You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "ch-sc (via GitHub)" <gi...@apache.org> on 2023/10/16 12:11:15 UTC
[PR] Add small column on empty projection [arrow-datafusion]
ch-sc opened a new pull request, #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833
## Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes #123` indicates that this PR will close issue #123.
-->
Improves #3214.
## Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes.
-->
If a projection is empty, we add the first column of the input schema since some parts of DataFusion still rely on at least having one column. Instead of selecting the first column from the input schema, these changes aim to select a column with a smaller memory size. The memory size is based on the data type.
## What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR.
-->
## Are these changes tested?
<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
2. Serve as another way to document the expected behavior of the code
If tests are not included in your PR, please explain why (for example, are they covered by existing tests)?
-->
Basic unit tests for new logic are included. All tests that involve query planning and empty projections execute this code.
## Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be updated before approving the PR.
-->
<!--
If there are any breaking changes to public APIs, please add the `api change` label.
-->
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1361792738
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
+ .sum(),
+ Union(fields, _) => fields
+ .iter()
+ .map(|(_, f)| nested_size(f.data_type(), nesting))
+ .sum(),
+ Map(field, _) => nested_size(field.data_type(), nesting),
+ RunEndEncoded(run_ends, values) => {
+ nested_size(run_ends.data_type(), nesting)
+ + nested_size(values.data_type(), nesting)
+ }
+ }
+}
+
+/// Find a field with a presumable small memory footprint based on its data type's memory size
+/// and the level of nesting.
+fn find_small_field(fields: &[DFField]) -> DFField {
+ let mut indexed_fields = fields
+ .iter()
+ .map(|f| {
+ let nesting = &mut 0;
+ let size = nested_size(f.field().data_type(), nesting);
+ (*nesting, size)
+ })
+ .enumerate()
+ .collect_vec();
+
+ let (i, _) = indexed_fields
+ .select_nth_unstable_by(
Review Comment:
We can use `min_by` instead of sorting?
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
+ .sum(),
+ Union(fields, _) => fields
+ .iter()
+ .map(|(_, f)| nested_size(f.data_type(), nesting))
+ .sum(),
+ Map(field, _) => nested_size(field.data_type(), nesting),
+ RunEndEncoded(run_ends, values) => {
+ nested_size(run_ends.data_type(), nesting)
+ + nested_size(values.data_type(), nesting)
+ }
+ }
+}
+
+/// Find a field with a presumable small memory footprint based on its data type's memory size
+/// and the level of nesting.
+fn find_small_field(fields: &[DFField]) -> DFField {
+ let mut indexed_fields = fields
+ .iter()
+ .map(|f| {
+ let nesting = &mut 0;
+ let size = nested_size(f.field().data_type(), nesting);
+ (*nesting, size)
+ })
+ .enumerate()
+ .collect_vec();
+
+ let (i, _) = indexed_fields
+ .select_nth_unstable_by(
Review Comment:
We can use `min_by` instead of selecting/sorting?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "ch-sc (via GitHub)" <gi...@apache.org>.
ch-sc commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1363540750
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
Review Comment:
Makes sense. Added a comment about variable sized types. Feel free to rephrase if you think something is missing.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1361799741
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
Review Comment:
In principle we could project a sub-field from a struct instead of the entire struct (all columns).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1361794458
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
+ .sum(),
+ Union(fields, _) => fields
+ .iter()
+ .map(|(_, f)| nested_size(f.data_type(), nesting))
+ .sum(),
+ Map(field, _) => nested_size(field.data_type(), nesting),
+ RunEndEncoded(run_ends, values) => {
+ nested_size(run_ends.data_type(), nesting)
+ + nested_size(values.data_type(), nesting)
+ }
+ }
+}
+
+/// Find a field with a presumable small memory footprint based on its data type's memory size
+/// and the level of nesting.
+fn find_small_field(fields: &[DFField]) -> DFField {
+ let mut indexed_fields = fields
+ .iter()
+ .map(|f| {
+ let nesting = &mut 0;
+ let size = nested_size(f.field().data_type(), nesting);
+ (*nesting, size)
+ })
+ .enumerate()
+ .collect_vec();
Review Comment:
We shouldn't need to collect to intermediate vec (i.e. use `min_by` on iterator)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "ch-sc (via GitHub)" <gi...@apache.org>.
ch-sc commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1363538324
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
+ .sum(),
+ Union(fields, _) => fields
+ .iter()
+ .map(|(_, f)| nested_size(f.data_type(), nesting))
+ .sum(),
+ Map(field, _) => nested_size(field.data_type(), nesting),
+ RunEndEncoded(run_ends, values) => {
+ nested_size(run_ends.data_type(), nesting)
+ + nested_size(values.data_type(), nesting)
+ }
+ }
+}
+
+/// Find a field with a presumable small memory footprint based on its data type's memory size
+/// and the level of nesting.
+fn find_small_field(fields: &[DFField]) -> DFField {
+ let mut indexed_fields = fields
+ .iter()
+ .map(|f| {
+ let nesting = &mut 0;
+ let size = nested_size(f.field().data_type(), nesting);
+ (*nesting, size)
+ })
+ .enumerate()
+ .collect_vec();
+
+ let (i, _) = indexed_fields
+ .select_nth_unstable_by(
Review Comment:
Yeah, much cleaner with `min_by`. Thanks! :)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#issuecomment-1768219426
Change seems non controversial and has some good tests, so merging seems fine.
Thank you @ch-sc 😊
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1361795405
##########
datafusion/sqllogictest/test_files/avro.slt:
##########
@@ -253,10 +253,10 @@ EXPLAIN SELECT count(*) from alltypes_plain
----
logical_plan
Aggregate: groupBy=[[]], aggr=[[COUNT(UInt8(1)) AS COUNT(*)]]
---TableScan: alltypes_plain projection=[id]
+--TableScan: alltypes_plain projection=[bool_col]
Review Comment:
👍
##########
datafusion/sqllogictest/test_files/avro.slt:
##########
@@ -253,10 +253,10 @@ EXPLAIN SELECT count(*) from alltypes_plain
----
logical_plan
Aggregate: groupBy=[[]], aggr=[[COUNT(UInt8(1)) AS COUNT(*)]]
---TableScan: alltypes_plain projection=[id]
+--TableScan: alltypes_plain projection=[bool_col]
Review Comment:
👍
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan merged PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "ch-sc (via GitHub)" <gi...@apache.org>.
ch-sc commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1363543733
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
Review Comment:
Good idea, I will play around with it. Though it sounds like a rare edge case to me where no other "smaller" type would be present in the schema!?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "ch-sc (via GitHub)" <gi...@apache.org>.
ch-sc commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1363543733
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
Review Comment:
Good idea, I will play around with this idea. Though it sounds like a rare edge case to me where no other "smaller" type would be present in the schema!?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1361797733
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
Review Comment:
Can we add a comment saying that variable-sized types are estimated using some heuristics?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Add small column on empty projection [arrow-datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #7833:
URL: https://github.com/apache/arrow-datafusion/pull/7833#discussion_r1363678450
##########
datafusion/optimizer/src/push_down_projection.rs:
##########
@@ -423,7 +426,93 @@ pub fn collect_projection_expr(projection: &Projection) -> HashMap<String, Expr>
.collect::<HashMap<_, _>>()
}
-// Get the projection exprs from columns in the order of the schema
+/// Accumulate the memory size of a data type measured in bits.
+///
+/// Nested types are traversed and increment `nesting` on every level.
+fn nested_size(data_type: &DataType, nesting: &mut usize) -> usize {
+ use DataType::*;
+ if data_type.is_primitive() {
+ return data_type.primitive_width().unwrap_or(1) * 8;
+ }
+
+ if data_type.is_nested() {
+ *nesting += 1;
+ }
+
+ match data_type {
+ Null => 0,
+ Boolean => 1,
+ Binary | Utf8 => 128,
+ LargeBinary | LargeUtf8 => 256,
+ FixedSizeBinary(bytes) => (*bytes * 8) as usize,
+ // primitive types
+ Int8
+ | Int16
+ | Int32
+ | Int64
+ | UInt8
+ | UInt16
+ | UInt32
+ | UInt64
+ | Float16
+ | Float32
+ | Float64
+ | Timestamp(_, _)
+ | Date32
+ | Date64
+ | Time32(_)
+ | Time64(_)
+ | Duration(_)
+ | Interval(_)
+ | Dictionary(_, _)
+ | Decimal128(_, _)
+ | Decimal256(_, _) => data_type.primitive_width().unwrap_or(1) * 8,
+ // nested types
+ List(f) => nested_size(f.data_type(), nesting),
+ FixedSizeList(_, s) => (s * 8) as usize,
+ LargeList(f) => nested_size(f.data_type(), nesting),
+ Struct(fields) => fields
+ .iter()
+ .map(|f| nested_size(f.data_type(), nesting))
Review Comment:
Yeah indeed :)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org