class="number">11</span>, <span class="string">"b"</span>, Type::Primitive(PrimitiveType::Long)).into(),
+<a href=#906 id=906 data-nosnippet>906</a> NestedField::required(<span class="number">12</span>, <span class="string">"c"</span>, Type::Primitive(PrimitiveType::Double)).into(),
+<a href=#907 id=907 data-nosnippet>907</a> NestedField::required(<span class="number">13</span>, <span class="string">"d"</span>, Type::Primitive(PrimitiveType::Int)).into(),
+<a href=#908 id=908 data-nosnippet>908</a> NestedField::optional(<span class="number">14</span>, <span class="string">"e"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#909 id=909 data-nosnippet>909</a> NestedField::required(<span class="number">15</span>, <span class="string">"f"</span>, Type::Primitive(PrimitiveType::String))
+<a href=#910 id=910 data-nosnippet>910</a> .with_initial_default(Literal::string(<span class="string">"(╯°□°)╯"</span>))
+<a href=#911 id=911 data-nosnippet>911</a> .into(),
+<a href=#912 id=912 data-nosnippet>912</a> ])
+<a href=#913 id=913 data-nosnippet>913</a> .build()
+<a href=#914 id=914 data-nosnippet>914</a> .unwrap()
+<a href=#915 id=915 data-nosnippet>915</a> }
<a href=#916 id=916 data-nosnippet>916</a>
-<a href=#917 id=917 data-nosnippet>917</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
-<a href=#918 id=918 data-nosnippet>918</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids).build();
-<a href=#919 id=919 data-nosnippet>919</a>
-<a href=#920 id=920 data-nosnippet>920</a> <span class="comment">// Create a Parquet RecordBatch with data for: name="John Doe", subdept="communications"
-<a href=#921 id=921 data-nosnippet>921</a> </span><span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
-<a href=#922 id=922 data-nosnippet>922</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"John Doe"</span>])),
-<a href=#923 id=923 data-nosnippet>923</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"communications"</span>])),
-<a href=#924 id=924 data-nosnippet>924</a> ])
-<a href=#925 id=925 data-nosnippet>925</a> .unwrap();
+<a href=#917 id=917 data-nosnippet>917</a> <span class="kw">fn </span>arrow_schema_already_same_as_target() -> Arc<ArrowSchema> {
+<a href=#918 id=918 data-nosnippet>918</a> Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#919 id=919 data-nosnippet>919</a> simple_field(<span class="string">"a"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"10"</span>),
+<a href=#920 id=920 data-nosnippet>920</a> simple_field(<span class="string">"b"</span>, DataType::Int64, <span class="bool-val">false</span>, <span class="string">"11"</span>),
+<a href=#921 id=921 data-nosnippet>921</a> simple_field(<span class="string">"c"</span>, DataType::Float64, <span class="bool-val">false</span>, <span class="string">"12"</span>),
+<a href=#922 id=922 data-nosnippet>922</a> simple_field(<span class="string">"e"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"14"</span>),
+<a href=#923 id=923 data-nosnippet>923</a> simple_field(<span class="string">"f"</span>, DataType::Utf8, <span class="bool-val">false</span>, <span class="string">"15"</span>),
+<a href=#924 id=924 data-nosnippet>924</a> ]))
+<a href=#925 id=925 data-nosnippet>925</a> }
<a href=#926 id=926 data-nosnippet>926</a>
-<a href=#927 id=927 data-nosnippet>927</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
-<a href=#928 id=928 data-nosnippet>928</a>
-<a href=#929 id=929 data-nosnippet>929</a> <span class="comment">// Verify the transformed RecordBatch has:
-<a href=#930 id=930 data-nosnippet>930</a> // - id=1 (from initial_default, not from Parquet)
-<a href=#931 id=931 data-nosnippet>931</a> // - name="John Doe" (from Parquet with correct field_id=2)
-<a href=#932 id=932 data-nosnippet>932</a> // - dept="hr" (from initial_default, not from Parquet)
-<a href=#933 id=933 data-nosnippet>933</a> // - subdept="communications" (from Parquet with correct field_id=4)
-<a href=#934 id=934 data-nosnippet>934</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">4</span>);
-<a href=#935 id=935 data-nosnippet>935</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">1</span>);
-<a href=#936 id=936 data-nosnippet>936</a>
-<a href=#937 id=937 data-nosnippet>937</a> <span class="kw">let </span>id_column = result
-<a href=#938 id=938 data-nosnippet>938</a> .column(<span class="number">0</span>)
-<a href=#939 id=939 data-nosnippet>939</a> .as_any()
-<a href=#940 id=940 data-nosnippet>940</a> .downcast_ref::<Int32Array>()
-<a href=#941 id=941 data-nosnippet>941</a> .unwrap();
-<a href=#942 id=942 data-nosnippet>942</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">1</span>);
-<a href=#943 id=943 data-nosnippet>943</a>
-<a href=#944 id=944 data-nosnippet>944</a> <span class="kw">let </span>name_column = result
-<a href=#945 id=945 data-nosnippet>945</a> .column(<span class="number">1</span>)
-<a href=#946 id=946 data-nosnippet>946</a> .as_any()
-<a href=#947 id=947 data-nosnippet>947</a> .downcast_ref::<StringArray>()
-<a href=#948 id=948 data-nosnippet>948</a> .unwrap();
-<a href=#949 id=949 data-nosnippet>949</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"John Doe"</span>);
+<a href=#927 id=927 data-nosnippet>927</a> <span class="kw">fn </span>arrow_schema_promotion_addition_and_renaming_required() -> Arc<ArrowSchema> {
+<a href=#928 id=928 data-nosnippet>928</a> Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#929 id=929 data-nosnippet>929</a> simple_field(<span class="string">"b"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"11"</span>),
+<a href=#930 id=930 data-nosnippet>930</a> simple_field(<span class="string">"c"</span>, DataType::Float32, <span class="bool-val">false</span>, <span class="string">"12"</span>),
+<a href=#931 id=931 data-nosnippet>931</a> simple_field(<span class="string">"d"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"13"</span>),
+<a href=#932 id=932 data-nosnippet>932</a> simple_field(<span class="string">"e_old"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"14"</span>),
+<a href=#933 id=933 data-nosnippet>933</a> ]))
+<a href=#934 id=934 data-nosnippet>934</a> }
+<a href=#935 id=935 data-nosnippet>935</a>
+<a href=#936 id=936 data-nosnippet>936</a> <span class="kw">fn </span>arrow_schema_no_promotion_addition_or_renaming_required() -> Arc<ArrowSchema> {
+<a href=#937 id=937 data-nosnippet>937</a> Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#938 id=938 data-nosnippet>938</a> simple_field(<span class="string">"d"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"13"</span>),
+<a href=#939 id=939 data-nosnippet>939</a> simple_field(<span class="string">"e"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"14"</span>),
+<a href=#940 id=940 data-nosnippet>940</a> ]))
+<a href=#941 id=941 data-nosnippet>941</a> }
+<a href=#942 id=942 data-nosnippet>942</a>
+<a href=#943 id=943 data-nosnippet>943</a> <span class="doccomment">/// Create a simple arrow field with metadata.
+<a href=#944 id=944 data-nosnippet>944</a> </span><span class="kw">fn </span>simple_field(name: <span class="kw-2">&</span>str, ty: DataType, nullable: bool, value: <span class="kw-2">&</span>str) -> Field {
+<a href=#945 id=945 data-nosnippet>945</a> Field::new(name, ty, nullable).with_metadata(HashMap::from([(
+<a href=#946 id=946 data-nosnippet>946</a> PARQUET_FIELD_ID_META_KEY.to_string(),
+<a href=#947 id=947 data-nosnippet>947</a> value.to_string(),
+<a href=#948 id=948 data-nosnippet>948</a> )]))
+<a href=#949 id=949 data-nosnippet>949</a> }
<a href=#950 id=950 data-nosnippet>950</a>
-<a href=#951 id=951 data-nosnippet>951</a> <span class="kw">let </span>dept_column = result
-<a href=#952 id=952 data-nosnippet>952</a> .column(<span class="number">2</span>)
-<a href=#953 id=953 data-nosnippet>953</a> .as_any()
-<a href=#954 id=954 data-nosnippet>954</a> .downcast_ref::<StringArray>()
-<a href=#955 id=955 data-nosnippet>955</a> .unwrap();
-<a href=#956 id=956 data-nosnippet>956</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">0</span>), <span class="string">"hr"</span>);
-<a href=#957 id=957 data-nosnippet>957</a>
-<a href=#958 id=958 data-nosnippet>958</a> <span class="kw">let </span>subdept_column = result
-<a href=#959 id=959 data-nosnippet>959</a> .column(<span class="number">3</span>)
-<a href=#960 id=960 data-nosnippet>960</a> .as_any()
-<a href=#961 id=961 data-nosnippet>961</a> .downcast_ref::<StringArray>()
-<a href=#962 id=962 data-nosnippet>962</a> .unwrap();
-<a href=#963 id=963 data-nosnippet>963</a> <span class="macro">assert_eq!</span>(subdept_column.value(<span class="number">0</span>), <span class="string">"communications"</span>);
-<a href=#964 id=964 data-nosnippet>964</a> }
-<a href=#965 id=965 data-nosnippet>965</a>
-<a href=#966 id=966 data-nosnippet>966</a> <span class="doccomment">/// Test for bucket partitioning where source columns must be read from data files.
-<a href=#967 id=967 data-nosnippet>967</a> ///
-<a href=#968 id=968 data-nosnippet>968</a> /// This test verifies correct implementation of the Iceberg spec's "Column Projection" rules:
-<a href=#969 id=969 data-nosnippet>969</a> /// > "Return the value from partition metadata if an **Identity Transform** exists for the field"
-<a href=#970 id=970 data-nosnippet>970</a> ///
-<a href=#971 id=971 data-nosnippet>971</a> /// # Why this test is critical
-<a href=#972 id=972 data-nosnippet>972</a> ///
-<a href=#973 id=973 data-nosnippet>973</a> /// The key insight is that partition metadata stores TRANSFORMED values, not source values:
-<a href=#974 id=974 data-nosnippet>974</a> /// - For `bucket(4, id)`, partition metadata has `id_bucket = 2` (the bucket number)
-<a href=#975 id=975 data-nosnippet>975</a> /// - The actual `id` column values (100, 200, 300) are ONLY in the data file
-<a href=#976 id=976 data-nosnippet>976</a> ///
-<a href=#977 id=977 data-nosnippet>977</a> /// If iceberg-rust incorrectly treated bucket-partitioned fields as constants, it would:
-<a href=#978 id=978 data-nosnippet>978</a> /// 1. Replace all `id` values with the constant `2` from partition metadata
-<a href=#979 id=979 data-nosnippet>979</a> /// 2. Break runtime filtering (e.g., `WHERE id = 100` would match no rows)
-<a href=#980 id=980 data-nosnippet>980</a> /// 3. Return incorrect query results
-<a href=#981 id=981 data-nosnippet>981</a> ///
-<a href=#982 id=982 data-nosnippet>982</a> /// # What this test verifies
-<a href=#983 id=983 data-nosnippet>983</a> ///
-<a href=#984 id=984 data-nosnippet>984</a> /// - Bucket-partitioned fields (e.g., `bucket(4, id)`) are read from the data file
-<a href=#985 id=985 data-nosnippet>985</a> /// - The source column `id` contains actual values (100, 200, 300), not constants
-<a href=#986 id=986 data-nosnippet>986</a> /// - Java's `PartitionUtil.constantsMap()` behavior is correctly replicated:
-<a href=#987 id=987 data-nosnippet>987</a> /// ```java
-<a href=#988 id=988 data-nosnippet>988</a> /// if (field.transform().isIdentity()) { // FALSE for bucket transforms
-<a href=#989 id=989 data-nosnippet>989</a> /// idToConstant.put(field.sourceId(), converted);
-<a href=#990 id=990 data-nosnippet>990</a> /// }
-<a href=#991 id=991 data-nosnippet>991</a> /// ```
-<a href=#992 id=992 data-nosnippet>992</a> ///
-<a href=#993 id=993 data-nosnippet>993</a> /// # Real-world impact
-<a href=#994 id=994 data-nosnippet>994</a> ///
-<a href=#995 id=995 data-nosnippet>995</a> /// This reproduces the failure scenario from Iceberg Java's TestRuntimeFiltering:
-<a href=#996 id=996 data-nosnippet>996</a> /// - Tables partitioned by `bucket(N, col)` are common for load balancing
-<a href=#997 id=997 data-nosnippet>997</a> /// - Queries filter on the source column: `SELECT * FROM tbl WHERE col = value`
-<a href=#998 id=998 data-nosnippet>998</a> /// - Runtime filtering pushes predicates down to Iceberg file scans
-<a href=#999 id=999 data-nosnippet>999</a> /// - Without this fix, the filter would match against constant partition values instead of data
-<a href=#1000 id=1000 data-nosnippet>1000</a> ///
-<a href=#1001 id=1001 data-nosnippet>1001</a> /// # References
-<a href=#1002 id=1002 data-nosnippet>1002</a> /// - Iceberg spec: format/spec.md "Column Projection" + "Partition Transforms"
-<a href=#1003 id=1003 data-nosnippet>1003</a> /// - Java impl: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java
-<a href=#1004 id=1004 data-nosnippet>1004</a> /// - Java test: spark/src/test/java/.../TestRuntimeFiltering.java
-<a href=#1005 id=1005 data-nosnippet>1005</a> </span><span class="attr">#[test]
-<a href=#1006 id=1006 data-nosnippet>1006</a> </span><span class="kw">fn </span>bucket_partitioning_reads_source_column_from_file() {
-<a href=#1007 id=1007 data-nosnippet>1007</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Struct, Transform};
-<a href=#1008 id=1008 data-nosnippet>1008</a>
-<a href=#1009 id=1009 data-nosnippet>1009</a> <span class="comment">// Table schema: id (data column), name (data column), id_bucket (partition column)
-<a href=#1010 id=1010 data-nosnippet>1010</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
-<a href=#1011 id=1011 data-nosnippet>1011</a> Schema::builder()
-<a href=#1012 id=1012 data-nosnippet>1012</a> .with_schema_id(<span class="number">0</span>)
-<a href=#1013 id=1013 data-nosnippet>1013</a> .with_fields(<span class="macro">vec!</span>[
-<a href=#1014 id=1014 data-nosnippet>1014</a> NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
-<a href=#1015 id=1015 data-nosnippet>1015</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#1016 id=1016 data-nosnippet>1016</a> ])
-<a href=#1017 id=1017 data-nosnippet>1017</a> .build()
-<a href=#1018 id=1018 data-nosnippet>1018</a> .unwrap(),
-<a href=#1019 id=1019 data-nosnippet>1019</a> );
-<a href=#1020 id=1020 data-nosnippet>1020</a>
-<a href=#1021 id=1021 data-nosnippet>1021</a> <span class="comment">// Partition spec: bucket(4, id) - the id field is bucketed
-<a href=#1022 id=1022 data-nosnippet>1022</a> </span><span class="kw">let </span>partition_spec = Arc::new(
-<a href=#1023 id=1023 data-nosnippet>1023</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
-<a href=#1024 id=1024 data-nosnippet>1024</a> .with_spec_id(<span class="number">0</span>)
-<a href=#1025 id=1025 data-nosnippet>1025</a> .add_partition_field(<span class="string">"id"</span>, <span class="string">"id_bucket"</span>, Transform::Bucket(<span class="number">4</span>))
-<a href=#1026 id=1026 data-nosnippet>1026</a> .unwrap()
-<a href=#1027 id=1027 data-nosnippet>1027</a> .build()
-<a href=#1028 id=1028 data-nosnippet>1028</a> .unwrap(),
-<a href=#1029 id=1029 data-nosnippet>1029</a> );
-<a href=#1030 id=1030 data-nosnippet>1030</a>
-<a href=#1031 id=1031 data-nosnippet>1031</a> <span class="comment">// Partition data: bucket value is 2
-<a href=#1032 id=1032 data-nosnippet>1032</a> // In Iceberg, partition data is a Struct where each field corresponds to a partition field
-<a href=#1033 id=1033 data-nosnippet>1033</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::int(<span class="number">2</span>))]);
-<a href=#1034 id=1034 data-nosnippet>1034</a>
-<a href=#1035 id=1035 data-nosnippet>1035</a> <span class="comment">// Parquet file contains both id and name columns
-<a href=#1036 id=1036 data-nosnippet>1036</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#1037 id=1037 data-nosnippet>1037</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
-<a href=#1038 id=1038 data-nosnippet>1038</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"2"</span>),
-<a href=#1039 id=1039 data-nosnippet>1039</a> ]));
-<a href=#1040 id=1040 data-nosnippet>1040</a>
-<a href=#1041 id=1041 data-nosnippet>1041</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>]; <span class="comment">// id, name
-<a href=#1042 id=1042 data-nosnippet>1042</a>
-<a href=#1043 id=1043 data-nosnippet>1043</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
-<a href=#1044 id=1044 data-nosnippet>1044</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
-<a href=#1045 id=1045 data-nosnippet>1045</a> .with_partition(partition_spec, partition_data)
-<a href=#1046 id=1046 data-nosnippet>1046</a> .build();
-<a href=#1047 id=1047 data-nosnippet>1047</a>
-<a href=#1048 id=1048 data-nosnippet>1048</a> <span class="comment">// Create a Parquet RecordBatch with actual data
-<a href=#1049 id=1049 data-nosnippet>1049</a> // The id column MUST be read from here, not treated as a constant
-<a href=#1050 id=1050 data-nosnippet>1050</a> </span><span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
-<a href=#1051 id=1051 data-nosnippet>1051</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>, <span class="number">300</span>])),
-<a href=#1052 id=1052 data-nosnippet>1052</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"Alice"</span>, <span class="string">"Bob"</span>, <span class="string">"Charlie"</span>])),
-<a href=#1053 id=1053 data-nosnippet>1053</a> ])
-<a href=#1054 id=1054 data-nosnippet>1054</a> .unwrap();
-<a href=#1055 id=1055 data-nosnippet>1055</a>
-<a href=#1056 id=1056 data-nosnippet>1056</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
-<a href=#1057 id=1057 data-nosnippet>1057</a>
-<a href=#1058 id=1058 data-nosnippet>1058</a> <span class="comment">// Verify the transformed RecordBatch correctly reads id from the file
-<a href=#1059 id=1059 data-nosnippet>1059</a> // (NOT as a constant from partition metadata)
-<a href=#1060 id=1060 data-nosnippet>1060</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">2</span>);
-<a href=#1061 id=1061 data-nosnippet>1061</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">3</span>);
-<a href=#1062 id=1062 data-nosnippet>1062</a>
-<a href=#1063 id=1063 data-nosnippet>1063</a> <span class="kw">let </span>id_column = result
-<a href=#1064 id=1064 data-nosnippet>1064</a> .column(<span class="number">0</span>)
-<a href=#1065 id=1065 data-nosnippet>1065</a> .as_any()
-<a href=#1066 id=1066 data-nosnippet>1066</a> .downcast_ref::<Int32Array>()
-<a href=#1067 id=1067 data-nosnippet>1067</a> .unwrap();
-<a href=#1068 id=1068 data-nosnippet>1068</a> <span class="comment">// These values MUST come from the Parquet file, not be replaced by constants
-<a href=#1069 id=1069 data-nosnippet>1069</a> </span><span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">100</span>);
-<a href=#1070 id=1070 data-nosnippet>1070</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">1</span>), <span class="number">200</span>);
-<a href=#1071 id=1071 data-nosnippet>1071</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">2</span>), <span class="number">300</span>);
-<a href=#1072 id=1072 data-nosnippet>1072</a>
-<a href=#1073 id=1073 data-nosnippet>1073</a> <span class="kw">let </span>name_column = result
-<a href=#1074 id=1074 data-nosnippet>1074</a> .column(<span class="number">1</span>)
-<a href=#1075 id=1075 data-nosnippet>1075</a> .as_any()
-<a href=#1076 id=1076 data-nosnippet>1076</a> .downcast_ref::<StringArray>()
-<a href=#1077 id=1077 data-nosnippet>1077</a> .unwrap();
-<a href=#1078 id=1078 data-nosnippet>1078</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
-<a href=#1079 id=1079 data-nosnippet>1079</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
-<a href=#1080 id=1080 data-nosnippet>1080</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">2</span>), <span class="string">"Charlie"</span>);
-<a href=#1081 id=1081 data-nosnippet>1081</a> }
-<a href=#1082 id=1082 data-nosnippet>1082</a>
-<a href=#1083 id=1083 data-nosnippet>1083</a> <span class="doccomment">/// Test that identity-transformed partition fields ARE treated as constants.
-<a href=#1084 id=1084 data-nosnippet>1084</a> ///
-<a href=#1085 id=1085 data-nosnippet>1085</a> /// This is the complement to `bucket_partitioning_reads_source_column_from_file`,
-<a href=#1086 id=1086 data-nosnippet>1086</a> /// verifying that constants_map() correctly identifies identity-transformed
-<a href=#1087 id=1087 data-nosnippet>1087</a> /// partition fields per the Iceberg spec.
-<a href=#1088 id=1088 data-nosnippet>1088</a> ///
-<a href=#1089 id=1089 data-nosnippet>1089</a> /// # Spec requirement (format/spec.md "Column Projection")
-<a href=#1090 id=1090 data-nosnippet>1090</a> ///
-<a href=#1091 id=1091 data-nosnippet>1091</a> /// > "Return the value from partition metadata if an Identity Transform exists for the field
-<a href=#1092 id=1092 data-nosnippet>1092</a> /// > and the partition value is present in the `partition` struct on `data_file` object
-<a href=#1093 id=1093 data-nosnippet>1093</a> /// > in the manifest. This allows for metadata only migrations of Hive tables."
-<a href=#1094 id=1094 data-nosnippet>1094</a> ///
-<a href=#1095 id=1095 data-nosnippet>1095</a> /// # Why identity transforms use constants
-<a href=#1096 id=1096 data-nosnippet>1096</a> ///
-<a href=#1097 id=1097 data-nosnippet>1097</a> /// Unlike bucket/truncate/year/etc., identity transforms don't modify the value:
-<a href=#1098 id=1098 data-nosnippet>1098</a> /// - `identity(dept)` stores the actual `dept` value in partition metadata
-<a href=#1099 id=1099 data-nosnippet>1099</a> /// - Partition metadata has `dept = "engineering"` (the real value, not a hash/bucket)
-<a href=#1100 id=1100 data-nosnippet>1100</a> /// - This value can be used directly without reading the data file
-<a href=#1101 id=1101 data-nosnippet>1101</a> ///
-<a href=#1102 id=1102 data-nosnippet>1102</a> /// # Performance benefit
-<a href=#1103 id=1103 data-nosnippet>1103</a> ///
-<a href=#1104 id=1104 data-nosnippet>1104</a> /// For Hive migrations where partition columns aren't in data files:
-<a href=#1105 id=1105 data-nosnippet>1105</a> /// - Partition metadata provides the column values
-<a href=#1106 id=1106 data-nosnippet>1106</a> /// - No need to read from data files (metadata-only query optimization)
-<a href=#1107 id=1107 data-nosnippet>1107</a> /// - Common pattern: `dept=engineering/subdept=backend/file.parquet`
-<a href=#1108 id=1108 data-nosnippet>1108</a> /// - `dept` and `subdept` are in directory structure, not in `file.parquet`
-<a href=#1109 id=1109 data-nosnippet>1109</a> /// - Iceberg populates these from partition metadata as constants
-<a href=#1110 id=1110 data-nosnippet>1110</a> ///
-<a href=#1111 id=1111 data-nosnippet>1111</a> /// # What this test verifies
-<a href=#1112 id=1112 data-nosnippet>1112</a> ///
-<a href=#1113 id=1113 data-nosnippet>1113</a> /// - Identity-partitioned fields use constants from partition metadata
-<a href=#1114 id=1114 data-nosnippet>1114</a> /// - The `dept` column is populated with `"engineering"` (not read from file)
-<a href=#1115 id=1115 data-nosnippet>1115</a> /// - Java's `PartitionUtil.constantsMap()` behavior is matched:
-<a href=#1116 id=1116 data-nosnippet>1116</a> /// ```java
-<a href=#1117 id=1117 data-nosnippet>1117</a> /// if (field.transform().isIdentity()) { // TRUE for identity
-<a href=#1118 id=1118 data-nosnippet>1118</a> /// idToConstant.put(field.sourceId(), converted);
-<a href=#1119 id=1119 data-nosnippet>1119</a> /// }
-<a href=#1120 id=1120 data-nosnippet>1120</a> /// ```
-<a href=#1121 id=1121 data-nosnippet>1121</a> ///
-<a href=#1122 id=1122 data-nosnippet>1122</a> /// # References
-<a href=#1123 id=1123 data-nosnippet>1123</a> /// - Iceberg spec: format/spec.md "Column Projection"
-<a href=#1124 id=1124 data-nosnippet>1124</a> /// - Java impl: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java
-<a href=#1125 id=1125 data-nosnippet>1125</a> </span><span class="attr">#[test]
-<a href=#1126 id=1126 data-nosnippet>1126</a> </span><span class="kw">fn </span>identity_partition_uses_constant_from_metadata() {
-<a href=#1127 id=1127 data-nosnippet>1127</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Struct, Transform};
-<a href=#1128 id=1128 data-nosnippet>1128</a>
-<a href=#1129 id=1129 data-nosnippet>1129</a> <span class="comment">// Table schema: id (data column), dept (partition column), name (data column)
-<a href=#1130 id=1130 data-nosnippet>1130</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
-<a href=#1131 id=1131 data-nosnippet>1131</a> Schema::builder()
-<a href=#1132 id=1132 data-nosnippet>1132</a> .with_schema_id(<span class="number">0</span>)
-<a href=#1133 id=1133 data-nosnippet>1133</a> .with_fields(<span class="macro">vec!</span>[
-<a href=#1134 id=1134 data-nosnippet>1134</a> NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
-<a href=#1135 id=1135 data-nosnippet>1135</a> NestedField::required(<span class="number">2</span>, <span class="string">"dept"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#1136 id=1136 data-nosnippet>1136</a> NestedField::optional(<span class="number">3</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#1137 id=1137 data-nosnippet>1137</a> ])
-<a href=#1138 id=1138 data-nosnippet>1138</a> .build()
-<a href=#1139 id=1139 data-nosnippet>1139</a> .unwrap(),
-<a href=#1140 id=1140 data-nosnippet>1140</a> );
-<a href=#1141 id=1141 data-nosnippet>1141</a>
-<a href=#1142 id=1142 data-nosnippet>1142</a> <span class="comment">// Partition spec: identity(dept) - the dept field uses identity transform
-<a href=#1143 id=1143 data-nosnippet>1143</a> </span><span class="kw">let </span>partition_spec = Arc::new(
-<a href=#1144 id=1144 data-nosnippet>1144</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
-<a href=#1145 id=1145 data-nosnippet>1145</a> .with_spec_id(<span class="number">0</span>)
-<a href=#1146 id=1146 data-nosnippet>1146</a> .add_partition_field(<span class="string">"dept"</span>, <span class="string">"dept"</span>, Transform::Identity)
-<a href=#1147 id=1147 data-nosnippet>1147</a> .unwrap()
-<a href=#1148 id=1148 data-nosnippet>1148</a> .build()
-<a href=#1149 id=1149 data-nosnippet>1149</a> .unwrap(),
-<a href=#1150 id=1150 data-nosnippet>1150</a> );
-<a href=#1151 id=1151 data-nosnippet>1151</a>
-<a href=#1152 id=1152 data-nosnippet>1152</a> <span class="comment">// Partition data: dept="engineering"
-<a href=#1153 id=1153 data-nosnippet>1153</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::string(<span class="string">"engineering"</span>))]);
-<a href=#1154 id=1154 data-nosnippet>1154</a>
-<a href=#1155 id=1155 data-nosnippet>1155</a> <span class="comment">// Parquet file contains only id and name (dept is in partition path)
-<a href=#1156 id=1156 data-nosnippet>1156</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#1157 id=1157 data-nosnippet>1157</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
-<a href=#1158 id=1158 data-nosnippet>1158</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"3"</span>),
-<a href=#1159 id=1159 data-nosnippet>1159</a> ]));
-<a href=#1160 id=1160 data-nosnippet>1160</a>
-<a href=#1161 id=1161 data-nosnippet>1161</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>]; <span class="comment">// id, dept, name
-<a href=#1162 id=1162 data-nosnippet>1162</a>
-<a href=#1163 id=1163 data-nosnippet>1163</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
-<a href=#1164 id=1164 data-nosnippet>1164</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
-<a href=#1165 id=1165 data-nosnippet>1165</a> .with_partition(partition_spec, partition_data)
-<a href=#1166 id=1166 data-nosnippet>1166</a> .build();
-<a href=#1167 id=1167 data-nosnippet>1167</a>
-<a href=#1168 id=1168 data-nosnippet>1168</a> <span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
-<a href=#1169 id=1169 data-nosnippet>1169</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>])),
-<a href=#1170 id=1170 data-nosnippet>1170</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"Alice"</span>, <span class="string">"Bob"</span>])),
-<a href=#1171 id=1171 data-nosnippet>1171</a> ])
-<a href=#1172 id=1172 data-nosnippet>1172</a> .unwrap();
-<a href=#1173 id=1173 data-nosnippet>1173</a>
-<a href=#1174 id=1174 data-nosnippet>1174</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
+<a href=#951 id=951 data-nosnippet>951</a> <span class="doccomment">/// Test for add_files with Parquet files that have NO field IDs (Hive tables).
+<a href=#952 id=952 data-nosnippet>952</a> ///
+<a href=#953 id=953 data-nosnippet>953</a> /// This reproduces the scenario from Iceberg spec where:
+<a href=#954 id=954 data-nosnippet>954</a> /// - Hive-style partitioned Parquet files are imported via add_files procedure
+<a href=#955 id=955 data-nosnippet>955</a> /// - Parquet files originally DO NOT have field IDs (typical for Hive tables)
+<a href=#956 id=956 data-nosnippet>956</a> /// - ArrowReader applies name mapping to assign correct Iceberg field IDs
+<a href=#957 id=957 data-nosnippet>957</a> /// - Iceberg schema assigns field IDs: id (1), name (2), dept (3), subdept (4)
+<a href=#958 id=958 data-nosnippet>958</a> /// - Partition columns (id, dept) have initial_default values
+<a href=#959 id=959 data-nosnippet>959</a> ///
+<a href=#960 id=960 data-nosnippet>960</a> /// Per the Iceberg spec (https://iceberg.apache.org/spec/#column-projection),
+<a href=#961 id=961 data-nosnippet>961</a> /// this scenario requires `schema.name-mapping.default` from table metadata
+<a href=#962 id=962 data-nosnippet>962</a> /// to correctly map Parquet columns by name to Iceberg field IDs.
+<a href=#963 id=963 data-nosnippet>963</a> /// This mapping is now applied in ArrowReader before data is processed.
+<a href=#964 id=964 data-nosnippet>964</a> ///
+<a href=#965 id=965 data-nosnippet>965</a> /// Expected behavior:
+<a href=#966 id=966 data-nosnippet>966</a> /// 1. id=1 (from initial_default) - spec rule #3
+<a href=#967 id=967 data-nosnippet>967</a> /// 2. name="John Doe" (from Parquet with field_id=2 assigned by reader) - found by field ID
+<a href=#968 id=968 data-nosnippet>968</a> /// 3. dept="hr" (from initial_default) - spec rule #3
+<a href=#969 id=969 data-nosnippet>969</a> /// 4. subdept="communications" (from Parquet with field_id=4 assigned by reader) - found by field ID
+<a href=#970 id=970 data-nosnippet>970</a> </span><span class="attr">#[test]
+<a href=#971 id=971 data-nosnippet>971</a> </span><span class="kw">fn </span>add_files_with_name_mapping_applied_in_reader() {
+<a href=#972 id=972 data-nosnippet>972</a> <span class="comment">// Iceberg schema after add_files: id (partition), name, dept (partition), subdept
+<a href=#973 id=973 data-nosnippet>973</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
+<a href=#974 id=974 data-nosnippet>974</a> Schema::builder()
+<a href=#975 id=975 data-nosnippet>975</a> .with_schema_id(<span class="number">0</span>)
+<a href=#976 id=976 data-nosnippet>976</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#977 id=977 data-nosnippet>977</a> NestedField::optional(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int))
+<a href=#978 id=978 data-nosnippet>978</a> .with_initial_default(Literal::int(<span class="number">1</span>))
+<a href=#979 id=979 data-nosnippet>979</a> .into(),
+<a href=#980 id=980 data-nosnippet>980</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#981 id=981 data-nosnippet>981</a> NestedField::optional(<span class="number">3</span>, <span class="string">"dept"</span>, Type::Primitive(PrimitiveType::String))
+<a href=#982 id=982 data-nosnippet>982</a> .with_initial_default(Literal::string(<span class="string">"hr"</span>))
+<a href=#983 id=983 data-nosnippet>983</a> .into(),
+<a href=#984 id=984 data-nosnippet>984</a> NestedField::optional(<span class="number">4</span>, <span class="string">"subdept"</span>, Type::Primitive(PrimitiveType::String))
+<a href=#985 id=985 data-nosnippet>985</a> .into(),
+<a href=#986 id=986 data-nosnippet>986</a> ])
+<a href=#987 id=987 data-nosnippet>987</a> .build()
+<a href=#988 id=988 data-nosnippet>988</a> .unwrap(),
+<a href=#989 id=989 data-nosnippet>989</a> );
+<a href=#990 id=990 data-nosnippet>990</a>
+<a href=#991 id=991 data-nosnippet>991</a> <span class="comment">// Simulate ArrowReader having applied name mapping:
+<a href=#992 id=992 data-nosnippet>992</a> // Original Parquet: name, subdept (NO field IDs)
+<a href=#993 id=993 data-nosnippet>993</a> // After reader.rs applies name mapping: name (field_id=2), subdept (field_id=4)
+<a href=#994 id=994 data-nosnippet>994</a> //
+<a href=#995 id=995 data-nosnippet>995</a> // Note: Partition columns (id, dept) are NOT in the Parquet file - they're in directory paths
+<a href=#996 id=996 data-nosnippet>996</a> </span><span class="kw">use </span>std::collections::HashMap;
+<a href=#997 id=997 data-nosnippet>997</a> <span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#998 id=998 data-nosnippet>998</a> Field::new(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>).with_metadata(HashMap::from([(
+<a href=#999 id=999 data-nosnippet>999</a> <span class="string">"PARQUET:field_id"</span>.to_string(),
+<a href=#1000 id=1000 data-nosnippet>1000</a> <span class="string">"2"</span>.to_string(),
+<a href=#1001 id=1001 data-nosnippet>1001</a> )])),
+<a href=#1002 id=1002 data-nosnippet>1002</a> Field::new(<span class="string">"subdept"</span>, DataType::Utf8, <span class="bool-val">true</span>).with_metadata(HashMap::from([(
+<a href=#1003 id=1003 data-nosnippet>1003</a> <span class="string">"PARQUET:field_id"</span>.to_string(),
+<a href=#1004 id=1004 data-nosnippet>1004</a> <span class="string">"4"</span>.to_string(),
+<a href=#1005 id=1005 data-nosnippet>1005</a> )])),
+<a href=#1006 id=1006 data-nosnippet>1006</a> ]));
+<a href=#1007 id=1007 data-nosnippet>1007</a>
+<a href=#1008 id=1008 data-nosnippet>1008</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>, <span class="number">4</span>]; <span class="comment">// id, name, dept, subdept
+<a href=#1009 id=1009 data-nosnippet>1009</a>
+<a href=#1010 id=1010 data-nosnippet>1010</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
+<a href=#1011 id=1011 data-nosnippet>1011</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids).build();
+<a href=#1012 id=1012 data-nosnippet>1012</a>
+<a href=#1013 id=1013 data-nosnippet>1013</a> <span class="comment">// Create a Parquet RecordBatch with data for: name="John Doe", subdept="communications"
+<a href=#1014 id=1014 data-nosnippet>1014</a> </span><span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
+<a href=#1015 id=1015 data-nosnippet>1015</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"John Doe"</span>])),
+<a href=#1016 id=1016 data-nosnippet>1016</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"communications"</span>])),
+<a href=#1017 id=1017 data-nosnippet>1017</a> ])
+<a href=#1018 id=1018 data-nosnippet>1018</a> .unwrap();
+<a href=#1019 id=1019 data-nosnippet>1019</a>
+<a href=#1020 id=1020 data-nosnippet>1020</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
+<a href=#1021 id=1021 data-nosnippet>1021</a>
+<a href=#1022 id=1022 data-nosnippet>1022</a> <span class="comment">// Verify the transformed RecordBatch has:
+<a href=#1023 id=1023 data-nosnippet>1023</a> // - id=1 (from initial_default, not from Parquet)
+<a href=#1024 id=1024 data-nosnippet>1024</a> // - name="John Doe" (from Parquet with correct field_id=2)
+<a href=#1025 id=1025 data-nosnippet>1025</a> // - dept="hr" (from initial_default, not from Parquet)
+<a href=#1026 id=1026 data-nosnippet>1026</a> // - subdept="communications" (from Parquet with correct field_id=4)
+<a href=#1027 id=1027 data-nosnippet>1027</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">4</span>);
+<a href=#1028 id=1028 data-nosnippet>1028</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">1</span>);
+<a href=#1029 id=1029 data-nosnippet>1029</a>
+<a href=#1030 id=1030 data-nosnippet>1030</a> <span class="kw">let </span>id_column = result
+<a href=#1031 id=1031 data-nosnippet>1031</a> .column(<span class="number">0</span>)
+<a href=#1032 id=1032 data-nosnippet>1032</a> .as_any()
+<a href=#1033 id=1033 data-nosnippet>1033</a> .downcast_ref::<Int32Array>()
+<a href=#1034 id=1034 data-nosnippet>1034</a> .unwrap();
+<a href=#1035 id=1035 data-nosnippet>1035</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">1</span>);
+<a href=#1036 id=1036 data-nosnippet>1036</a>
+<a href=#1037 id=1037 data-nosnippet>1037</a> <span class="kw">let </span>name_column = result
+<a href=#1038 id=1038 data-nosnippet>1038</a> .column(<span class="number">1</span>)
+<a href=#1039 id=1039 data-nosnippet>1039</a> .as_any()
+<a href=#1040 id=1040 data-nosnippet>1040</a> .downcast_ref::<StringArray>()
+<a href=#1041 id=1041 data-nosnippet>1041</a> .unwrap();
+<a href=#1042 id=1042 data-nosnippet>1042</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"John Doe"</span>);
+<a href=#1043 id=1043 data-nosnippet>1043</a>
+<a href=#1044 id=1044 data-nosnippet>1044</a> <span class="kw">let </span>dept_column = result
+<a href=#1045 id=1045 data-nosnippet>1045</a> .column(<span class="number">2</span>)
+<a href=#1046 id=1046 data-nosnippet>1046</a> .as_any()
+<a href=#1047 id=1047 data-nosnippet>1047</a> .downcast_ref::<StringArray>()
+<a href=#1048 id=1048 data-nosnippet>1048</a> .unwrap();
+<a href=#1049 id=1049 data-nosnippet>1049</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">0</span>), <span class="string">"hr"</span>);
+<a href=#1050 id=1050 data-nosnippet>1050</a>
+<a href=#1051 id=1051 data-nosnippet>1051</a> <span class="kw">let </span>subdept_column = result
+<a href=#1052 id=1052 data-nosnippet>1052</a> .column(<span class="number">3</span>)
+<a href=#1053 id=1053 data-nosnippet>1053</a> .as_any()
+<a href=#1054 id=1054 data-nosnippet>1054</a> .downcast_ref::<StringArray>()
+<a href=#1055 id=1055 data-nosnippet>1055</a> .unwrap();
+<a href=#1056 id=1056 data-nosnippet>1056</a> <span class="macro">assert_eq!</span>(subdept_column.value(<span class="number">0</span>), <span class="string">"communications"</span>);
+<a href=#1057 id=1057 data-nosnippet>1057</a> }
+<a href=#1058 id=1058 data-nosnippet>1058</a>
+<a href=#1059 id=1059 data-nosnippet>1059</a> <span class="doccomment">/// Test for bucket partitioning where source columns must be read from data files.
+<a href=#1060 id=1060 data-nosnippet>1060</a> ///
+<a href=#1061 id=1061 data-nosnippet>1061</a> /// This test verifies correct implementation of the Iceberg spec's "Column Projection" rules:
+<a href=#1062 id=1062 data-nosnippet>1062</a> /// > "Return the value from partition metadata if an **Identity Transform** exists for the field"
+<a href=#1063 id=1063 data-nosnippet>1063</a> ///
+<a href=#1064 id=1064 data-nosnippet>1064</a> /// # Why this test is critical
+<a href=#1065 id=1065 data-nosnippet>1065</a> ///
+<a href=#1066 id=1066 data-nosnippet>1066</a> /// The key insight is that partition metadata stores TRANSFORMED values, not source values:
+<a href=#1067 id=1067 data-nosnippet>1067</a> /// - For `bucket(4, id)`, partition metadata has `id_bucket = 2` (the bucket number)
+<a href=#1068 id=1068 data-nosnippet>1068</a> /// - The actual `id` column values (100, 200, 300) are ONLY in the data file
+<a href=#1069 id=1069 data-nosnippet>1069</a> ///
+<a href=#1070 id=1070 data-nosnippet>1070</a> /// If iceberg-rust incorrectly treated bucket-partitioned fields as constants, it would:
+<a href=#1071 id=1071 data-nosnippet>1071</a> /// 1. Replace all `id` values with the constant `2` from partition metadata
+<a href=#1072 id=1072 data-nosnippet>1072</a> /// 2. Break runtime filtering (e.g., `WHERE id = 100` would match no rows)
+<a href=#1073 id=1073 data-nosnippet>1073</a> /// 3. Return incorrect query results
+<a href=#1074 id=1074 data-nosnippet>1074</a> ///
+<a href=#1075 id=1075 data-nosnippet>1075</a> /// # What this test verifies
+<a href=#1076 id=1076 data-nosnippet>1076</a> ///
+<a href=#1077 id=1077 data-nosnippet>1077</a> /// - Bucket-partitioned fields (e.g., `bucket(4, id)`) are read from the data file
+<a href=#1078 id=1078 data-nosnippet>1078</a> /// - The source column `id` contains actual values (100, 200, 300), not constants
+<a href=#1079 id=1079 data-nosnippet>1079</a> /// - Java's `PartitionUtil.constantsMap()` behavior is correctly replicated:
+<a href=#1080 id=1080 data-nosnippet>1080</a> /// ```java
+<a href=#1081 id=1081 data-nosnippet>1081</a> /// if (field.transform().isIdentity()) { // FALSE for bucket transforms
+<a href=#1082 id=1082 data-nosnippet>1082</a> /// idToConstant.put(field.sourceId(), converted);
+<a href=#1083 id=1083 data-nosnippet>1083</a> /// }
+<a href=#1084 id=1084 data-nosnippet>1084</a> /// ```
+<a href=#1085 id=1085 data-nosnippet>1085</a> ///
+<a href=#1086 id=1086 data-nosnippet>1086</a> /// # Real-world impact
+<a href=#1087 id=1087 data-nosnippet>1087</a> ///
+<a href=#1088 id=1088 data-nosnippet>1088</a> /// This reproduces the failure scenario from Iceberg Java's TestRuntimeFiltering:
+<a href=#1089 id=1089 data-nosnippet>1089</a> /// - Tables partitioned by `bucket(N, col)` are common for load balancing
+<a href=#1090 id=1090 data-nosnippet>1090</a> /// - Queries filter on the source column: `SELECT * FROM tbl WHERE col = value`
+<a href=#1091 id=1091 data-nosnippet>1091</a> /// - Runtime filtering pushes predicates down to Iceberg file scans
+<a href=#1092 id=1092 data-nosnippet>1092</a> /// - Without this fix, the filter would match against constant partition values instead of data
+<a href=#1093 id=1093 data-nosnippet>1093</a> ///
+<a href=#1094 id=1094 data-nosnippet>1094</a> /// # References
+<a href=#1095 id=1095 data-nosnippet>1095</a> /// - Iceberg spec: format/spec.md "Column Projection" + "Partition Transforms"
+<a href=#1096 id=1096 data-nosnippet>1096</a> /// - Java impl: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java
+<a href=#1097 id=1097 data-nosnippet>1097</a> /// - Java test: spark/src/test/java/.../TestRuntimeFiltering.java
+<a href=#1098 id=1098 data-nosnippet>1098</a> </span><span class="attr">#[test]
+<a href=#1099 id=1099 data-nosnippet>1099</a> </span><span class="kw">fn </span>bucket_partitioning_reads_source_column_from_file() {
+<a href=#1100 id=1100 data-nosnippet>1100</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Struct, Transform};
+<a href=#1101 id=1101 data-nosnippet>1101</a>
+<a href=#1102 id=1102 data-nosnippet>1102</a> <span class="comment">// Table schema: id (data column), name (data column), id_bucket (partition column)
+<a href=#1103 id=1103 data-nosnippet>1103</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
+<a href=#1104 id=1104 data-nosnippet>1104</a> Schema::builder()
+<a href=#1105 id=1105 data-nosnippet>1105</a> .with_schema_id(<span class="number">0</span>)
+<a href=#1106 id=1106 data-nosnippet>1106</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#1107 id=1107 data-nosnippet>1107</a> NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
+<a href=#1108 id=1108 data-nosnippet>1108</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#1109 id=1109 data-nosnippet>1109</a> ])
+<a href=#1110 id=1110 data-nosnippet>1110</a> .build()
+<a href=#1111 id=1111 data-nosnippet>1111</a> .unwrap(),
+<a href=#1112 id=1112 data-nosnippet>1112</a> );
+<a href=#1113 id=1113 data-nosnippet>1113</a>
+<a href=#1114 id=1114 data-nosnippet>1114</a> <span class="comment">// Partition spec: bucket(4, id) - the id field is bucketed
+<a href=#1115 id=1115 data-nosnippet>1115</a> </span><span class="kw">let </span>partition_spec = Arc::new(
+<a href=#1116 id=1116 data-nosnippet>1116</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
+<a href=#1117 id=1117 data-nosnippet>1117</a> .with_spec_id(<span class="number">0</span>)
+<a href=#1118 id=1118 data-nosnippet>1118</a> .add_partition_field(<span class="string">"id"</span>, <span class="string">"id_bucket"</span>, Transform::Bucket(<span class="number">4</span>))
+<a href=#1119 id=1119 data-nosnippet>1119</a> .unwrap()
+<a href=#1120 id=1120 data-nosnippet>1120</a> .build()
+<a href=#1121 id=1121 data-nosnippet>1121</a> .unwrap(),
+<a href=#1122 id=1122 data-nosnippet>1122</a> );
+<a href=#1123 id=1123 data-nosnippet>1123</a>
+<a href=#1124 id=1124 data-nosnippet>1124</a> <span class="comment">// Partition data: bucket value is 2
+<a href=#1125 id=1125 data-nosnippet>1125</a> // In Iceberg, partition data is a Struct where each field corresponds to a partition field
+<a href=#1126 id=1126 data-nosnippet>1126</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::int(<span class="number">2</span>))]);
+<a href=#1127 id=1127 data-nosnippet>1127</a>
+<a href=#1128 id=1128 data-nosnippet>1128</a> <span class="comment">// Parquet file contains both id and name columns
+<a href=#1129 id=1129 data-nosnippet>1129</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#1130 id=1130 data-nosnippet>1130</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
+<a href=#1131 id=1131 data-nosnippet>1131</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"2"</span>),
+<a href=#1132 id=1132 data-nosnippet>1132</a> ]));
+<a href=#1133 id=1133 data-nosnippet>1133</a>
+<a href=#1134 id=1134 data-nosnippet>1134</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>]; <span class="comment">// id, name
+<a href=#1135 id=1135 data-nosnippet>1135</a>
+<a href=#1136 id=1136 data-nosnippet>1136</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
+<a href=#1137 id=1137 data-nosnippet>1137</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
+<a href=#1138 id=1138 data-nosnippet>1138</a> .with_partition(partition_spec, partition_data)
+<a href=#1139 id=1139 data-nosnippet>1139</a> .build();
+<a href=#1140 id=1140 data-nosnippet>1140</a>
+<a href=#1141 id=1141 data-nosnippet>1141</a> <span class="comment">// Create a Parquet RecordBatch with actual data
+<a href=#1142 id=1142 data-nosnippet>1142</a> // The id column MUST be read from here, not treated as a constant
+<a href=#1143 id=1143 data-nosnippet>1143</a> </span><span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
+<a href=#1144 id=1144 data-nosnippet>1144</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>, <span class="number">300</span>])),
+<a href=#1145 id=1145 data-nosnippet>1145</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"Alice"</span>, <span class="string">"Bob"</span>, <span class="string">"Charlie"</span>])),
+<a href=#1146 id=1146 data-nosnippet>1146</a> ])
+<a href=#1147 id=1147 data-nosnippet>1147</a> .unwrap();
+<a href=#1148 id=1148 data-nosnippet>1148</a>
+<a href=#1149 id=1149 data-nosnippet>1149</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
+<a href=#1150 id=1150 data-nosnippet>1150</a>
+<a href=#1151 id=1151 data-nosnippet>1151</a> <span class="comment">// Verify the transformed RecordBatch correctly reads id from the file
+<a href=#1152 id=1152 data-nosnippet>1152</a> // (NOT as a constant from partition metadata)
+<a href=#1153 id=1153 data-nosnippet>1153</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">2</span>);
+<a href=#1154 id=1154 data-nosnippet>1154</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">3</span>);
+<a href=#1155 id=1155 data-nosnippet>1155</a>
+<a href=#1156 id=1156 data-nosnippet>1156</a> <span class="kw">let </span>id_column = result
+<a href=#1157 id=1157 data-nosnippet>1157</a> .column(<span class="number">0</span>)
+<a href=#1158 id=1158 data-nosnippet>1158</a> .as_any()
+<a href=#1159 id=1159 data-nosnippet>1159</a> .downcast_ref::<Int32Array>()
+<a href=#1160 id=1160 data-nosnippet>1160</a> .unwrap();
+<a href=#1161 id=1161 data-nosnippet>1161</a> <span class="comment">// These values MUST come from the Parquet file, not be replaced by constants
+<a href=#1162 id=1162 data-nosnippet>1162</a> </span><span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">100</span>);
+<a href=#1163 id=1163 data-nosnippet>1163</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">1</span>), <span class="number">200</span>);
+<a href=#1164 id=1164 data-nosnippet>1164</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">2</span>), <span class="number">300</span>);
+<a href=#1165 id=1165 data-nosnippet>1165</a>
+<a href=#1166 id=1166 data-nosnippet>1166</a> <span class="kw">let </span>name_column = result
+<a href=#1167 id=1167 data-nosnippet>1167</a> .column(<span class="number">1</span>)
+<a href=#1168 id=1168 data-nosnippet>1168</a> .as_any()
+<a href=#1169 id=1169 data-nosnippet>1169</a> .downcast_ref::<StringArray>()
+<a href=#1170 id=1170 data-nosnippet>1170</a> .unwrap();
+<a href=#1171 id=1171 data-nosnippet>1171</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
+<a href=#1172 id=1172 data-nosnippet>1172</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
+<a href=#1173 id=1173 data-nosnippet>1173</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">2</span>), <span class="string">"Charlie"</span>);
+<a href=#1174 id=1174 data-nosnippet>1174</a> }
<a href=#1175 id=1175 data-nosnippet>1175</a>
-<a href=#1176 id=1176 data-nosnippet>1176</a> <span class="comment">// Verify the dept column is populated with the constant from partition metadata
-<a href=#1177 id=1177 data-nosnippet>1177</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">3</span>);
-<a href=#1178 id=1178 data-nosnippet>1178</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">2</span>);
-<a href=#1179 id=1179 data-nosnippet>1179</a>
-<a href=#1180 id=1180 data-nosnippet>1180</a> <span class="kw">let </span>id_column = result
-<a href=#1181 id=1181 data-nosnippet>1181</a> .column(<span class="number">0</span>)
-<a href=#1182 id=1182 data-nosnippet>1182</a> .as_any()
-<a href=#1183 id=1183 data-nosnippet>1183</a> .downcast_ref::<Int32Array>()
-<a href=#1184 id=1184 data-nosnippet>1184</a> .unwrap();
-<a href=#1185 id=1185 data-nosnippet>1185</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">100</span>);
-<a href=#1186 id=1186 data-nosnippet>1186</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">1</span>), <span class="number">200</span>);
-<a href=#1187 id=1187 data-nosnippet>1187</a>
-<a href=#1188 id=1188 data-nosnippet>1188</a> <span class="kw">let </span>dept_column = result
-<a href=#1189 id=1189 data-nosnippet>1189</a> .column(<span class="number">1</span>)
-<a href=#1190 id=1190 data-nosnippet>1190</a> .as_any()
-<a href=#1191 id=1191 data-nosnippet>1191</a> .downcast_ref::<StringArray>()
-<a href=#1192 id=1192 data-nosnippet>1192</a> .unwrap();
-<a href=#1193 id=1193 data-nosnippet>1193</a> <span class="comment">// This value MUST come from partition metadata (constant)
-<a href=#1194 id=1194 data-nosnippet>1194</a> </span><span class="macro">assert_eq!</span>(dept_column.value(<span class="number">0</span>), <span class="string">"engineering"</span>);
-<a href=#1195 id=1195 data-nosnippet>1195</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">1</span>), <span class="string">"engineering"</span>);
-<a href=#1196 id=1196 data-nosnippet>1196</a>
-<a href=#1197 id=1197 data-nosnippet>1197</a> <span class="kw">let </span>name_column = result
-<a href=#1198 id=1198 data-nosnippet>1198</a> .column(<span class="number">2</span>)
-<a href=#1199 id=1199 data-nosnippet>1199</a> .as_any()
-<a href=#1200 id=1200 data-nosnippet>1200</a> .downcast_ref::<StringArray>()
-<a href=#1201 id=1201 data-nosnippet>1201</a> .unwrap();
-<a href=#1202 id=1202 data-nosnippet>1202</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
-<a href=#1203 id=1203 data-nosnippet>1203</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
-<a href=#1204 id=1204 data-nosnippet>1204</a> }
-<a href=#1205 id=1205 data-nosnippet>1205</a>
-<a href=#1206 id=1206 data-nosnippet>1206</a> <span class="doccomment">/// Test bucket partitioning with renamed source column.
-<a href=#1207 id=1207 data-nosnippet>1207</a> ///
-<a href=#1208 id=1208 data-nosnippet>1208</a> /// This verifies correct behavior for TestRuntimeFiltering.testRenamedSourceColumnTable() in Iceberg Java.
-<a href=#1209 id=1209 data-nosnippet>1209</a> /// When a source column is renamed after partitioning is established, field-ID-based mapping
-<a href=#1210 id=1210 data-nosnippet>1210</a> /// must still correctly identify the column in Parquet files.
-<a href=#1211 id=1211 data-nosnippet>1211</a> ///
-<a href=#1212 id=1212 data-nosnippet>1212</a> /// # Scenario
-<a href=#1213 id=1213 data-nosnippet>1213</a> ///
-<a href=#1214 id=1214 data-nosnippet>1214</a> /// 1. Table created with `bucket(4, id)` partitioning
-<a href=#1215 id=1215 data-nosnippet>1215</a> /// 2. Data written to Parquet files (field_id=1, name="id")
-<a href=#1216 id=1216 data-nosnippet>1216</a> /// 3. Column renamed: `ALTER TABLE ... RENAME COLUMN id TO row_id`
-<a href=#1217 id=1217 data-nosnippet>1217</a> /// 4. Iceberg schema now has: field_id=1, name="row_id"
-<a href=#1218 id=1218 data-nosnippet>1218</a> /// 5. Parquet files still have: field_id=1, name="id"
-<a href=#1219 id=1219 data-nosnippet>1219</a> ///
-<a href=#1220 id=1220 data-nosnippet>1220</a> /// # Expected Behavior Per Iceberg Spec
-<a href=#1221 id=1221 data-nosnippet>1221</a> ///
-<a href=#1222 id=1222 data-nosnippet>1222</a> /// Per the Iceberg spec "Column Projection" section and Java's PartitionUtil.constantsMap():
-<a href=#1223 id=1223 data-nosnippet>1223</a> /// - Bucket transforms are NON-identity, so partition metadata stores bucket numbers (0-3), not source values
-<a href=#1224 id=1224 data-nosnippet>1224</a> /// - Source columns for non-identity transforms MUST be read from data files
-<a href=#1225 id=1225 data-nosnippet>1225</a> /// - Field-ID-based mapping should find the column by field_id=1 (ignoring name mismatch)
-<a href=#1226 id=1226 data-nosnippet>1226</a> /// - Runtime filtering on `row_id` should work correctly
-<a href=#1227 id=1227 data-nosnippet>1227</a> ///
-<a href=#1228 id=1228 data-nosnippet>1228</a> /// # What This Tests
-<a href=#1229 id=1229 data-nosnippet>1229</a> ///
-<a href=#1230 id=1230 data-nosnippet>1230</a> /// This test ensures that when FileScanTask provides partition_spec and partition_data:
-<a href=#1231 id=1231 data-nosnippet>1231</a> /// - constants_map() correctly identifies that bucket(4, row_id) is NOT an identity transform
-<a href=#1232 id=1232 data-nosnippet>1232</a> /// - The source column (field_id=1) is NOT added to constants_map
-<a href=#1233 id=1233 data-nosnippet>1233</a> /// - Field-ID-based mapping reads actual values from the Parquet file
-<a href=#1234 id=1234 data-nosnippet>1234</a> /// - Values [100, 200, 300] are read, not replaced with bucket constant 2
-<a href=#1235 id=1235 data-nosnippet>1235</a> ///
-<a href=#1236 id=1236 data-nosnippet>1236</a> /// # References
-<a href=#1237 id=1237 data-nosnippet>1237</a> /// - Java test: spark/src/test/java/.../TestRuntimeFiltering.java::testRenamedSourceColumnTable
-<a href=#1238 id=1238 data-nosnippet>1238</a> /// - Java impl: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java::constantsMap()
-<a href=#1239 id=1239 data-nosnippet>1239</a> /// - Iceberg spec: format/spec.md "Column Projection" section
-<a href=#1240 id=1240 data-nosnippet>1240</a> </span><span class="attr">#[test]
-<a href=#1241 id=1241 data-nosnippet>1241</a> </span><span class="kw">fn </span>test_bucket_partitioning_with_renamed_source_column() {
-<a href=#1242 id=1242 data-nosnippet>1242</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Struct, Transform};
-<a href=#1243 id=1243 data-nosnippet>1243</a>
-<a href=#1244 id=1244 data-nosnippet>1244</a> <span class="comment">// Iceberg schema after rename: row_id (was id), name
-<a href=#1245 id=1245 data-nosnippet>1245</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
-<a href=#1246 id=1246 data-nosnippet>1246</a> Schema::builder()
-<a href=#1247 id=1247 data-nosnippet>1247</a> .with_schema_id(<span class="number">0</span>)
-<a href=#1248 id=1248 data-nosnippet>1248</a> .with_fields(<span class="macro">vec!</span>[
-<a href=#1249 id=1249 data-nosnippet>1249</a> NestedField::required(<span class="number">1</span>, <span class="string">"row_id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
-<a href=#1250 id=1250 data-nosnippet>1250</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#1251 id=1251 data-nosnippet>1251</a> ])
-<a href=#1252 id=1252 data-nosnippet>1252</a> .build()
-<a href=#1253 id=1253 data-nosnippet>1253</a> .unwrap(),
-<a href=#1254 id=1254 data-nosnippet>1254</a> );
+<a href=#1176 id=1176 data-nosnippet>1176</a> <span class="doccomment">/// Test that identity-transformed partition fields ARE treated as constants.
+<a href=#1177 id=1177 data-nosnippet>1177</a> ///
+<a href=#1178 id=1178 data-nosnippet>1178</a> /// This is the complement to `bucket_partitioning_reads_source_column_from_file`,
+<a href=#1179 id=1179 data-nosnippet>1179</a> /// verifying that constants_map() correctly identifies identity-transformed
+<a href=#1180 id=1180 data-nosnippet>1180</a> /// partition fields per the Iceberg spec.
+<a href=#1181 id=1181 data-nosnippet>1181</a> ///
+<a href=#1182 id=1182 data-nosnippet>1182</a> /// # Spec requirement (format/spec.md "Column Projection")
+<a href=#1183 id=1183 data-nosnippet>1183</a> ///
+<a href=#1184 id=1184 data-nosnippet>1184</a> /// > "Return the value from partition metadata if an Identity Transform exists for the field
+<a href=#1185 id=1185 data-nosnippet>1185</a> /// > and the partition value is present in the `partition` struct on `data_file` object
+<a href=#1186 id=1186 data-nosnippet>1186</a> /// > in the manifest. This allows for metadata only migrations of Hive tables."
+<a href=#1187 id=1187 data-nosnippet>1187</a> ///
+<a href=#1188 id=1188 data-nosnippet>1188</a> /// # Why identity transforms use constants
+<a href=#1189 id=1189 data-nosnippet>1189</a> ///
+<a href=#1190 id=1190 data-nosnippet>1190</a> /// Unlike bucket/truncate/year/etc., identity transforms don't modify the value:
+<a href=#1191 id=1191 data-nosnippet>1191</a> /// - `identity(dept)` stores the actual `dept` value in partition metadata
+<a href=#1192 id=1192 data-nosnippet>1192</a> /// - Partition metadata has `dept = "engineering"` (the real value, not a hash/bucket)
+<a href=#1193 id=1193 data-nosnippet>1193</a> /// - This value can be used directly without reading the data file
+<a href=#1194 id=1194 data-nosnippet>1194</a> ///
+<a href=#1195 id=1195 data-nosnippet>1195</a> /// # Performance benefit
+<a href=#1196 id=1196 data-nosnippet>1196</a> ///
+<a href=#1197 id=1197 data-nosnippet>1197</a> /// For Hive migrations where partition columns aren't in data files:
+<a href=#1198 id=1198 data-nosnippet>1198</a> /// - Partition metadata provides the column values
+<a href=#1199 id=1199 data-nosnippet>1199</a> /// - No need to read from data files (metadata-only query optimization)
+<a href=#1200 id=1200 data-nosnippet>1200</a> /// - Common pattern: `dept=engineering/subdept=backend/file.parquet`
+<a href=#1201 id=1201 data-nosnippet>1201</a> /// - `dept` and `subdept` are in directory structure, not in `file.parquet`
+<a href=#1202 id=1202 data-nosnippet>1202</a> /// - Iceberg populates these from partition metadata as constants
+<a href=#1203 id=1203 data-nosnippet>1203</a> ///
+<a href=#1204 id=1204 data-nosnippet>1204</a> /// # What this test verifies
+<a href=#1205 id=1205 data-nosnippet>1205</a> ///
+<a href=#1206 id=1206 data-nosnippet>1206</a> /// - Identity-partitioned fields use constants from partition metadata
+<a href=#1207 id=1207 data-nosnippet>1207</a> /// - The `dept` column is populated with `"engineering"` (not read from file)
+<a href=#1208 id=1208 data-nosnippet>1208</a> /// - Java's `PartitionUtil.constantsMap()` behavior is matched:
+<a href=#1209 id=1209 data-nosnippet>1209</a> /// ```java
+<a href=#1210 id=1210 data-nosnippet>1210</a> /// if (field.transform().isIdentity()) { // TRUE for identity
+<a href=#1211 id=1211 data-nosnippet>1211</a> /// idToConstant.put(field.sourceId(), converted);
+<a href=#1212 id=1212 data-nosnippet>1212</a> /// }
+<a href=#1213 id=1213 data-nosnippet>1213</a> /// ```
+<a href=#1214 id=1214 data-nosnippet>1214</a> ///
+<a href=#1215 id=1215 data-nosnippet>1215</a> /// # References
+<a href=#1216 id=1216 data-nosnippet>1216</a> /// - Iceberg spec: format/spec.md "Column Projection"
+<a href=#1217 id=1217 data-nosnippet>1217</a> /// - Java impl: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java
+<a href=#1218 id=1218 data-nosnippet>1218</a> </span><span class="attr">#[test]
+<a href=#1219 id=1219 data-nosnippet>1219</a> </span><span class="kw">fn </span>identity_partition_uses_constant_from_metadata() {
+<a href=#1220 id=1220 data-nosnippet>1220</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Struct, Transform};
+<a href=#1221 id=1221 data-nosnippet>1221</a>
+<a href=#1222 id=1222 data-nosnippet>1222</a> <span class="comment">// Table schema: id (data column), dept (partition column), name (data column)
+<a href=#1223 id=1223 data-nosnippet>1223</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
+<a href=#1224 id=1224 data-nosnippet>1224</a> Schema::builder()
+<a href=#1225 id=1225 data-nosnippet>1225</a> .with_schema_id(<span class="number">0</span>)
+<a href=#1226 id=1226 data-nosnippet>1226</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#1227 id=1227 data-nosnippet>1227</a> NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
+<a href=#1228 id=1228 data-nosnippet>1228</a> NestedField::required(<span class="number">2</span>, <span class="string">"dept"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#1229 id=1229 data-nosnippet>1229</a> NestedField::optional(<span class="number">3</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#1230 id=1230 data-nosnippet>1230</a> ])
+<a href=#1231 id=1231 data-nosnippet>1231</a> .build()
+<a href=#1232 id=1232 data-nosnippet>1232</a> .unwrap(),
+<a href=#1233 id=1233 data-nosnippet>1233</a> );
+<a href=#1234 id=1234 data-nosnippet>1234</a>
+<a href=#1235 id=1235 data-nosnippet>1235</a> <span class="comment">// Partition spec: identity(dept) - the dept field uses identity transform
+<a href=#1236 id=1236 data-nosnippet>1236</a> </span><span class="kw">let </span>partition_spec = Arc::new(
+<a href=#1237 id=1237 data-nosnippet>1237</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
+<a href=#1238 id=1238 data-nosnippet>1238</a> .with_spec_id(<span class="number">0</span>)
+<a href=#1239 id=1239 data-nosnippet>1239</a> .add_partition_field(<span class="string">"dept"</span>, <span class="string">"dept"</span>, Transform::Identity)
+<a href=#1240 id=1240 data-nosnippet>1240</a> .unwrap()
+<a href=#1241 id=1241 data-nosnippet>1241</a> .build()
+<a href=#1242 id=1242 data-nosnippet>1242</a> .unwrap(),
+<a href=#1243 id=1243 data-nosnippet>1243</a> );
+<a href=#1244 id=1244 data-nosnippet>1244</a>
+<a href=#1245 id=1245 data-nosnippet>1245</a> <span class="comment">// Partition data: dept="engineering"
+<a href=#1246 id=1246 data-nosnippet>1246</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::string(<span class="string">"engineering"</span>))]);
+<a href=#1247 id=1247 data-nosnippet>1247</a>
+<a href=#1248 id=1248 data-nosnippet>1248</a> <span class="comment">// Parquet file contains only id and name (dept is in partition path)
+<a href=#1249 id=1249 data-nosnippet>1249</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#1250 id=1250 data-nosnippet>1250</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
+<a href=#1251 id=1251 data-nosnippet>1251</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"3"</span>),
+<a href=#1252 id=1252 data-nosnippet>1252</a> ]));
+<a href=#1253 id=1253 data-nosnippet>1253</a>
+<a href=#1254 id=1254 data-nosnippet>1254</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>]; <span class="comment">// id, dept, name
<a href=#1255 id=1255 data-nosnippet>1255</a>
-<a href=#1256 id=1256 data-nosnippet>1256</a> <span class="comment">// Partition spec: bucket(4, row_id) - but source_id still points to field_id=1
-<a href=#1257 id=1257 data-nosnippet>1257</a> </span><span class="kw">let </span>partition_spec = Arc::new(
-<a href=#1258 id=1258 data-nosnippet>1258</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
-<a href=#1259 id=1259 data-nosnippet>1259</a> .with_spec_id(<span class="number">0</span>)
-<a href=#1260 id=1260 data-nosnippet>1260</a> .add_partition_field(<span class="string">"row_id"</span>, <span class="string">"row_id_bucket"</span>, Transform::Bucket(<span class="number">4</span>))
-<a href=#1261 id=1261 data-nosnippet>1261</a> .unwrap()
-<a href=#1262 id=1262 data-nosnippet>1262</a> .build()
-<a href=#1263 id=1263 data-nosnippet>1263</a> .unwrap(),
-<a href=#1264 id=1264 data-nosnippet>1264</a> );
-<a href=#1265 id=1265 data-nosnippet>1265</a>
-<a href=#1266 id=1266 data-nosnippet>1266</a> <span class="comment">// Partition data: bucket value is 2
-<a href=#1267 id=1267 data-nosnippet>1267</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::int(<span class="number">2</span>))]);
+<a href=#1256 id=1256 data-nosnippet>1256</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
+<a href=#1257 id=1257 data-nosnippet>1257</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
+<a href=#1258 id=1258 data-nosnippet>1258</a> .with_partition(partition_spec, partition_data)
+<a href=#1259 id=1259 data-nosnippet>1259</a> .build();
+<a href=#1260 id=1260 data-nosnippet>1260</a>
+<a href=#1261 id=1261 data-nosnippet>1261</a> <span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
+<a href=#1262 id=1262 data-nosnippet>1262</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>])),
+<a href=#1263 id=1263 data-nosnippet>1263</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"Alice"</span>, <span class="string">"Bob"</span>])),
+<a href=#1264 id=1264 data-nosnippet>1264</a> ])
+<a href=#1265 id=1265 data-nosnippet>1265</a> .unwrap();
+<a href=#1266 id=1266 data-nosnippet>1266</a>
+<a href=#1267 id=1267 data-nosnippet>1267</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
<a href=#1268 id=1268 data-nosnippet>1268</a>
-<a href=#1269 id=1269 data-nosnippet>1269</a> <span class="comment">// Parquet file has OLD column name "id" but SAME field_id=1
-<a href=#1270 id=1270 data-nosnippet>1270</a> // Field-ID-based mapping should find this despite name mismatch
-<a href=#1271 id=1271 data-nosnippet>1271</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#1272 id=1272 data-nosnippet>1272</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
-<a href=#1273 id=1273 data-nosnippet>1273</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"2"</span>),
-<a href=#1274 id=1274 data-nosnippet>1274</a> ]));
-<a href=#1275 id=1275 data-nosnippet>1275</a>
-<a href=#1276 id=1276 data-nosnippet>1276</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>]; <span class="comment">// row_id (field_id=1), name (field_id=2)
-<a href=#1277 id=1277 data-nosnippet>1277</a>
-<a href=#1278 id=1278 data-nosnippet>1278</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
-<a href=#1279 id=1279 data-nosnippet>1279</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
-<a href=#1280 id=1280 data-nosnippet>1280</a> .with_partition(partition_spec, partition_data)
-<a href=#1281 id=1281 data-nosnippet>1281</a> .build();
-<a href=#1282 id=1282 data-nosnippet>1282</a>
-<a href=#1283 id=1283 data-nosnippet>1283</a> <span class="comment">// Create a Parquet RecordBatch with actual data
-<a href=#1284 id=1284 data-nosnippet>1284</a> // Despite column rename, data should be read via field_id=1
-<a href=#1285 id=1285 data-nosnippet>1285</a> </span><span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
-<a href=#1286 id=1286 data-nosnippet>1286</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>, <span class="number">300</span>])),
-<a href=#1287 id=1287 data-nosnippet>1287</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"Alice"</span>, <span class="string">"Bob"</span>, <span class="string">"Charlie"</span>])),
-<a href=#1288 id=1288 data-nosnippet>1288</a> ])
-<a href=#1289 id=1289 data-nosnippet>1289</a> .unwrap();
-<a href=#1290 id=1290 data-nosnippet>1290</a>
-<a href=#1291 id=1291 data-nosnippet>1291</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
-<a href=#1292 id=1292 data-nosnippet>1292</a>
-<a href=#1293 id=1293 data-nosnippet>1293</a> <span class="comment">// Verify the transformed RecordBatch correctly reads data despite name mismatch
-<a href=#1294 id=1294 data-nosnippet>1294</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">2</span>);
-<a href=#1295 id=1295 data-nosnippet>1295</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">3</span>);
-<a href=#1296 id=1296 data-nosnippet>1296</a>
-<a href=#1297 id=1297 data-nosnippet>1297</a> <span class="kw">let </span>row_id_column = result
-<a href=#1298 id=1298 data-nosnippet>1298</a> .column(<span class="number">0</span>)
-<a href=#1299 id=1299 data-nosnippet>1299</a> .as_any()
-<a href=#1300 id=1300 data-nosnippet>1300</a> .downcast_ref::<Int32Array>()
-<a href=#1301 id=1301 data-nosnippet>1301</a> .unwrap();
-<a href=#1302 id=1302 data-nosnippet>1302</a> <span class="comment">// These values MUST come from the Parquet file via field_id=1,
-<a href=#1303 id=1303 data-nosnippet>1303</a> // not be replaced by the bucket constant (2)
-<a href=#1304 id=1304 data-nosnippet>1304</a> </span><span class="macro">assert_eq!</span>(row_id_column.value(<span class="number">0</span>), <span class="number">100</span>);
-<a href=#1305 id=1305 data-nosnippet>1305</a> <span class="macro">assert_eq!</span>(row_id_column.value(<span class="number">1</span>), <span class="number">200</span>);
-<a href=#1306 id=1306 data-nosnippet>1306</a> <span class="macro">assert_eq!</span>(row_id_column.value(<span class="number">2</span>), <span class="number">300</span>);
-<a href=#1307 id=1307 data-nosnippet>1307</a>
-<a href=#1308 id=1308 data-nosnippet>1308</a> <span class="kw">let </span>name_column = result
-<a href=#1309 id=1309 data-nosnippet>1309</a> .column(<span class="number">1</span>)
-<a href=#1310 id=1310 data-nosnippet>1310</a> .as_any()
-<a href=#1311 id=1311 data-nosnippet>1311</a> .downcast_ref::<StringArray>()
-<a href=#1312 id=1312 data-nosnippet>1312</a> .unwrap();
-<a href=#1313 id=1313 data-nosnippet>1313</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
-<a href=#1314 id=1314 data-nosnippet>1314</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
-<a href=#1315 id=1315 data-nosnippet>1315</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">2</span>), <span class="string">"Charlie"</span>);
-<a href=#1316 id=1316 data-nosnippet>1316</a> }
-<a href=#1317 id=1317 data-nosnippet>1317</a>
-<a href=#1318 id=1318 data-nosnippet>1318</a> <span class="doccomment">/// Comprehensive integration test that verifies all 4 Iceberg spec rules work correctly.
-<a href=#1319 id=1319 data-nosnippet>1319</a> ///
-<a href=#1320 id=1320 data-nosnippet>1320</a> /// Per the Iceberg spec (https://iceberg.apache.org/spec/#column-projection),
-<a href=#1321 id=1321 data-nosnippet>1321</a> /// "Values for field ids which are not present in a data file must be resolved
-<a href=#1322 id=1322 data-nosnippet>1322</a> /// according the following rules:"
-<a href=#1323 id=1323 data-nosnippet>1323</a> ///
-<a href=#1324 id=1324 data-nosnippet>1324</a> /// This test creates a scenario where each rule is exercised:
-<a href=#1325 id=1325 data-nosnippet>1325</a> /// - Rule #1: dept (identity-partitioned) -> constant from partition metadata
-<a href=#1326 id=1326 data-nosnippet>1326</a> /// - Rule #2: data (via name mapping) -> read from Parquet file by name
-<a href=#1327 id=1327 data-nosnippet>1327</a> /// - Rule #3: category (initial_default) -> use default value
-<a href=#1328 id=1328 data-nosnippet>1328</a> /// - Rule #4: notes (no default) -> return null
-<a href=#1329 id=1329 data-nosnippet>1329</a> ///
-<a href=#1330 id=1330 data-nosnippet>1330</a> /// # References
-<a href=#1331 id=1331 data-nosnippet>1331</a> /// - Iceberg spec: format/spec.md "Column Projection" section
-<a href=#1332 id=1332 data-nosnippet>1332</a> </span><span class="attr">#[test]
-<a href=#1333 id=1333 data-nosnippet>1333</a> </span><span class="kw">fn </span>test_all_four_spec_rules() {
-<a href=#1334 id=1334 data-nosnippet>1334</a> <span class="kw">use </span><span class="kw">crate</span>::spec::Transform;
-<a href=#1335 id=1335 data-nosnippet>1335</a>
-<a href=#1336 id=1336 data-nosnippet>1336</a> <span class="comment">// Iceberg schema with columns designed to exercise each spec rule
-<a href=#1337 id=1337 data-nosnippet>1337</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
-<a href=#1338 id=1338 data-nosnippet>1338</a> Schema::builder()
-<a href=#1339 id=1339 data-nosnippet>1339</a> .with_schema_id(<span class="number">0</span>)
-<a href=#1340 id=1340 data-nosnippet>1340</a> .with_fields(<span class="macro">vec!</span>[
-<a href=#1341 id=1341 data-nosnippet>1341</a> <span class="comment">// Field in Parquet by field ID (normal case)
-<a href=#1342 id=1342 data-nosnippet>1342</a> </span>NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
-<a href=#1343 id=1343 data-nosnippet>1343</a> <span class="comment">// Rule #1: Identity-partitioned field - should use partition metadata
-<a href=#1344 id=1344 data-nosnippet>1344</a> </span>NestedField::required(<span class="number">2</span>, <span class="string">"dept"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#1345 id=1345 data-nosnippet>1345</a> <span class="comment">// Rule #2: Field resolved by name mapping (ArrowReader already applied)
-<a href=#1346 id=1346 data-nosnippet>1346</a> </span>NestedField::required(<span class="number">3</span>, <span class="string">"data"</span>, Type::Primitive(PrimitiveType::String)).into(),
-<a href=#1347 id=1347 data-nosnippet>1347</a> <span class="comment">// Rule #3: Field with initial_default
-<a href=#1348 id=1348 data-nosnippet>1348</a> </span>NestedField::optional(<span class="number">4</span>, <span class="string">"category"</span>, Type::Primitive(PrimitiveType::String))
-<a href=#1349 id=1349 data-nosnippet>1349</a> .with_initial_default(Literal::string(<span class="string">"default_category"</span>))
-<a href=#1350 id=1350 data-nosnippet>1350</a> .into(),
-<a href=#1351 id=1351 data-nosnippet>1351</a> <span class="comment">// Rule #4: Field with no default - should be null
-<a href=#1352 id=1352 data-nosnippet>1352</a> </span>NestedField::optional(<span class="number">5</span>, <span class="string">"notes"</span>, Type::Primitive(PrimitiveType::String))
-<a href=#1353 id=1353 data-nosnippet>1353</a> .into(),
-<a href=#1354 id=1354 data-nosnippet>1354</a> ])
+<a href=#1269 id=1269 data-nosnippet>1269</a> <span class="comment">// Verify the dept column is populated with the constant from partition metadata
+<a href=#1270 id=1270 data-nosnippet>1270</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">3</span>);
+<a href=#1271 id=1271 data-nosnippet>1271</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">2</span>);
+<a href=#1272 id=1272 data-nosnippet>1272</a>
+<a href=#1273 id=1273 data-nosnippet>1273</a> <span class="kw">let </span>id_column = result
+<a href=#1274 id=1274 data-nosnippet>1274</a> .column(<span class="number">0</span>)
+<a href=#1275 id=1275 data-nosnippet>1275</a> .as_any()
+<a href=#1276 id=1276 data-nosnippet>1276</a> .downcast_ref::<Int32Array>()
+<a href=#1277 id=1277 data-nosnippet>1277</a> .unwrap();
+<a href=#1278 id=1278 data-nosnippet>1278</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">100</span>);
+<a href=#1279 id=1279 data-nosnippet>1279</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">1</span>), <span class="number">200</span>);
+<a href=#1280 id=1280 data-nosnippet>1280</a>
+<a href=#1281 id=1281 data-nosnippet>1281</a> <span class="kw">let </span>dept_column = result
+<a href=#1282 id=1282 data-nosnippet>1282</a> .column(<span class="number">1</span>)
+<a href=#1283 id=1283 data-nosnippet>1283</a> .as_any()
+<a href=#1284 id=1284 data-nosnippet>1284</a> .downcast_ref::<StringArray>()
+<a href=#1285 id=1285 data-nosnippet>1285</a> .unwrap();
+<a href=#1286 id=1286 data-nosnippet>1286</a> <span class="comment">// This value MUST come from partition metadata (constant)
+<a href=#1287 id=1287 data-nosnippet>1287</a> </span><span class="macro">assert_eq!</span>(dept_column.value(<span class="number">0</span>), <span class="string">"engineering"</span>);
+<a href=#1288 id=1288 data-nosnippet>1288</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">1</span>), <span class="string">"engineering"</span>);
+<a href=#1289 id=1289 data-nosnippet>1289</a>
+<a href=#1290 id=1290 data-nosnippet>1290</a> <span class="kw">let </span>name_column = result
+<a href=#1291 id=1291 data-nosnippet>1291</a> .column(<span class="number">2</span>)
+<a href=#1292 id=1292 data-nosnippet>1292</a> .as_any()
+<a href=#1293 id=1293 data-nosnippet>1293</a> .downcast_ref::<StringArray>()
+<a href=#1294 id=1294 data-nosnippet>1294</a> .unwrap();
+<a href=#1295 id=1295 data-nosnippet>1295</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
+<a href=#1296 id=1296 data-nosnippet>1296</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
+<a href=#1297 id=1297 data-nosnippet>1297</a> }
+<a href=#1298 id=1298 data-nosnippet>1298</a>
+<a href=#1299 id=1299 data-nosnippet>1299</a> <span class="doccomment">/// Test bucket partitioning with renamed source column.
+<a href=#1300 id=1300 data-nosnippet>1300</a> ///
+<a href=#1301 id=1301 data-nosnippet>1301</a> /// This verifies correct behavior for TestRuntimeFiltering.testRenamedSourceColumnTable() in Iceberg Java.
+<a href=#1302 id=1302 data-nosnippet>1302</a> /// When a source column is renamed after partitioning is established, field-ID-based mapping
+<a href=#1303 id=1303 data-nosnippet>1303</a> /// must still correctly identify the column in Parquet files.
+<a href=#1304 id=1304 data-nosnippet>1304</a> ///
+<a href=#1305 id=1305 data-nosnippet>1305</a> /// # Scenario
+<a href=#1306 id=1306 data-nosnippet>1306</a> ///
+<a href=#1307 id=1307 data-nosnippet>1307</a> /// 1. Table created with `bucket(4, id)` partitioning
+<a href=#1308 id=1308 data-nosnippet>1308</a> /// 2. Data written to Parquet files (field_id=1, name="id")
+<a href=#1309 id=1309 data-nosnippet>1309</a> /// 3. Column renamed: `ALTER TABLE ... RENAME COLUMN id TO row_id`
+<a href=#1310 id=1310 data-nosnippet>1310</a> /// 4. Iceberg schema now has: field_id=1, name="row_id"
+<a href=#1311 id=1311 data-nosnippet>1311</a> /// 5. Parquet files still have: field_id=1, name="id"
+<a href=#1312 id=1312 data-nosnippet>1312</a> ///
+<a href=#1313 id=1313 data-nosnippet>1313</a> /// # Expected Behavior Per Iceberg Spec
+<a href=#1314 id=1314 data-nosnippet>1314</a> ///
+<a href=#1315 id=1315 data-nosnippet>1315</a> /// Per the Iceberg spec "Column Projection" section and Java's PartitionUtil.constantsMap():
+<a href=#1316 id=1316 data-nosnippet>1316</a> /// - Bucket transforms are NON-identity, so partition metadata stores bucket numbers (0-3), not source values
+<a href=#1317 id=1317 data-nosnippet>1317</a> /// - Source columns for non-identity transforms MUST be read from data files
+<a href=#1318 id=1318 data-nosnippet>1318</a> /// - Field-ID-based mapping should find the column by field_id=1 (ignoring name mismatch)
+<a href=#1319 id=1319 data-nosnippet>1319</a> /// - Runtime filtering on `row_id` should work correctly
+<a href=#1320 id=1320 data-nosnippet>1320</a> ///
+<a href=#1321 id=1321 data-nosnippet>1321</a> /// # What This Tests
+<a href=#1322 id=1322 data-nosnippet>1322</a> ///
+<a href=#1323 id=1323 data-nosnippet>1323</a> /// This test ensures that when FileScanTask provides partition_spec and partition_data:
+<a href=#1324 id=1324 data-nosnippet>1324</a> /// - constants_map() correctly identifies that bucket(4, row_id) is NOT an identity transform
+<a href=#1325 id=1325 data-nosnippet>1325</a> /// - The source column (field_id=1) is NOT added to constants_map
+<a href=#1326 id=1326 data-nosnippet>1326</a> /// - Field-ID-based mapping reads actual values from the Parquet file
+<a href=#1327 id=1327 data-nosnippet>1327</a> /// - Values [100, 200, 300] are read, not replaced with bucket constant 2
+<a href=#1328 id=1328 data-nosnippet>1328</a> ///
+<a href=#1329 id=1329 data-nosnippet>1329</a> /// # References
+<a href=#1330 id=1330 data-nosnippet>1330</a> /// - Java test: spark/src/test/java/.../TestRuntimeFiltering.java::testRenamedSourceColumnTable
+<a href=#1331 id=1331 data-nosnippet>1331</a> /// - Java impl: core/src/main/java/org/apache/iceberg/util/PartitionUtil.java::constantsMap()
+<a href=#1332 id=1332 data-nosnippet>1332</a> /// - Iceberg spec: format/spec.md "Column Projection" section
+<a href=#1333 id=1333 data-nosnippet>1333</a> </span><span class="attr">#[test]
+<a href=#1334 id=1334 data-nosnippet>1334</a> </span><span class="kw">fn </span>test_bucket_partitioning_with_renamed_source_column() {
+<a href=#1335 id=1335 data-nosnippet>1335</a> <span class="kw">use </span><span class="kw">crate</span>::spec::{Struct, Transform};
+<a href=#1336 id=1336 data-nosnippet>1336</a>
+<a href=#1337 id=1337 data-nosnippet>1337</a> <span class="comment">// Iceberg schema after rename: row_id (was id), name
+<a href=#1338 id=1338 data-nosnippet>1338</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
+<a href=#1339 id=1339 data-nosnippet>1339</a> Schema::builder()
+<a href=#1340 id=1340 data-nosnippet>1340</a> .with_schema_id(<span class="number">0</span>)
+<a href=#1341 id=1341 data-nosnippet>1341</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#1342 id=1342 data-nosnippet>1342</a> NestedField::required(<span class="number">1</span>, <span class="string">"row_id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
+<a href=#1343 id=1343 data-nosnippet>1343</a> NestedField::optional(<span class="number">2</span>, <span class="string">"name"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#1344 id=1344 data-nosnippet>1344</a> ])
+<a href=#1345 id=1345 data-nosnippet>1345</a> .build()
+<a href=#1346 id=1346 data-nosnippet>1346</a> .unwrap(),
+<a href=#1347 id=1347 data-nosnippet>1347</a> );
+<a href=#1348 id=1348 data-nosnippet>1348</a>
+<a href=#1349 id=1349 data-nosnippet>1349</a> <span class="comment">// Partition spec: bucket(4, row_id) - but source_id still points to field_id=1
+<a href=#1350 id=1350 data-nosnippet>1350</a> </span><span class="kw">let </span>partition_spec = Arc::new(
+<a href=#1351 id=1351 data-nosnippet>1351</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
+<a href=#1352 id=1352 data-nosnippet>1352</a> .with_spec_id(<span class="number">0</span>)
+<a href=#1353 id=1353 data-nosnippet>1353</a> .add_partition_field(<span class="string">"row_id"</span>, <span class="string">"row_id_bucket"</span>, Transform::Bucket(<span class="number">4</span>))
+<a href=#1354 id=1354 data-nosnippet>1354</a> .unwrap()
<a href=#1355 id=1355 data-nosnippet>1355</a> .build()
<a href=#1356 id=1356 data-nosnippet>1356</a> .unwrap(),
<a href=#1357 id=1357 data-nosnippet>1357</a> );
<a href=#1358 id=1358 data-nosnippet>1358</a>
-<a href=#1359 id=1359 data-nosnippet>1359</a> <span class="comment">// Partition spec: identity transform on dept
-<a href=#1360 id=1360 data-nosnippet>1360</a> </span><span class="kw">let </span>partition_spec = Arc::new(
-<a href=#1361 id=1361 data-nosnippet>1361</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
-<a href=#1362 id=1362 data-nosnippet>1362</a> .with_spec_id(<span class="number">0</span>)
-<a href=#1363 id=1363 data-nosnippet>1363</a> .add_partition_field(<span class="string">"dept"</span>, <span class="string">"dept"</span>, Transform::Identity)
-<a href=#1364 id=1364 data-nosnippet>1364</a> .unwrap()
-<a href=#1365 id=1365 data-nosnippet>1365</a> .build()
-<a href=#1366 id=1366 data-nosnippet>1366</a> .unwrap(),
-<a href=#1367 id=1367 data-nosnippet>1367</a> );
+<a href=#1359 id=1359 data-nosnippet>1359</a> <span class="comment">// Partition data: bucket value is 2
+<a href=#1360 id=1360 data-nosnippet>1360</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::int(<span class="number">2</span>))]);
+<a href=#1361 id=1361 data-nosnippet>1361</a>
+<a href=#1362 id=1362 data-nosnippet>1362</a> <span class="comment">// Parquet file has OLD column name "id" but SAME field_id=1
+<a href=#1363 id=1363 data-nosnippet>1363</a> // Field-ID-based mapping should find this despite name mismatch
+<a href=#1364 id=1364 data-nosnippet>1364</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#1365 id=1365 data-nosnippet>1365</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
+<a href=#1366 id=1366 data-nosnippet>1366</a> simple_field(<span class="string">"name"</span>, DataType::Utf8, <span class="bool-val">true</span>, <span class="string">"2"</span>),
+<a href=#1367 id=1367 data-nosnippet>1367</a> ]));
<a href=#1368 id=1368 data-nosnippet>1368</a>
-<a href=#1369 id=1369 data-nosnippet>1369</a> <span class="comment">// Partition data: dept="engineering"
-<a href=#1370 id=1370 data-nosnippet>1370</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::string(<span class="string">"engineering"</span>))]);
-<a href=#1371 id=1371 data-nosnippet>1371</a>
-<a href=#1372 id=1372 data-nosnippet>1372</a> <span class="comment">// Parquet schema: simulates post-ArrowReader state where name mapping already applied
-<a href=#1373 id=1373 data-nosnippet>1373</a> // Has id (field_id=1) and data (field_id=3, assigned by ArrowReader via name mapping)
-<a href=#1374 id=1374 data-nosnippet>1374</a> // Missing: dept (in partition), category (has default), notes (no default)
-<a href=#1375 id=1375 data-nosnippet>1375</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
-<a href=#1376 id=1376 data-nosnippet>1376</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
-<a href=#1377 id=1377 data-nosnippet>1377</a> simple_field(<span class="string">"data"</span>, DataType::Utf8, <span class="bool-val">false</span>, <span class="string">"3"</span>),
-<a href=#1378 id=1378 data-nosnippet>1378</a> ]));
-<a href=#1379 id=1379 data-nosnippet>1379</a>
-<a href=#1380 id=1380 data-nosnippet>1380</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>, <span class="number">4</span>, <span class="number">5</span>]; <span class="comment">// id, dept, data, category, notes
-<a href=#1381 id=1381 data-nosnippet>1381</a>
-<a href=#1382 id=1382 data-nosnippet>1382</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
-<a href=#1383 id=1383 data-nosnippet>1383</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
-<a href=#1384 id=1384 data-nosnippet>1384</a> .with_partition(partition_spec, partition_data)
-<a href=#1385 id=1385 data-nosnippet>1385</a> .build();
-<a href=#1386 id=1386 data-nosnippet>1386</a>
-<a href=#1387 id=1387 data-nosnippet>1387</a> <span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
-<a href=#1388 id=1388 data-nosnippet>1388</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>])),
-<a href=#1389 id=1389 data-nosnippet>1389</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"value1"</span>, <span class="string">"value2"</span>])),
-<a href=#1390 id=1390 data-nosnippet>1390</a> ])
-<a href=#1391 id=1391 data-nosnippet>1391</a> .unwrap();
-<a href=#1392 id=1392 data-nosnippet>1392</a>
-<a href=#1393 id=1393 data-nosnippet>1393</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
-<a href=#1394 id=1394 data-nosnippet>1394</a>
-<a href=#1395 id=1395 data-nosnippet>1395</a> <span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">5</span>);
-<a href=#1396 id=1396 data-nosnippet>1396</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">2</span>);
-<a href=#1397 id=1397 data-nosnippet>1397</a>
-<a href=#1398 id=1398 data-nosnippet>1398</a> <span class="comment">// Verify each column demonstrates the correct spec rule:
-<a href=#1399 id=1399 data-nosnippet>1399</a>
-<a href=#1400 id=1400 data-nosnippet>1400</a> // Normal case: id from Parquet by field ID
-<a href=#1401 id=1401 data-nosnippet>1401</a> </span><span class="kw">let </span>id_column = result
-<a href=#1402 id=1402 data-nosnippet>1402</a> .column(<span class="number">0</span>)
+<a href=#1369 id=1369 data-nosnippet>1369</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>]; <span class="comment">// row_id (field_id=1), name (field_id=2)
+<a href=#1370 id=1370 data-nosnippet>1370</a>
+<a href=#1371 id=1371 data-nosnippet>1371</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
+<a href=#1372 id=1372 data-nosnippet>1372</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
+<a href=#1373 id=1373 data-nosnippet>1373</a> .with_partition(partition_spec, partition_data)
+<a href=#1374 id=1374 data-nosnippet>1374</a> .build();
+<a href=#1375 id=1375 data-nosnippet>1375</a>
+<a href=#1376 id=1376 data-nosnippet>1376</a> <span class="comment">// Create a Parquet RecordBatch with actual data
+<a href=#1377 id=1377 data-nosnippet>1377</a> // Despite column rename, data should be read via field_id=1
+<a href=#1378 id=1378 data-nosnippet>1378</a> </span><span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
+<a href=#1379 id=1379 data-nosnippet>1379</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>, <span class="number">300</span>])),
+<a href=#1380 id=1380 data-nosnippet>1380</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"Alice"</span>, <span class="string">"Bob"</span>, <span class="string">"Charlie"</span>])),
+<a href=#1381 id=1381 data-nosnippet>1381</a> ])
+<a href=#1382 id=1382 data-nosnippet>1382</a> .unwrap();
+<a href=#1383 id=1383 data-nosnippet>1383</a>
+<a href=#1384 id=1384 data-nosnippet>1384</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
+<a href=#1385 id=1385 data-nosnippet>1385</a>
+<a href=#1386 id=1386 data-nosnippet>1386</a> <span class="comment">// Verify the transformed RecordBatch correctly reads data despite name mismatch
+<a href=#1387 id=1387 data-nosnippet>1387</a> </span><span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">2</span>);
+<a href=#1388 id=1388 data-nosnippet>1388</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">3</span>);
+<a href=#1389 id=1389 data-nosnippet>1389</a>
+<a href=#1390 id=1390 data-nosnippet>1390</a> <span class="kw">let </span>row_id_column = result
+<a href=#1391 id=1391 data-nosnippet>1391</a> .column(<span class="number">0</span>)
+<a href=#1392 id=1392 data-nosnippet>1392</a> .as_any()
+<a href=#1393 id=1393 data-nosnippet>1393</a> .downcast_ref::<Int32Array>()
+<a href=#1394 id=1394 data-nosnippet>1394</a> .unwrap();
+<a href=#1395 id=1395 data-nosnippet>1395</a> <span class="comment">// These values MUST come from the Parquet file via field_id=1,
+<a href=#1396 id=1396 data-nosnippet>1396</a> // not be replaced by the bucket constant (2)
+<a href=#1397 id=1397 data-nosnippet>1397</a> </span><span class="macro">assert_eq!</span>(row_id_column.value(<span class="number">0</span>), <span class="number">100</span>);
+<a href=#1398 id=1398 data-nosnippet>1398</a> <span class="macro">assert_eq!</span>(row_id_column.value(<span class="number">1</span>), <span class="number">200</span>);
+<a href=#1399 id=1399 data-nosnippet>1399</a> <span class="macro">assert_eq!</span>(row_id_column.value(<span class="number">2</span>), <span class="number">300</span>);
+<a href=#1400 id=1400 data-nosnippet>1400</a>
+<a href=#1401 id=1401 data-nosnippet>1401</a> <span class="kw">let </span>name_column = result
+<a href=#1402 id=1402 data-nosnippet>1402</a> .column(<span class="number">1</span>)
<a href=#1403 id=1403 data-nosnippet>1403</a> .as_any()
-<a href=#1404 id=1404 data-nosnippet>1404</a> .downcast_ref::<Int32Array>()
+<a href=#1404 id=1404 data-nosnippet>1404</a> .downcast_ref::<StringArray>()
<a href=#1405 id=1405 data-nosnippet>1405</a> .unwrap();
-<a href=#1406 id=1406 data-nosnippet>1406</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">100</span>);
-<a href=#1407 id=1407 data-nosnippet>1407</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">1</span>), <span class="number">200</span>);
-<a href=#1408 id=1408 data-nosnippet>1408</a>
-<a href=#1409 id=1409 data-nosnippet>1409</a> <span class="comment">// Rule #1: dept from partition metadata (identity transform)
-<a href=#1410 id=1410 data-nosnippet>1410</a> </span><span class="kw">let </span>dept_column = result
-<a href=#1411 id=1411 data-nosnippet>1411</a> .column(<span class="number">1</span>)
-<a href=#1412 id=1412 data-nosnippet>1412</a> .as_any()
-<a href=#1413 id=1413 data-nosnippet>1413</a> .downcast_ref::<StringArray>()
-<a href=#1414 id=1414 data-nosnippet>1414</a> .unwrap();
-<a href=#1415 id=1415 data-nosnippet>1415</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">0</span>), <span class="string">"engineering"</span>);
-<a href=#1416 id=1416 data-nosnippet>1416</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">1</span>), <span class="string">"engineering"</span>);
-<a href=#1417 id=1417 data-nosnippet>1417</a>
-<a href=#1418 id=1418 data-nosnippet>1418</a> <span class="comment">// Rule #2: data from Parquet via name mapping
-<a href=#1419 id=1419 data-nosnippet>1419</a> </span><span class="kw">let </span>data_column = result
-<a href=#1420 id=1420 data-nosnippet>1420</a> .column(<span class="number">2</span>)
-<a href=#1421 id=1421 data-nosnippet>1421</a> .as_any()
-<a href=#1422 id=1422 data-nosnippet>1422</a> .downcast_ref::<StringArray>()
-<a href=#1423 id=1423 data-nosnippet>1423</a> .unwrap();
-<a href=#1424 id=1424 data-nosnippet>1424</a> <span class="macro">assert_eq!</span>(data_column.value(<span class="number">0</span>), <span class="string">"value1"</span>);
-<a href=#1425 id=1425 data-nosnippet>1425</a> <span class="macro">assert_eq!</span>(data_column.value(<span class="number">1</span>), <span class="string">"value2"</span>);
-<a href=#1426 id=1426 data-nosnippet>1426</a>
-<a href=#1427 id=1427 data-nosnippet>1427</a> <span class="comment">// Rule #3: category from initial_default
-<a href=#1428 id=1428 data-nosnippet>1428</a> </span><span class="kw">let </span>category_column = result
-<a href=#1429 id=1429 data-nosnippet>1429</a> .column(<span class="number">3</span>)
-<a href=#1430 id=1430 data-nosnippet>1430</a> .as_any()
-<a href=#1431 id=1431 data-nosnippet>1431</a> .downcast_ref::<StringArray>()
-<a href=#1432 id=1432 data-nosnippet>1432</a> .unwrap();
-<a href=#1433 id=1433 data-nosnippet>1433</a> <span class="macro">assert_eq!</span>(category_column.value(<span class="number">0</span>), <span class="string">"default_category"</span>);
-<a href=#1434 id=1434 data-nosnippet>1434</a> <span class="macro">assert_eq!</span>(category_column.value(<span class="number">1</span>), <span class="string">"default_category"</span>);
-<a href=#1435 id=1435 data-nosnippet>1435</a>
-<a href=#1436 id=1436 data-nosnippet>1436</a> <span class="comment">// Rule #4: notes is null (no default, not in Parquet, not in partition)
-<a href=#1437 id=1437 data-nosnippet>1437</a> </span><span class="kw">let </span>notes_column = result
-<a href=#1438 id=1438 data-nosnippet>1438</a> .column(<span class="number">4</span>)
-<a href=#1439 id=1439 data-nosnippet>1439</a> .as_any()
-<a href=#1440 id=1440 data-nosnippet>1440</a> .downcast_ref::<StringArray>()
-<a href=#1441 id=1441 data-nosnippet>1441</a> .unwrap();
-<a href=#1442 id=1442 data-nosnippet>1442</a> <span class="macro">assert!</span>(notes_column.is_null(<span class="number">0</span>));
-<a href=#1443 id=1443 data-nosnippet>1443</a> <span class="macro">assert!</span>(notes_column.is_null(<span class="number">1</span>));
-<a href=#1444 id=1444 data-nosnippet>1444</a> }
-<a href=#1445 id=1445 data-nosnippet>1445</a>}</code></pre></div></section></main></body></html>
\ No newline at end of file
+<a href=#1406 id=1406 data-nosnippet>1406</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">0</span>), <span class="string">"Alice"</span>);
+<a href=#1407 id=1407 data-nosnippet>1407</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">1</span>), <span class="string">"Bob"</span>);
+<a href=#1408 id=1408 data-nosnippet>1408</a> <span class="macro">assert_eq!</span>(name_column.value(<span class="number">2</span>), <span class="string">"Charlie"</span>);
+<a href=#1409 id=1409 data-nosnippet>1409</a> }
+<a href=#1410 id=1410 data-nosnippet>1410</a>
+<a href=#1411 id=1411 data-nosnippet>1411</a> <span class="doccomment">/// Comprehensive integration test that verifies all 4 Iceberg spec rules work correctly.
+<a href=#1412 id=1412 data-nosnippet>1412</a> ///
+<a href=#1413 id=1413 data-nosnippet>1413</a> /// Per the Iceberg spec (https://iceberg.apache.org/spec/#column-projection),
+<a href=#1414 id=1414 data-nosnippet>1414</a> /// "Values for field ids which are not present in a data file must be resolved
+<a href=#1415 id=1415 data-nosnippet>1415</a> /// according the following rules:"
+<a href=#1416 id=1416 data-nosnippet>1416</a> ///
+<a href=#1417 id=1417 data-nosnippet>1417</a> /// This test creates a scenario where each rule is exercised:
+<a href=#1418 id=1418 data-nosnippet>1418</a> /// - Rule #1: dept (identity-partitioned) -> constant from partition metadata
+<a href=#1419 id=1419 data-nosnippet>1419</a> /// - Rule #2: data (via name mapping) -> read from Parquet file by name
+<a href=#1420 id=1420 data-nosnippet>1420</a> /// - Rule #3: category (initial_default) -> use default value
+<a href=#1421 id=1421 data-nosnippet>1421</a> /// - Rule #4: notes (no default) -> return null
+<a href=#1422 id=1422 data-nosnippet>1422</a> ///
+<a href=#1423 id=1423 data-nosnippet>1423</a> /// # References
+<a href=#1424 id=1424 data-nosnippet>1424</a> /// - Iceberg spec: format/spec.md "Column Projection" section
+<a href=#1425 id=1425 data-nosnippet>1425</a> </span><span class="attr">#[test]
+<a href=#1426 id=1426 data-nosnippet>1426</a> </span><span class="kw">fn </span>test_all_four_spec_rules() {
+<a href=#1427 id=1427 data-nosnippet>1427</a> <span class="kw">use </span><span class="kw">crate</span>::spec::Transform;
+<a href=#1428 id=1428 data-nosnippet>1428</a>
+<a href=#1429 id=1429 data-nosnippet>1429</a> <span class="comment">// Iceberg schema with columns designed to exercise each spec rule
+<a href=#1430 id=1430 data-nosnippet>1430</a> </span><span class="kw">let </span>snapshot_schema = Arc::new(
+<a href=#1431 id=1431 data-nosnippet>1431</a> Schema::builder()
+<a href=#1432 id=1432 data-nosnippet>1432</a> .with_schema_id(<span class="number">0</span>)
+<a href=#1433 id=1433 data-nosnippet>1433</a> .with_fields(<span class="macro">vec!</span>[
+<a href=#1434 id=1434 data-nosnippet>1434</a> <span class="comment">// Field in Parquet by field ID (normal case)
+<a href=#1435 id=1435 data-nosnippet>1435</a> </span>NestedField::required(<span class="number">1</span>, <span class="string">"id"</span>, Type::Primitive(PrimitiveType::Int)).into(),
+<a href=#1436 id=1436 data-nosnippet>1436</a> <span class="comment">// Rule #1: Identity-partitioned field - should use partition metadata
+<a href=#1437 id=1437 data-nosnippet>1437</a> </span>NestedField::required(<span class="number">2</span>, <span class="string">"dept"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#1438 id=1438 data-nosnippet>1438</a> <span class="comment">// Rule #2: Field resolved by name mapping (ArrowReader already applied)
+<a href=#1439 id=1439 data-nosnippet>1439</a> </span>NestedField::required(<span class="number">3</span>, <span class="string">"data"</span>, Type::Primitive(PrimitiveType::String)).into(),
+<a href=#1440 id=1440 data-nosnippet>1440</a> <span class="comment">// Rule #3: Field with initial_default
+<a href=#1441 id=1441 data-nosnippet>1441</a> </span>NestedField::optional(<span class="number">4</span>, <span class="string">"category"</span>, Type::Primitive(PrimitiveType::String))
+<a href=#1442 id=1442 data-nosnippet>1442</a> .with_initial_default(Literal::string(<span class="string">"default_category"</span>))
+<a href=#1443 id=1443 data-nosnippet>1443</a> .into(),
+<a href=#1444 id=1444 data-nosnippet>1444</a> <span class="comment">// Rule #4: Field with no default - should be null
+<a href=#1445 id=1445 data-nosnippet>1445</a> </span>NestedField::optional(<span class="number">5</span>, <span class="string">"notes"</span>, Type::Primitive(PrimitiveType::String))
+<a href=#1446 id=1446 data-nosnippet>1446</a> .into(),
+<a href=#1447 id=1447 data-nosnippet>1447</a> ])
+<a href=#1448 id=1448 data-nosnippet>1448</a> .build()
+<a href=#1449 id=1449 data-nosnippet>1449</a> .unwrap(),
+<a href=#1450 id=1450 data-nosnippet>1450</a> );
+<a href=#1451 id=1451 data-nosnippet>1451</a>
+<a href=#1452 id=1452 data-nosnippet>1452</a> <span class="comment">// Partition spec: identity transform on dept
+<a href=#1453 id=1453 data-nosnippet>1453</a> </span><span class="kw">let </span>partition_spec = Arc::new(
+<a href=#1454 id=1454 data-nosnippet>1454</a> <span class="kw">crate</span>::spec::PartitionSpec::builder(snapshot_schema.clone())
+<a href=#1455 id=1455 data-nosnippet>1455</a> .with_spec_id(<span class="number">0</span>)
+<a href=#1456 id=1456 data-nosnippet>1456</a> .add_partition_field(<span class="string">"dept"</span>, <span class="string">"dept"</span>, Transform::Identity)
+<a href=#1457 id=1457 data-nosnippet>1457</a> .unwrap()
+<a href=#1458 id=1458 data-nosnippet>1458</a> .build()
+<a href=#1459 id=1459 data-nosnippet>1459</a> .unwrap(),
+<a href=#1460 id=1460 data-nosnippet>1460</a> );
+<a href=#1461 id=1461 data-nosnippet>1461</a>
+<a href=#1462 id=1462 data-nosnippet>1462</a> <span class="comment">// Partition data: dept="engineering"
+<a href=#1463 id=1463 data-nosnippet>1463</a> </span><span class="kw">let </span>partition_data = Struct::from_iter(<span class="macro">vec!</span>[<span class="prelude-val">Some</span>(Literal::string(<span class="string">"engineering"</span>))]);
+<a href=#1464 id=1464 data-nosnippet>1464</a>
+<a href=#1465 id=1465 data-nosnippet>1465</a> <span class="comment">// Parquet schema: simulates post-ArrowReader state where name mapping already applied
+<a href=#1466 id=1466 data-nosnippet>1466</a> // Has id (field_id=1) and data (field_id=3, assigned by ArrowReader via name mapping)
+<a href=#1467 id=1467 data-nosnippet>1467</a> // Missing: dept (in partition), category (has default), notes (no default)
+<a href=#1468 id=1468 data-nosnippet>1468</a> </span><span class="kw">let </span>parquet_schema = Arc::new(ArrowSchema::new(<span class="macro">vec!</span>[
+<a href=#1469 id=1469 data-nosnippet>1469</a> simple_field(<span class="string">"id"</span>, DataType::Int32, <span class="bool-val">false</span>, <span class="string">"1"</span>),
+<a href=#1470 id=1470 data-nosnippet>1470</a> simple_field(<span class="string">"data"</span>, DataType::Utf8, <span class="bool-val">false</span>, <span class="string">"3"</span>),
+<a href=#1471 id=1471 data-nosnippet>1471</a> ]));
+<a href=#1472 id=1472 data-nosnippet>1472</a>
+<a href=#1473 id=1473 data-nosnippet>1473</a> <span class="kw">let </span>projected_field_ids = [<span class="number">1</span>, <span class="number">2</span>, <span class="number">3</span>, <span class="number">4</span>, <span class="number">5</span>]; <span class="comment">// id, dept, data, category, notes
+<a href=#1474 id=1474 data-nosnippet>1474</a>
+<a href=#1475 id=1475 data-nosnippet>1475</a> </span><span class="kw">let </span><span class="kw-2">mut </span>transformer =
+<a href=#1476 id=1476 data-nosnippet>1476</a> RecordBatchTransformerBuilder::new(snapshot_schema, <span class="kw-2">&</span>projected_field_ids)
+<a href=#1477 id=1477 data-nosnippet>1477</a> .with_partition(partition_spec, partition_data)
+<a href=#1478 id=1478 data-nosnippet>1478</a> .build();
+<a href=#1479 id=1479 data-nosnippet>1479</a>
+<a href=#1480 id=1480 data-nosnippet>1480</a> <span class="kw">let </span>parquet_batch = RecordBatch::try_new(parquet_schema, <span class="macro">vec!</span>[
+<a href=#1481 id=1481 data-nosnippet>1481</a> Arc::new(Int32Array::from(<span class="macro">vec!</span>[<span class="number">100</span>, <span class="number">200</span>])),
+<a href=#1482 id=1482 data-nosnippet>1482</a> Arc::new(StringArray::from(<span class="macro">vec!</span>[<span class="string">"value1"</span>, <span class="string">"value2"</span>])),
+<a href=#1483 id=1483 data-nosnippet>1483</a> ])
+<a href=#1484 id=1484 data-nosnippet>1484</a> .unwrap();
+<a href=#1485 id=1485 data-nosnippet>1485</a>
+<a href=#1486 id=1486 data-nosnippet>1486</a> <span class="kw">let </span>result = transformer.process_record_batch(parquet_batch).unwrap();
+<a href=#1487 id=1487 data-nosnippet>1487</a>
+<a href=#1488 id=1488 data-nosnippet>1488</a> <span class="macro">assert_eq!</span>(result.num_columns(), <span class="number">5</span>);
+<a href=#1489 id=1489 data-nosnippet>1489</a> <span class="macro">assert_eq!</span>(result.num_rows(), <span class="number">2</span>);
+<a href=#1490 id=1490 data-nosnippet>1490</a>
+<a href=#1491 id=1491 data-nosnippet>1491</a> <span class="comment">// Verify each column demonstrates the correct spec rule:
+<a href=#1492 id=1492 data-nosnippet>1492</a>
+<a href=#1493 id=1493 data-nosnippet>1493</a> // Normal case: id from Parquet by field ID
+<a href=#1494 id=1494 data-nosnippet>1494</a> </span><span class="kw">let </span>id_column = result
+<a href=#1495 id=1495 data-nosnippet>1495</a> .column(<span class="number">0</span>)
+<a href=#1496 id=1496 data-nosnippet>1496</a> .as_any()
+<a href=#1497 id=1497 data-nosnippet>1497</a> .downcast_ref::<Int32Array>()
+<a href=#1498 id=1498 data-nosnippet>1498</a> .unwrap();
+<a href=#1499 id=1499 data-nosnippet>1499</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">0</span>), <span class="number">100</span>);
+<a href=#1500 id=1500 data-nosnippet>1500</a> <span class="macro">assert_eq!</span>(id_column.value(<span class="number">1</span>), <span class="number">200</span>);
+<a href=#1501 id=1501 data-nosnippet>1501</a>
+<a href=#1502 id=1502 data-nosnippet>1502</a> <span class="comment">// Rule #1: dept from partition metadata (identity transform)
+<a href=#1503 id=1503 data-nosnippet>1503</a> </span><span class="kw">let </span>dept_column = result
+<a href=#1504 id=1504 data-nosnippet>1504</a> .column(<span class="number">1</span>)
+<a href=#1505 id=1505 data-nosnippet>1505</a> .as_any()
+<a href=#1506 id=1506 data-nosnippet>1506</a> .downcast_ref::<StringArray>()
+<a href=#1507 id=1507 data-nosnippet>1507</a> .unwrap();
+<a href=#1508 id=1508 data-nosnippet>1508</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">0</span>), <span class="string">"engineering"</span>);
+<a href=#1509 id=1509 data-nosnippet>1509</a> <span class="macro">assert_eq!</span>(dept_column.value(<span class="number">1</span>), <span class="string">"engineering"</span>);
+<a href=#1510 id=1510 data-nosnippet>1510</a>
+<a href=#1511 id=1511 data-nosnippet>1511</a> <span class="comment">// Rule #2: data from Parquet via name mapping
+<a href=#1512 id=1512 data-nosnippet>1512</a> </span><span class="kw">let </span>data_column = result
+<a href=#1513 id=1513 data-nosnippet>1513</a> .column(<span class="number">2</span>)
+<a href=#1514 id=1514 data-nosnippet>1514</a> .as_any()
+<a href=#1515 id=1515 data-nosnippet>1515</a> .downcast_ref::<StringArray>()
+<a href=#1516 id=1516 data-nosnippet>1516</a> .unwrap();
+<a href=#1517 id=1517 data-nosnippet>1517</a> <span class="macro">assert_eq!</span>(data_column.value(<span class="number">0</span>), <span class="string">"value1"</span>);
+<a href=#1518 id=1518 data-nosnippet>1518</a> <span class="macro">assert_eq!</span>(data_column.value(<span class="number">1</span>), <span class="string">"value2"</span>);
+<a href=#1519 id=1519 data-nosnippet>1519</a>
+<a href=#1520 id=1520 data-nosnippet>1520</a> <span class="comment">// Rule #3: category from initial_default
+<a href=#1521 id=1521 data-nosnippet>1521</a> </span><span class="kw">let </span>category_column = result
+<a href=#1522 id=1522 data-nosnippet>1522</a> .column(<span class="number">3</span>)
+<a href=#1523 id=1523 data-nosnippet>1523</a> .as_any()
+<a href=#1524 id=1524 data-nosnippet>1524</a> .downcast_ref::<StringArray>()
+<a href=#1525 id=1525 data-nosnippet>1525</a> .unwrap();
+<a href=#1526 id=1526 data-nosnippet>1526</a> <span class="macro">assert_eq!</span>(category_column.value(<span class="number">0</span>), <span class="string">"default_category"</span>);
+<a href=#1527 id=1527 data-nosnippet>1527</a> <span class="macro">assert_eq!</span>(category_column.value(<span class="number">1</span>), <span class="string">"default_category"</span>);
+<a href=#1528 id=1528 data-nosnippet>1528</a>
+<a href=#1529 id=1529 data-nosnippet>1529</a> <span class="comment">// Rule #4: notes is null (no default, not in Parquet, not in partition)
+<a href=#1530 id=1530 data-nosnippet>1530</a> </span><span class="kw">let </span>notes_column = result
+<a href=#1531 id=1531 data-nosnippet>1531</a> .column(<span class="number">4</span>)
+<a href=#1532 id=1532 data-nosnippet>1532</a> .as_any()
+<a href=#1533 id=1533 data-nosnippet>1533</a> .downcast_ref::<StringArray>()
+<a href=#1534 id=1534 data-nosnippet>1534</a> .unwrap();
+<a href=#1535 id=1535 data-nosnippet>1535</a> <span class="macro">assert!</span>(notes_column.is_null(<span class="number">0</span>));
+<a href=#1536 id=1536 data-nosnippet>1536</a> <span class="macro">assert!</span>(notes_column.is_null(<span class="number">1</span>));
+<a href=#1537 id=1537 data-nosnippet>1537</a> }
+<a href=#1538 id=1538 data-nosnippet>1538</a>}</code></pre></div></section></main></body></html>
\ No newline at end of file