You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@arrow.apache.org by tu...@apache.org on 2022/10/06 09:51:33 UTC

[arrow-rs] branch master updated: Convert rows to arrays (#2677) (#2826)

This is an automated email from the ASF dual-hosted git repository.

tustvold pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/arrow-rs.git


The following commit(s) were added to refs/heads/master by this push:
     new f8c403722 Convert rows to arrays (#2677) (#2826)
f8c403722 is described below

commit f8c40372242ab096cd53118021d016cac4f0daed
Author: Raphael Taylor-Davies <17...@users.noreply.github.com>
AuthorDate: Thu Oct 6 10:51:27 2022 +0100

    Convert rows to arrays (#2677) (#2826)
    
    * Convert rows to arrays (#2677)
    
    * Review feedback
    
    * Clippy
---
 arrow/benches/row_format.rs | 140 ++++--------
 arrow/src/row/dictionary.rs | 337 ++++++++++++++++++++++++++++
 arrow/src/row/fixed.rs      | 239 ++++++++++++++++++--
 arrow/src/row/interner.rs   |   4 +-
 arrow/src/row/mod.rs        | 528 ++++++++++++++++++++++++++++++++++++--------
 arrow/src/row/variable.rs   | 140 +++++++++++-
 6 files changed, 1158 insertions(+), 230 deletions(-)

diff --git a/arrow/benches/row_format.rs b/arrow/benches/row_format.rs
index ec872c127..ff505781a 100644
--- a/arrow/benches/row_format.rs
+++ b/arrow/benches/row_format.rs
@@ -20,161 +20,95 @@ extern crate criterion;
 extern crate core;
 
 use arrow::array::ArrayRef;
-use arrow::datatypes::{DataType, Int64Type, UInt64Type};
+use arrow::datatypes::{Int64Type, UInt64Type};
 use arrow::row::{RowConverter, SortField};
 use arrow::util::bench_util::{
     create_primitive_array, create_string_array_with_len, create_string_dict_array,
 };
 use arrow_array::types::Int32Type;
+use arrow_array::Array;
 use criterion::{black_box, Criterion};
 use std::sync::Arc;
 
-fn row_bench(c: &mut Criterion) {
-    let cols = vec![Arc::new(create_primitive_array::<UInt64Type>(4096, 0.)) as ArrayRef];
+fn do_bench(c: &mut Criterion, name: &str, cols: Vec<ArrayRef>) {
+    let fields: Vec<_> = cols
+        .iter()
+        .map(|x| SortField::new(x.data_type().clone()))
+        .collect();
 
-    c.bench_function("row_batch 4096 u64(0)", |b| {
+    c.bench_function(&format!("convert_columns {}", name), |b| {
         b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::UInt64)]);
-            black_box(converter.convert_columns(&cols))
+            let mut converter = RowConverter::new(fields.clone());
+            black_box(converter.convert_columns(&cols).unwrap())
         });
     });
 
-    let cols = vec![Arc::new(create_primitive_array::<Int64Type>(4096, 0.)) as ArrayRef];
+    let mut converter = RowConverter::new(fields);
+    let rows = converter.convert_columns(&cols).unwrap();
 
-    c.bench_function("row_batch 4096 i64(0)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Int64)]);
-            black_box(converter.convert_columns(&cols))
-        });
+    c.bench_function(&format!("convert_rows {}", name), |b| {
+        b.iter(|| black_box(converter.convert_rows(&rows).unwrap()));
     });
+}
+
+fn row_bench(c: &mut Criterion) {
+    let cols = vec![Arc::new(create_primitive_array::<UInt64Type>(4096, 0.)) as ArrayRef];
+    do_bench(c, "4096 u64(0)", cols);
+
+    let cols = vec![Arc::new(create_primitive_array::<Int64Type>(4096, 0.)) as ArrayRef];
+    do_bench(c, "4096 i64(0)", cols);
 
     let cols =
         vec![Arc::new(create_string_array_with_len::<i32>(4096, 0., 10)) as ArrayRef];
-
-    c.bench_function("row_batch 4096 string(10, 0)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
+    do_bench(c, "4096 string(10, 0)", cols);
 
     let cols =
         vec![Arc::new(create_string_array_with_len::<i32>(4096, 0., 30)) as ArrayRef];
-
-    c.bench_function("row_batch 4096 string(30, 0)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
+    do_bench(c, "4096 string(30, 0)", cols);
 
     let cols =
         vec![Arc::new(create_string_array_with_len::<i32>(4096, 0., 100)) as ArrayRef];
-
-    c.bench_function("row_batch 4096 string(100, 0)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
+    do_bench(c, "4096 string(100, 0)", cols);
 
     let cols =
         vec![Arc::new(create_string_array_with_len::<i32>(4096, 0.5, 100)) as ArrayRef];
-
-    c.bench_function("row_batch 4096 string(100, 0.5)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
+    do_bench(c, "4096 string(100, 0.5)", cols);
 
     let cols =
         vec![Arc::new(create_string_dict_array::<Int32Type>(4096, 0., 10)) as ArrayRef];
-
-    c.bench_function("row_batch 4096 string_dictionary(10, 0)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
+    do_bench(c, "4096 string_dictionary(10, 0)", cols);
 
     let cols =
         vec![Arc::new(create_string_dict_array::<Int32Type>(4096, 0., 30)) as ArrayRef];
-
-    c.bench_function("row_batch 4096 string_dictionary(30, 0)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
+    do_bench(c, "4096 string_dictionary(30, 0)", cols);
 
     let cols =
         vec![Arc::new(create_string_dict_array::<Int32Type>(4096, 0., 100)) as ArrayRef];
-
-    c.bench_function("row_batch 4096 string_dictionary(100, 0)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
+    do_bench(c, "4096 string_dictionary(100, 0)", cols);
 
     let cols =
         vec![Arc::new(create_string_dict_array::<Int32Type>(4096, 0.5, 100)) as ArrayRef];
+    do_bench(c, "4096 string_dictionary(100, 0.5)", cols);
 
-    c.bench_function("row_batch 4096 string_dictionary(100, 0.5)", |b| {
-        b.iter(|| {
-            let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-            black_box(converter.convert_columns(&cols))
-        });
-    });
-
-    let cols = [
+    let cols = vec![
         Arc::new(create_string_array_with_len::<i32>(4096, 0.5, 20)) as ArrayRef,
         Arc::new(create_string_array_with_len::<i32>(4096, 0., 30)) as ArrayRef,
         Arc::new(create_string_array_with_len::<i32>(4096, 0., 100)) as ArrayRef,
         Arc::new(create_primitive_array::<Int64Type>(4096, 0.)) as ArrayRef,
     ];
-
-    let fields = [
-        SortField::new(DataType::Utf8),
-        SortField::new(DataType::Utf8),
-        SortField::new(DataType::Utf8),
-        SortField::new(DataType::Int64),
-    ];
-
-    c.bench_function(
-        "row_batch 4096 string(20, 0.5), string(30, 0), string(100, 0), i64(0)",
-        |b| {
-            b.iter(|| {
-                let mut converter = RowConverter::new(fields.to_vec());
-                black_box(converter.convert_columns(&cols))
-            });
-        },
+    do_bench(
+        c,
+        "4096 string(20, 0.5), string(30, 0), string(100, 0), i64(0)",
+        cols,
     );
 
-    let cols = [
+    let cols = vec![
         Arc::new(create_string_dict_array::<Int32Type>(4096, 0.5, 20)) as ArrayRef,
         Arc::new(create_string_dict_array::<Int32Type>(4096, 0., 30)) as ArrayRef,
         Arc::new(create_string_dict_array::<Int32Type>(4096, 0., 100)) as ArrayRef,
         Arc::new(create_primitive_array::<Int64Type>(4096, 0.)) as ArrayRef,
     ];
-
-    let fields = [
-        SortField::new(DataType::Utf8),
-        SortField::new(DataType::Utf8),
-        SortField::new(DataType::Utf8),
-        SortField::new(DataType::Int64),
-    ];
-
-    c.bench_function(
-        "row_batch 4096 string_dictionary(20, 0.5), string_dictionary(30, 0), string_dictionary(100, 0), i64(0)",
-        |b| {
-            b.iter(|| {
-                let mut converter = RowConverter::new(fields.to_vec());
-                black_box(converter.convert_columns(&cols))
-            });
-        },
-    );
+    do_bench(c, "4096 4096 string_dictionary(20, 0.5), string_dictionary(30, 0), string_dictionary(100, 0), i64(0)", cols);
 }
 
 criterion_group!(benches, row_bench);
diff --git a/arrow/src/row/dictionary.rs b/arrow/src/row/dictionary.rs
new file mode 100644
index 000000000..4a048fbce
--- /dev/null
+++ b/arrow/src/row/dictionary.rs
@@ -0,0 +1,337 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements.  See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership.  The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License.  You may obtain a copy of the License at
+//
+//   http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied.  See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+use crate::compute::SortOptions;
+use crate::row::fixed::{FixedLengthEncoding, FromSlice, RawDecimal};
+use crate::row::interner::{Interned, OrderPreservingInterner};
+use crate::row::{null_sentinel, Rows};
+use arrow_array::builder::*;
+use arrow_array::cast::*;
+use arrow_array::types::*;
+use arrow_array::*;
+use arrow_buffer::{ArrowNativeType, MutableBuffer, ToByteSlice};
+use arrow_data::{ArrayData, ArrayDataBuilder};
+use arrow_schema::{ArrowError, DataType, IntervalUnit, TimeUnit};
+use std::collections::hash_map::Entry;
+use std::collections::HashMap;
+
+/// Computes the dictionary mapping for the given dictionary values
+pub fn compute_dictionary_mapping(
+    interner: &mut OrderPreservingInterner,
+    values: &ArrayRef,
+) -> Result<Vec<Option<Interned>>, ArrowError> {
+    Ok(downcast_primitive_array! {
+        values => interner
+            .intern(values.iter().map(|x| x.map(|x| x.encode()))),
+        DataType::Binary => {
+            let iter = as_generic_binary_array::<i64>(values).iter();
+            interner.intern(iter)
+        }
+        DataType::LargeBinary => {
+            let iter = as_generic_binary_array::<i64>(values).iter();
+            interner.intern(iter)
+        }
+        DataType::Utf8 => {
+            let iter = as_string_array(values).iter().map(|x| x.map(|x| x.as_bytes()));
+            interner.intern(iter)
+        }
+        DataType::LargeUtf8 => {
+            let iter = as_largestring_array(values).iter().map(|x| x.map(|x| x.as_bytes()));
+            interner.intern(iter)
+        }
+        t => return Err(ArrowError::NotYetImplemented(format!("dictionary value {} is not supported", t))),
+    })
+}
+
+/// Dictionary types are encoded as
+///
+/// - single `0_u8` if null
+/// - the bytes of the corresponding normalized key including the null terminator
+pub fn encode_dictionary<K: ArrowDictionaryKeyType>(
+    out: &mut Rows,
+    column: &DictionaryArray<K>,
+    normalized_keys: &[Option<&[u8]>],
+    opts: SortOptions,
+) {
+    for (offset, k) in out.offsets.iter_mut().skip(1).zip(column.keys()) {
+        match k.and_then(|k| normalized_keys[k.as_usize()]) {
+            Some(normalized_key) => {
+                let end_offset = *offset + 1 + normalized_key.len();
+                out.buffer[*offset] = 1;
+                out.buffer[*offset + 1..end_offset].copy_from_slice(normalized_key);
+                // Negate if descending
+                if opts.descending {
+                    out.buffer[*offset..end_offset]
+                        .iter_mut()
+                        .for_each(|v| *v = !*v)
+                }
+                *offset = end_offset;
+            }
+            None => {
+                out.buffer[*offset] = null_sentinel(opts);
+                *offset += 1;
+            }
+        }
+    }
+}
+
+/// Decodes a string array from `rows` with the provided `options`
+///
+/// # Safety
+///
+/// `interner` must contain valid data for the provided `value_type`
+pub unsafe fn decode_dictionary<K: ArrowDictionaryKeyType>(
+    interner: &OrderPreservingInterner,
+    value_type: &DataType,
+    options: SortOptions,
+    rows: &mut [&[u8]],
+) -> Result<DictionaryArray<K>, ArrowError> {
+    let len = rows.len();
+    let mut dictionary: HashMap<Interned, K::Native> = HashMap::with_capacity(len);
+
+    let null_sentinel = null_sentinel(options);
+
+    // If descending, the null terminator will have been negated
+    let null_terminator = match options.descending {
+        true => 0xFF,
+        false => 0_u8,
+    };
+
+    let mut null_builder = BooleanBufferBuilder::new(len);
+    let mut keys = BufferBuilder::<K::Native>::new(len);
+    let mut values = Vec::with_capacity(len);
+    let mut null_count = 0;
+    let mut key_scratch = Vec::new();
+
+    for row in rows {
+        if row[0] == null_sentinel {
+            null_builder.append(false);
+            null_count += 1;
+            *row = &row[1..];
+            keys.append(K::Native::default());
+            continue;
+        }
+
+        let key_offset = row
+            .iter()
+            .skip(1)
+            .position(|x| *x == null_terminator)
+            .unwrap();
+
+        // Extract the normalized key including the null terminator
+        let key = &row[1..key_offset + 2];
+        *row = &row[key_offset + 2..];
+
+        let interned = match options.descending {
+            true => {
+                // If options.descending the normalized key will have been
+                // negated we must first reverse this
+                key_scratch.clear();
+                key_scratch.extend_from_slice(key);
+                key_scratch.iter_mut().for_each(|o| *o = !*o);
+                interner.lookup(&key_scratch).unwrap()
+            }
+            false => interner.lookup(key).unwrap(),
+        };
+
+        let k = match dictionary.entry(interned) {
+            Entry::Vacant(v) => {
+                let k = values.len();
+                values.push(interner.value(interned));
+                let key = K::Native::from_usize(k)
+                    .ok_or(ArrowError::DictionaryKeyOverflowError)?;
+                *v.insert(key)
+            }
+            Entry::Occupied(o) => *o.get(),
+        };
+
+        keys.append(k);
+        null_builder.append(true);
+    }
+
+    let child = match &value_type {
+        DataType::Null => NullArray::new(values.len()).into_data(),
+        DataType::Boolean => decode_bool(&values),
+        DataType::Int8 => decode_primitive::<Int8Type>(&values),
+        DataType::Int16 => decode_primitive::<Int16Type>(&values),
+        DataType::Int32 => decode_primitive::<Int32Type>(&values),
+        DataType::Int64 => decode_primitive::<Int64Type>(&values),
+        DataType::UInt8 => decode_primitive::<UInt8Type>(&values),
+        DataType::UInt16 => decode_primitive::<UInt16Type>(&values),
+        DataType::UInt32 => decode_primitive::<UInt32Type>(&values),
+        DataType::UInt64 => decode_primitive::<UInt64Type>(&values),
+        DataType::Float16 => decode_primitive::<Float16Type>(&values),
+        DataType::Float32 => decode_primitive::<Float32Type>(&values),
+        DataType::Float64 => decode_primitive::<Float64Type>(&values),
+        DataType::Timestamp(TimeUnit::Second, _) => {
+            decode_primitive::<TimestampSecondType>(&values)
+        }
+        DataType::Timestamp(TimeUnit::Millisecond, _) => {
+            decode_primitive::<TimestampMillisecondType>(&values)
+        }
+        DataType::Timestamp(TimeUnit::Microsecond, _) => {
+            decode_primitive::<TimestampMicrosecondType>(&values)
+        }
+        DataType::Timestamp(TimeUnit::Nanosecond, _) => {
+            decode_primitive::<TimestampNanosecondType>(&values)
+        }
+        DataType::Date32 => decode_primitive::<Date32Type>(&values),
+        DataType::Date64 => decode_primitive::<Date64Type>(&values),
+        DataType::Time32(t) => match t {
+            TimeUnit::Second => decode_primitive::<Time32SecondType>(&values),
+            TimeUnit::Millisecond => decode_primitive::<Time32MillisecondType>(&values),
+            _ => unreachable!(),
+        },
+        DataType::Time64(t) => match t {
+            TimeUnit::Microsecond => decode_primitive::<Time64MicrosecondType>(&values),
+            TimeUnit::Nanosecond => decode_primitive::<Time64NanosecondType>(&values),
+            _ => unreachable!(),
+        },
+        DataType::Duration(TimeUnit::Second) => {
+            decode_primitive::<DurationSecondType>(&values)
+        }
+        DataType::Duration(TimeUnit::Millisecond) => {
+            decode_primitive::<DurationMillisecondType>(&values)
+        }
+        DataType::Duration(TimeUnit::Microsecond) => {
+            decode_primitive::<DurationMicrosecondType>(&values)
+        }
+        DataType::Duration(TimeUnit::Nanosecond) => {
+            decode_primitive::<DurationNanosecondType>(&values)
+        }
+        DataType::Interval(IntervalUnit::DayTime) => {
+            decode_primitive::<IntervalDayTimeType>(&values)
+        }
+        DataType::Interval(IntervalUnit::MonthDayNano) => {
+            decode_primitive::<IntervalMonthDayNanoType>(&values)
+        }
+        DataType::Interval(IntervalUnit::YearMonth) => {
+            decode_primitive::<IntervalYearMonthType>(&values)
+        }
+        DataType::Decimal128(p, s) => {
+            decode_decimal::<16, Decimal128Type>(&values, *p, *s)
+        }
+        DataType::Decimal256(p, s) => {
+            decode_decimal::<32, Decimal256Type>(&values, *p, *s)
+        }
+        DataType::Utf8 => decode_string::<i32>(&values),
+        DataType::LargeUtf8 => decode_string::<i64>(&values),
+        DataType::Binary => decode_binary::<i32>(&values),
+        DataType::LargeBinary => decode_binary::<i64>(&values),
+        _ => {
+            return Err(ArrowError::NotYetImplemented(format!(
+                "decoding dictionary values of {}",
+                value_type
+            )))
+        }
+    };
+
+    let data_type =
+        DataType::Dictionary(Box::new(K::DATA_TYPE), Box::new(value_type.clone()));
+
+    let builder = ArrayDataBuilder::new(data_type)
+        .len(len)
+        .null_bit_buffer(Some(null_builder.finish()))
+        .null_count(null_count)
+        .add_buffer(keys.finish())
+        .add_child_data(child);
+
+    Ok(DictionaryArray::from(builder.build_unchecked()))
+}
+
+/// Decodes a binary array from dictionary values
+///
+/// # Safety
+///
+/// Values must be valid UTF-8
+fn decode_binary<O: OffsetSizeTrait>(values: &[&[u8]]) -> ArrayData {
+    let capacity = values.iter().map(|x| x.len()).sum();
+    let mut builder = GenericBinaryBuilder::<O>::with_capacity(values.len(), capacity);
+    for v in values {
+        builder.append_value(v)
+    }
+    builder.finish().into_data()
+}
+
+/// Decodes a string array from dictionary values
+///
+/// # Safety
+///
+/// Values must be valid UTF-8
+unsafe fn decode_string<O: OffsetSizeTrait>(values: &[&[u8]]) -> ArrayData {
+    let d = match O::IS_LARGE {
+        true => DataType::LargeUtf8,
+        false => DataType::Utf8,
+    };
+
+    decode_binary::<O>(values)
+        .into_builder()
+        .data_type(d)
+        .build_unchecked()
+}
+
+/// Decodes a boolean array from dictionary values
+fn decode_bool(values: &[&[u8]]) -> ArrayData {
+    let mut builder = BooleanBufferBuilder::new(values.len());
+    for value in values {
+        builder.append(bool::decode([value[0]]))
+    }
+
+    let builder = ArrayDataBuilder::new(DataType::Boolean)
+        .len(values.len())
+        .add_buffer(builder.finish());
+
+    // SAFETY: Buffers correct length
+    unsafe { builder.build_unchecked() }
+}
+
+/// Decodes a fixed length type array from dictionary values
+fn decode_fixed<T: FixedLengthEncoding + ToByteSlice>(
+    values: &[&[u8]],
+    data_type: DataType,
+) -> ArrayData {
+    let mut buffer = MutableBuffer::new(std::mem::size_of::<T>() * values.len());
+
+    for value in values {
+        let value = T::Encoded::from_slice(value, false);
+        buffer.push(T::decode(value))
+    }
+
+    let builder = ArrayDataBuilder::new(data_type)
+        .len(values.len())
+        .add_buffer(buffer.into());
+
+    // SAFETY: Buffers correct length
+    unsafe { builder.build_unchecked() }
+}
+
+/// Decodes a `PrimitiveArray` from dictionary values
+fn decode_primitive<T: ArrowPrimitiveType>(values: &[&[u8]]) -> ArrayData
+where
+    T::Native: FixedLengthEncoding,
+{
+    decode_fixed::<T::Native>(values, T::DATA_TYPE)
+}
+
+/// Decodes a `DecimalArray` from dictionary values
+fn decode_decimal<const N: usize, T: DecimalType>(
+    values: &[&[u8]],
+    precision: u8,
+    scale: u8,
+) -> ArrayData {
+    decode_fixed::<RawDecimal<N>>(values, T::TYPE_CONSTRUCTOR(precision, scale))
+}
diff --git a/arrow/src/row/fixed.rs b/arrow/src/row/fixed.rs
index 04b9a30ec..9952ee094 100644
--- a/arrow/src/row/fixed.rs
+++ b/arrow/src/row/fixed.rs
@@ -18,18 +18,39 @@
 use crate::array::PrimitiveArray;
 use crate::compute::SortOptions;
 use crate::datatypes::ArrowPrimitiveType;
-use crate::row::Rows;
-use crate::util::decimal::{Decimal128, Decimal256};
+use crate::row::{null_sentinel, Rows};
+use arrow_array::types::DecimalType;
+use arrow_array::{BooleanArray, DecimalArray};
+use arrow_buffer::{bit_util, MutableBuffer, ToByteSlice};
+use arrow_data::{ArrayData, ArrayDataBuilder};
+use arrow_schema::DataType;
 use half::f16;
 
+pub trait FromSlice {
+    fn from_slice(slice: &[u8], invert: bool) -> Self;
+}
+
+impl<const N: usize> FromSlice for [u8; N] {
+    #[inline]
+    fn from_slice(slice: &[u8], invert: bool) -> Self {
+        let mut t: Self = slice.try_into().unwrap();
+        if invert {
+            t.iter_mut().for_each(|o| *o = !*o);
+        }
+        t
+    }
+}
+
 /// Encodes a value of a particular fixed width type into bytes according to the rules
 /// described on [`super::RowConverter`]
 pub trait FixedLengthEncoding: Copy {
     const ENCODED_LEN: usize = 1 + std::mem::size_of::<Self::Encoded>();
 
-    type Encoded: Sized + Copy + AsRef<[u8]> + AsMut<[u8]>;
+    type Encoded: Sized + Copy + FromSlice + AsRef<[u8]> + AsMut<[u8]>;
 
     fn encode(self) -> Self::Encoded;
+
+    fn decode(encoded: Self::Encoded) -> Self;
 }
 
 impl FixedLengthEncoding for bool {
@@ -38,6 +59,10 @@ impl FixedLengthEncoding for bool {
     fn encode(self) -> [u8; 1] {
         [self as u8]
     }
+
+    fn decode(encoded: Self::Encoded) -> Self {
+        encoded[0] != 0
+    }
 }
 
 macro_rules! encode_signed {
@@ -51,6 +76,12 @@ macro_rules! encode_signed {
                 b[0] ^= 0x80;
                 b
             }
+
+            fn decode(mut encoded: Self::Encoded) -> Self {
+                // Toggle top "sign" bit
+                encoded[0] ^= 0x80;
+                Self::from_be_bytes(encoded)
+            }
         }
     };
 }
@@ -69,6 +100,10 @@ macro_rules! encode_unsigned {
             fn encode(self) -> [u8; $n] {
                 self.to_be_bytes()
             }
+
+            fn decode(encoded: Self::Encoded) -> Self {
+                Self::from_be_bytes(encoded)
+            }
         }
     };
 }
@@ -87,6 +122,12 @@ impl FixedLengthEncoding for f16 {
         let val = s ^ (((s >> 15) as u16) >> 1) as i16;
         val.encode()
     }
+
+    fn decode(encoded: Self::Encoded) -> Self {
+        let bits = i16::decode(encoded);
+        let val = bits ^ (((bits >> 15) as u16) >> 1) as i16;
+        Self::from_bits(val as u16)
+    }
 }
 
 impl FixedLengthEncoding for f32 {
@@ -98,6 +139,12 @@ impl FixedLengthEncoding for f32 {
         let val = s ^ (((s >> 31) as u32) >> 1) as i32;
         val.encode()
     }
+
+    fn decode(encoded: Self::Encoded) -> Self {
+        let bits = i32::decode(encoded);
+        let val = bits ^ (((bits >> 31) as u32) >> 1) as i32;
+        Self::from_bits(val as u32)
+    }
 }
 
 impl FixedLengthEncoding for f64 {
@@ -109,32 +156,44 @@ impl FixedLengthEncoding for f64 {
         let val = s ^ (((s >> 63) as u64) >> 1) as i64;
         val.encode()
     }
+
+    fn decode(encoded: Self::Encoded) -> Self {
+        let bits = i64::decode(encoded);
+        let val = bits ^ (((bits >> 63) as u64) >> 1) as i64;
+        Self::from_bits(val as u64)
+    }
 }
 
-impl FixedLengthEncoding for Decimal128 {
-    type Encoded = [u8; 16];
+pub type RawDecimal128 = RawDecimal<16>;
+pub type RawDecimal256 = RawDecimal<32>;
 
-    fn encode(self) -> [u8; 16] {
-        let mut val = *self.raw_value();
-        // Convert to big endian representation
-        val.reverse();
-        // Toggle top "sign" bit to ensure consistent sort order
-        val[0] ^= 0x80;
-        val
+/// The raw bytes of a decimal
+#[derive(Copy, Clone)]
+pub struct RawDecimal<const N: usize>(pub [u8; N]);
+
+impl<const N: usize> ToByteSlice for RawDecimal<N> {
+    fn to_byte_slice(&self) -> &[u8] {
+        &self.0
     }
 }
 
-impl FixedLengthEncoding for Decimal256 {
-    type Encoded = [u8; 32];
+impl<const N: usize> FixedLengthEncoding for RawDecimal<N> {
+    type Encoded = [u8; N];
 
-    fn encode(self) -> [u8; 32] {
-        let mut val = *self.raw_value();
+    fn encode(self) -> [u8; N] {
+        let mut val = self.0;
         // Convert to big endian representation
         val.reverse();
         // Toggle top "sign" bit to ensure consistent sort order
         val[0] ^= 0x80;
         val
     }
+
+    fn decode(mut encoded: Self::Encoded) -> Self {
+        encoded[0] ^= 0x80;
+        encoded.reverse();
+        Self(encoded)
+    }
 }
 
 /// Returns the total encoded length (including null byte) for a value of type `T::Native`
@@ -166,9 +225,153 @@ pub fn encode<T: FixedLengthEncoding, I: IntoIterator<Item = Option<T>>>(
                 encoded.as_mut().iter_mut().for_each(|v| *v = !*v)
             }
             to_write[1..].copy_from_slice(encoded.as_ref())
-        } else if !opts.nulls_first {
-            out.buffer[*offset] = 0xFF;
+        } else {
+            out.buffer[*offset] = null_sentinel(opts);
         }
         *offset = end_offset;
     }
 }
+
+/// Splits `len` bytes from `src`
+#[inline]
+fn split_off<'a>(src: &mut &'a [u8], len: usize) -> &'a [u8] {
+    let v = &src[..len];
+    *src = &src[len..];
+    v
+}
+
+/// Decodes a `BooleanArray` from rows
+pub fn decode_bool(rows: &mut [&[u8]], options: SortOptions) -> BooleanArray {
+    let true_val = match options.descending {
+        true => !1,
+        false => 1,
+    };
+
+    let len = rows.len();
+
+    let mut null_count = 0;
+    let mut nulls = MutableBuffer::new(bit_util::ceil(len, 64) * 8);
+    let mut values = MutableBuffer::new(bit_util::ceil(len, 64) * 8);
+
+    let chunks = len / 64;
+    let remainder = len % 64;
+    for chunk in 0..chunks {
+        let mut null_packed = 0;
+        let mut values_packed = 0;
+
+        for bit_idx in 0..64 {
+            let i = split_off(&mut rows[bit_idx + chunk * 64], 2);
+            let (null, value) = (i[0] == 1, i[1] == true_val);
+            null_count += !null as usize;
+            null_packed |= (null as u64) << bit_idx;
+            values_packed |= (value as u64) << bit_idx;
+        }
+
+        nulls.push(null_packed);
+        values.push(values_packed);
+    }
+
+    if remainder != 0 {
+        let mut null_packed = 0;
+        let mut values_packed = 0;
+
+        for bit_idx in 0..remainder {
+            let i = split_off(&mut rows[bit_idx + chunks * 64], 2);
+            let (null, value) = (i[0] == 1, i[1] == true_val);
+            null_count += !null as usize;
+            null_packed |= (null as u64) << bit_idx;
+            values_packed |= (value as u64) << bit_idx;
+        }
+
+        nulls.push(null_packed);
+        values.push(values_packed);
+    }
+
+    let builder = ArrayDataBuilder::new(DataType::Boolean)
+        .len(rows.len())
+        .null_count(null_count)
+        .add_buffer(values.into())
+        .null_bit_buffer(Some(nulls.into()));
+
+    // SAFETY:
+    // Buffers are the correct length
+    unsafe { BooleanArray::from(builder.build_unchecked()) }
+}
+
+/// Decodes a `ArrayData` from rows based on the provided `FixedLengthEncoding` `T`
+fn decode_fixed<T: FixedLengthEncoding + ToByteSlice>(
+    rows: &mut [&[u8]],
+    data_type: DataType,
+    options: SortOptions,
+) -> ArrayData {
+    let len = rows.len();
+
+    let mut null_count = 0;
+    let mut nulls = MutableBuffer::new(bit_util::ceil(len, 64) * 8);
+    let mut values = MutableBuffer::new(std::mem::size_of::<T>() * len);
+
+    let chunks = len / 64;
+    let remainder = len % 64;
+    for chunk in 0..chunks {
+        let mut null_packed = 0;
+
+        for bit_idx in 0..64 {
+            let i = split_off(&mut rows[bit_idx + chunk * 64], T::ENCODED_LEN);
+            let null = i[0] == 1;
+            null_count += !null as usize;
+            null_packed |= (null as u64) << bit_idx;
+
+            let value = T::Encoded::from_slice(&i[1..], options.descending);
+            values.push(T::decode(value));
+        }
+
+        nulls.push(null_packed);
+    }
+
+    if remainder != 0 {
+        let mut null_packed = 0;
+
+        for bit_idx in 0..remainder {
+            let i = split_off(&mut rows[bit_idx + chunks * 64], T::ENCODED_LEN);
+            let null = i[0] == 1;
+            null_count += !null as usize;
+            null_packed |= (null as u64) << bit_idx;
+
+            let value = T::Encoded::from_slice(&i[1..], options.descending);
+            values.push(T::decode(value));
+        }
+
+        nulls.push(null_packed);
+    }
+
+    let builder = ArrayDataBuilder::new(data_type)
+        .len(rows.len())
+        .null_count(null_count)
+        .add_buffer(values.into())
+        .null_bit_buffer(Some(nulls.into()));
+
+    // SAFETY: Buffers correct length
+    unsafe { builder.build_unchecked() }
+}
+
+/// Decodes a `DecimalArray` from rows
+pub fn decode_decimal<const N: usize, T: DecimalType>(
+    rows: &mut [&[u8]],
+    options: SortOptions,
+    precision: u8,
+    scale: u8,
+) -> DecimalArray<T> {
+    decode_fixed::<RawDecimal<N>>(rows, T::TYPE_CONSTRUCTOR(precision, scale), options)
+        .into()
+}
+
+/// Decodes a `PrimitiveArray` from rows
+pub fn decode_primitive<T: ArrowPrimitiveType>(
+    rows: &mut [&[u8]],
+    options: SortOptions,
+) -> PrimitiveArray<T>
+where
+    T::Native: FixedLengthEncoding + ToByteSlice,
+{
+    decode_fixed::<T::Native>(rows, T::DATA_TYPE, options).into()
+}
diff --git a/arrow/src/row/interner.rs b/arrow/src/row/interner.rs
index e48670984..156d23465 100644
--- a/arrow/src/row/interner.rs
+++ b/arrow/src/row/interner.rs
@@ -22,7 +22,7 @@ use std::num::NonZeroU32;
 use std::ops::Index;
 
 /// An interned value
-#[derive(Debug, Copy, Clone, PartialEq, Eq)]
+#[derive(Debug, Copy, Clone, PartialEq, Eq, Hash)]
 pub struct Interned(NonZeroU32); // We use NonZeroU32 so that `Option<Interned>` is 32 bits
 
 /// A byte array interner that generates normalized keys that are sorted with respect
@@ -132,7 +132,6 @@ impl OrderPreservingInterner {
 
     /// Converts a normalized key returned by [`Self::normalized_key`] to [`Interned`]
     /// returning `None` if it cannot be found
-    #[allow(dead_code)]
     pub fn lookup(&self, normalized_key: &[u8]) -> Option<Interned> {
         let len = normalized_key.len();
 
@@ -159,7 +158,6 @@ impl OrderPreservingInterner {
     }
 
     /// Returns the interned value for a given [`Interned`]
-    #[allow(dead_code)]
     pub fn value(&self, key: Interned) -> &[u8] {
         self.values.index(key)
     }
diff --git a/arrow/src/row/mod.rs b/arrow/src/row/mod.rs
index 88c8a9166..f5ac57032 100644
--- a/arrow/src/row/mod.rs
+++ b/arrow/src/row/mod.rs
@@ -17,17 +17,28 @@
 
 //! A comparable row-oriented representation of a collection of [`Array`]
 
-use crate::array::{
-    as_boolean_array, as_generic_binary_array, as_largestring_array, as_string_array,
-    Array, ArrayRef, Decimal128Array, Decimal256Array,
-};
+use std::cmp::Ordering;
+use std::hash::{Hash, Hasher};
+use std::sync::Arc;
+
+use arrow_array::cast::*;
+use arrow_array::*;
+
 use crate::compute::SortOptions;
 use crate::datatypes::*;
 use crate::error::{ArrowError, Result};
-use crate::row::interner::{Interned, OrderPreservingInterner};
-use crate::util::decimal::{Decimal128, Decimal256};
+use crate::row::dictionary::{
+    compute_dictionary_mapping, decode_dictionary, encode_dictionary,
+};
+use crate::row::fixed::{
+    decode_bool, decode_decimal, decode_primitive, RawDecimal, RawDecimal128,
+    RawDecimal256,
+};
+use crate::row::interner::OrderPreservingInterner;
+use crate::row::variable::{decode_binary, decode_string};
 use crate::{downcast_dictionary_array, downcast_primitive_array};
 
+mod dictionary;
 mod fixed;
 mod interner;
 mod variable;
@@ -134,13 +145,13 @@ mod variable;
 /// [byte stuffing]:[https://en.wikipedia.org/wiki/High-Level_Data_Link_Control#Asynchronous_framing]
 #[derive(Debug)]
 pub struct RowConverter {
-    fields: Vec<SortField>,
+    fields: Arc<[SortField]>,
     /// interning state for column `i`, if column`i` is a dictionary
     interners: Vec<Option<Box<OrderPreservingInterner>>>,
 }
 
 /// Configure the data type and sort order for a given column
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, PartialEq, Eq)]
 pub struct SortField {
     /// Sort options
     options: SortOptions,
@@ -164,7 +175,10 @@ impl RowConverter {
     /// Create a new [`RowConverter`] with the provided schema
     pub fn new(fields: Vec<SortField>) -> Self {
         let interners = (0..fields.len()).map(|_| None).collect();
-        Self { fields, interners }
+        Self {
+            fields: fields.into(),
+            interners,
+        }
     }
 
     /// Convert [`ArrayRef`] columns into [`Rows`]
@@ -186,7 +200,7 @@ impl RowConverter {
         let dictionaries = columns
             .iter()
             .zip(&mut self.interners)
-            .zip(&self.fields)
+            .zip(self.fields.iter())
             .map(|((column, interner), field)| {
                 if !column.data_type().equals_datatype(&field.data_type) {
                     return Err(ArrowError::InvalidArgumentError(format!(
@@ -214,10 +228,10 @@ impl RowConverter {
             })
             .collect::<Result<Vec<_>>>()?;
 
-        let mut rows = new_empty_rows(columns, &dictionaries)?;
+        let mut rows = new_empty_rows(columns, &dictionaries, Arc::clone(&self.fields))?;
 
         for ((column, field), dictionary) in
-            columns.iter().zip(&self.fields).zip(dictionaries)
+            columns.iter().zip(self.fields.iter()).zip(dictionaries)
         {
             // We encode a column at a time to minimise dispatch overheads
             encode_column(&mut rows, column, field.options, dictionary.as_deref())
@@ -227,11 +241,44 @@ impl RowConverter {
             assert_eq!(*rows.offsets.last().unwrap(), rows.buffer.len());
             rows.offsets
                 .windows(2)
-                .for_each(|w| assert!(w[0] < w[1], "offsets should be monotonic"));
+                .for_each(|w| assert!(w[0] <= w[1], "offsets should be monotonic"));
         }
 
         Ok(rows)
     }
+
+    /// Convert [`Rows`] columns into [`ArrayRef`]
+    ///
+    /// # Panics
+    ///
+    /// Panics if the rows were not produced by this [`RowConverter`]
+    pub fn convert_rows<'a, I>(&self, rows: I) -> Result<Vec<ArrayRef>>
+    where
+        I: IntoIterator<Item = Row<'a>>,
+    {
+        let mut rows: Vec<_> = rows
+            .into_iter()
+            .map(|row| {
+                assert!(
+                    Arc::ptr_eq(row.fields, &self.fields),
+                    "rows were not produced by this RowConverter"
+                );
+
+                row.data
+            })
+            .collect();
+
+        self.fields
+            .iter()
+            .zip(&self.interners)
+            .map(|(field, interner)| {
+                // SAFETY
+                // We have validated that the rows came from this [`RowConverter`]
+                // and therefore must be valid
+                unsafe { decode_column(field, &mut rows, interner.as_deref()) }
+            })
+            .collect()
+    }
 }
 
 /// A row-oriented representation of arrow data, that is normalized for comparison
@@ -243,13 +290,18 @@ pub struct Rows {
     buffer: Box<[u8]>,
     /// Row `i` has data `&buffer[offsets[i]..offsets[i+1]]`
     offsets: Box<[usize]>,
+    /// The schema for these rows
+    fields: Arc<[SortField]>,
 }
 
 impl Rows {
     pub fn row(&self, row: usize) -> Row<'_> {
         let end = self.offsets[row + 1];
         let start = self.offsets[row];
-        Row(&self.buffer[start..end])
+        Row {
+            data: &self.buffer[start..end],
+            fields: &self.fields,
+        }
     }
 
     pub fn num_rows(&self) -> usize {
@@ -257,54 +309,127 @@ impl Rows {
     }
 }
 
+impl<'a> IntoIterator for &'a Rows {
+    type Item = Row<'a>;
+    type IntoIter = RowsIter<'a>;
+
+    fn into_iter(self) -> Self::IntoIter {
+        RowsIter {
+            rows: self,
+            start: 0,
+            end: self.num_rows(),
+        }
+    }
+}
+
+/// An iterator over [`Rows`]
+#[derive(Debug)]
+pub struct RowsIter<'a> {
+    rows: &'a Rows,
+    start: usize,
+    end: usize,
+}
+
+impl<'a> Iterator for RowsIter<'a> {
+    type Item = Row<'a>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        if self.end == self.start {
+            return None;
+        }
+        let row = self.rows.row(self.start);
+        self.start += 1;
+        Some(row)
+    }
+
+    fn size_hint(&self) -> (usize, Option<usize>) {
+        let len = self.len();
+        (len, Some(len))
+    }
+}
+
+impl<'a> ExactSizeIterator for RowsIter<'a> {
+    fn len(&self) -> usize {
+        self.end - self.start
+    }
+}
+
+impl<'a> DoubleEndedIterator for RowsIter<'a> {
+    fn next_back(&mut self) -> Option<Self::Item> {
+        if self.end == self.start {
+            return None;
+        }
+        let row = self.rows.row(self.end);
+        self.end -= 1;
+        Some(row)
+    }
+}
+
 /// A comparable representation of a row
 ///
 /// Two [`Row`] can be compared if they both belong to [`Rows`] returned by calls to
 /// [`RowConverter::convert_columns`] on the same [`RowConverter`]
 ///
 /// Otherwise any ordering established by comparing the [`Row`] is arbitrary
-#[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
-pub struct Row<'a>(&'a [u8]);
+#[derive(Debug, Copy, Clone)]
+pub struct Row<'a> {
+    data: &'a [u8],
+    fields: &'a Arc<[SortField]>,
+}
+
+// Manually derive these as don't wish to include `fields`
+
+impl<'a> PartialEq for Row<'a> {
+    #[inline]
+    fn eq(&self, other: &Self) -> bool {
+        self.data.eq(other.data)
+    }
+}
+
+impl<'a> Eq for Row<'a> {}
+
+impl<'a> PartialOrd for Row<'a> {
+    #[inline]
+    fn partial_cmp(&self, other: &Self) -> Option<Ordering> {
+        self.data.partial_cmp(other.data)
+    }
+}
+
+impl<'a> Ord for Row<'a> {
+    #[inline]
+    fn cmp(&self, other: &Self) -> Ordering {
+        self.data.cmp(other.data)
+    }
+}
+
+impl<'a> Hash for Row<'a> {
+    #[inline]
+    fn hash<H: Hasher>(&self, state: &mut H) {
+        self.data.hash(state)
+    }
+}
 
 impl<'a> AsRef<[u8]> for Row<'a> {
+    #[inline]
     fn as_ref(&self) -> &[u8] {
-        self.0
+        self.data
     }
 }
 
-/// Computes the dictionary mapping for the given dictionary values
-fn compute_dictionary_mapping(
-    interner: &mut OrderPreservingInterner,
-    values: &ArrayRef,
-) -> Result<Vec<Option<Interned>>> {
-    use fixed::FixedLengthEncoding;
-    Ok(downcast_primitive_array! {
-        values => interner
-            .intern(values.iter().map(|x| x.map(|x| x.encode()))),
-        DataType::Binary => {
-            let iter = as_generic_binary_array::<i64>(values).iter();
-            interner.intern(iter)
-        }
-        DataType::LargeBinary => {
-            let iter = as_generic_binary_array::<i64>(values).iter();
-            interner.intern(iter)
-        }
-        DataType::Utf8 => {
-            let iter = as_string_array(values).iter().map(|x| x.map(|x| x.as_bytes()));
-            interner.intern(iter)
-        }
-        DataType::LargeUtf8 => {
-            let iter = as_largestring_array(values).iter().map(|x| x.map(|x| x.as_bytes()));
-            interner.intern(iter)
-        }
-        t => return Err(ArrowError::NotYetImplemented(format!("dictionary value {} is not supported", t))),
-    })
+/// Returns the null sentinel, negated if `invert` is true
+#[inline]
+fn null_sentinel(options: SortOptions) -> u8 {
+    match options.nulls_first {
+        true => 0,
+        false => 0xFF,
+    }
 }
 
 /// Computes the length of each encoded [`Rows`] and returns an empty [`Rows`]
 fn new_empty_rows(
     cols: &[ArrayRef],
     dictionaries: &[Option<Vec<Option<&[u8]>>>],
+    fields: Arc<[SortField]>,
 ) -> Result<Rows> {
     use fixed::FixedLengthEncoding;
 
@@ -314,10 +439,10 @@ fn new_empty_rows(
     for (array, dict) in cols.iter().zip(dictionaries) {
         downcast_primitive_array! {
             array => lengths.iter_mut().for_each(|x| *x += fixed::encoded_len(array)),
-            DataType::Null => lengths.iter_mut().for_each(|x| *x += 1),
+            DataType::Null => {},
             DataType::Boolean => lengths.iter_mut().for_each(|x| *x += bool::ENCODED_LEN),
-            DataType::Decimal128(_, _) => lengths.iter_mut().for_each(|x| *x += Decimal128::ENCODED_LEN),
-            DataType::Decimal256(_, _) => lengths.iter_mut().for_each(|x| *x += Decimal256::ENCODED_LEN),
+            DataType::Decimal128(_, _) => lengths.iter_mut().for_each(|x| *x += RawDecimal128::ENCODED_LEN),
+            DataType::Decimal256(_, _) => lengths.iter_mut().for_each(|x| *x += RawDecimal256::ENCODED_LEN),
             DataType::Binary => as_generic_binary_array::<i32>(array)
                 .iter()
                 .zip(lengths.iter_mut())
@@ -383,6 +508,7 @@ fn new_empty_rows(
     Ok(Rows {
         buffer: buffer.into(),
         offsets: offsets.into(),
+        fields,
     })
 }
 
@@ -395,20 +521,28 @@ fn encode_column(
 ) {
     downcast_primitive_array! {
         column => fixed::encode(out, column, opts),
-        DataType::Null => {
-            fixed::encode(out, std::iter::repeat(None::<bool>).take(column.len()), opts)
-        }
+        DataType::Null => {}
         DataType::Boolean => fixed::encode(out, as_boolean_array(column), opts),
-        DataType::Decimal128(_, _) => fixed::encode(
-            out,
-            column.as_any().downcast_ref::<Decimal128Array>().unwrap(),
-            opts,
-        ),
-        DataType::Decimal256(_, _) => fixed::encode(
-            out,
-            column.as_any().downcast_ref::<Decimal256Array>().unwrap(),
-            opts,
-        ),
+        DataType::Decimal128(_, _) => {
+            let iter = column
+                .as_any()
+                .downcast_ref::<Decimal128Array>()
+                .unwrap()
+                .into_iter()
+                .map(|x| x.map(|x| RawDecimal(*x.raw_value())));
+
+            fixed::encode(out, iter, opts)
+        },
+        DataType::Decimal256(_, _) => {
+            let iter = column
+                .as_any()
+                .downcast_ref::<Decimal256Array>()
+                .unwrap()
+                .into_iter()
+                .map(|x| x.map(|x| RawDecimal(*x.raw_value())));
+
+            fixed::encode(out, iter, opts)
+        },
         DataType::Binary => {
             variable::encode(out, as_generic_binary_array::<i32>(column).iter(), opts)
         }
@@ -428,37 +562,183 @@ fn encode_column(
             opts,
         ),
         DataType::Dictionary(_, _) => downcast_dictionary_array! {
-            column => {
-                let dict = dictionary.unwrap();
-                for (offset, k) in out.offsets.iter_mut().skip(1).zip(column.keys()) {
-                    match k.and_then(|k| dict[k as usize]) {
-                        Some(v) => {
-                            let end_offset = *offset + 1 + v.len();
-                            out.buffer[*offset] = 1;
-                            out.buffer[*offset+1..end_offset].copy_from_slice(v);
-                            if opts.descending {
-                                out.buffer[*offset..end_offset].iter_mut().for_each(|v| *v = !*v)
-                            }
-                            *offset = end_offset;
-                        }
-                        None => {
-                            if !opts.nulls_first {
-                                out.buffer[*offset] = 0xFF;
-                            }
-                            *offset += 1;
-                        }
-                    }
-                }
-            },
+            column => encode_dictionary(out, column, dictionary.unwrap(), opts),
             _ => unreachable!()
         }
         t => unimplemented!("not yet implemented: {}", t)
     }
 }
 
+/// Decodes a the provided `field` from `rows`
+///
+/// # Safety
+///
+/// Rows must contain valid data for the provided field
+unsafe fn decode_column(
+    field: &SortField,
+    rows: &mut [&[u8]],
+    interner: Option<&OrderPreservingInterner>,
+) -> Result<ArrayRef> {
+    let options = field.options;
+    let array: ArrayRef = match &field.data_type {
+        DataType::Null => Arc::new(NullArray::new(rows.len())),
+        DataType::Boolean => Arc::new(decode_bool(rows, options)),
+        DataType::Int8 => Arc::new(decode_primitive::<Int8Type>(rows, options)),
+        DataType::Int16 => Arc::new(decode_primitive::<Int16Type>(rows, options)),
+        DataType::Int32 => Arc::new(decode_primitive::<Int32Type>(rows, options)),
+        DataType::Int64 => Arc::new(decode_primitive::<Int64Type>(rows, options)),
+        DataType::UInt8 => Arc::new(decode_primitive::<UInt8Type>(rows, options)),
+        DataType::UInt16 => Arc::new(decode_primitive::<UInt16Type>(rows, options)),
+        DataType::UInt32 => Arc::new(decode_primitive::<UInt32Type>(rows, options)),
+        DataType::UInt64 => Arc::new(decode_primitive::<UInt64Type>(rows, options)),
+        DataType::Float16 => Arc::new(decode_primitive::<Float16Type>(rows, options)),
+        DataType::Float32 => Arc::new(decode_primitive::<Float32Type>(rows, options)),
+        DataType::Float64 => Arc::new(decode_primitive::<Float64Type>(rows, options)),
+        DataType::Timestamp(TimeUnit::Second, _) => {
+            Arc::new(decode_primitive::<TimestampSecondType>(rows, options))
+        }
+        DataType::Timestamp(TimeUnit::Millisecond, _) => {
+            Arc::new(decode_primitive::<TimestampMillisecondType>(rows, options))
+        }
+        DataType::Timestamp(TimeUnit::Microsecond, _) => {
+            Arc::new(decode_primitive::<TimestampMicrosecondType>(rows, options))
+        }
+        DataType::Timestamp(TimeUnit::Nanosecond, _) => {
+            Arc::new(decode_primitive::<TimestampNanosecondType>(rows, options))
+        }
+        DataType::Date32 => Arc::new(decode_primitive::<Date32Type>(rows, options)),
+        DataType::Date64 => Arc::new(decode_primitive::<Date64Type>(rows, options)),
+        DataType::Time32(t) => match t {
+            TimeUnit::Second => {
+                Arc::new(decode_primitive::<Time32SecondType>(rows, options))
+            }
+            TimeUnit::Millisecond => {
+                Arc::new(decode_primitive::<Time32MillisecondType>(rows, options))
+            }
+            _ => unreachable!(),
+        },
+        DataType::Time64(t) => match t {
+            TimeUnit::Microsecond => {
+                Arc::new(decode_primitive::<Time64MicrosecondType>(rows, options))
+            }
+            TimeUnit::Nanosecond => {
+                Arc::new(decode_primitive::<Time64NanosecondType>(rows, options))
+            }
+            _ => unreachable!(),
+        },
+        DataType::Duration(TimeUnit::Second) => {
+            Arc::new(decode_primitive::<DurationSecondType>(rows, options))
+        }
+        DataType::Duration(TimeUnit::Millisecond) => {
+            Arc::new(decode_primitive::<DurationMillisecondType>(rows, options))
+        }
+        DataType::Duration(TimeUnit::Microsecond) => {
+            Arc::new(decode_primitive::<DurationMicrosecondType>(rows, options))
+        }
+        DataType::Duration(TimeUnit::Nanosecond) => {
+            Arc::new(decode_primitive::<DurationNanosecondType>(rows, options))
+        }
+        DataType::Interval(IntervalUnit::DayTime) => {
+            Arc::new(decode_primitive::<IntervalDayTimeType>(rows, options))
+        }
+        DataType::Interval(IntervalUnit::MonthDayNano) => {
+            Arc::new(decode_primitive::<IntervalMonthDayNanoType>(rows, options))
+        }
+        DataType::Interval(IntervalUnit::YearMonth) => {
+            Arc::new(decode_primitive::<IntervalYearMonthType>(rows, options))
+        }
+        DataType::Binary => Arc::new(decode_binary::<i32>(rows, options)),
+        DataType::LargeBinary => Arc::new(decode_binary::<i64>(rows, options)),
+        DataType::Utf8 => Arc::new(decode_string::<i32>(rows, options)),
+        DataType::LargeUtf8 => Arc::new(decode_string::<i64>(rows, options)),
+        DataType::Decimal128(p, s) => {
+            Arc::new(decode_decimal::<16, Decimal128Type>(rows, options, *p, *s))
+        }
+        DataType::Decimal256(p, s) => {
+            Arc::new(decode_decimal::<32, Decimal256Type>(rows, options, *p, *s))
+        }
+        DataType::Dictionary(k, v) => match k.as_ref() {
+            DataType::Int8 => Arc::new(decode_dictionary::<Int8Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            DataType::Int16 => Arc::new(decode_dictionary::<Int16Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            DataType::Int32 => Arc::new(decode_dictionary::<Int32Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            DataType::Int64 => Arc::new(decode_dictionary::<Int64Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            DataType::UInt8 => Arc::new(decode_dictionary::<UInt8Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            DataType::UInt16 => Arc::new(decode_dictionary::<UInt16Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            DataType::UInt32 => Arc::new(decode_dictionary::<UInt32Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            DataType::UInt64 => Arc::new(decode_dictionary::<UInt64Type>(
+                interner.unwrap(),
+                v.as_ref(),
+                options,
+                rows,
+            )?),
+            _ => {
+                return Err(ArrowError::InvalidArgumentError(format!(
+                    "{} is not a valid dictionary key type",
+                    field.data_type
+                )));
+            }
+        },
+        DataType::FixedSizeBinary(_)
+        | DataType::List(_)
+        | DataType::FixedSizeList(_, _)
+        | DataType::LargeList(_)
+        | DataType::Struct(_)
+        | DataType::Union(_, _, _)
+        | DataType::Map(_, _) => {
+            return Err(ArrowError::NotYetImplemented(format!(
+                "converting {} row is not supported",
+                field.data_type
+            )))
+        }
+    };
+    Ok(array)
+}
+
 #[cfg(test)]
 mod tests {
-    use super::*;
+    use std::sync::Arc;
+
+    use rand::distributions::uniform::SampleUniform;
+    use rand::distributions::{Distribution, Standard};
+    use rand::{thread_rng, Rng};
+
+    use arrow_array::NullArray;
+
     use crate::array::{
         BinaryArray, BooleanArray, DictionaryArray, Float32Array, GenericStringArray,
         Int16Array, Int32Array, OffsetSizeTrait, PrimitiveArray,
@@ -466,10 +746,8 @@ mod tests {
     };
     use crate::compute::{LexicographicalComparator, SortColumn};
     use crate::util::display::array_value_to_string;
-    use rand::distributions::uniform::SampleUniform;
-    use rand::distributions::{Distribution, Standard};
-    use rand::{thread_rng, Rng};
-    use std::sync::Arc;
+
+    use super::*;
 
     #[test]
     fn test_fixed_width() {
@@ -525,19 +803,29 @@ mod tests {
         assert!(rows.row(0) < rows.row(1));
         assert!(rows.row(3) < rows.row(0));
         assert!(rows.row(4) < rows.row(1));
-        assert!(rows.row(5) < rows.row(4))
+        assert!(rows.row(5) < rows.row(4));
+
+        let back = converter.convert_rows(&rows).unwrap();
+        for (expected, actual) in cols.iter().zip(&back) {
+            assert_eq!(expected, actual);
+        }
     }
 
     #[test]
     fn test_bool() {
         let mut converter = RowConverter::new(vec![SortField::new(DataType::Boolean)]);
 
-        let col = Arc::new(BooleanArray::from_iter([None, Some(false), Some(true)]));
-        let rows = converter.convert_columns(&[col]).unwrap();
+        let col = Arc::new(BooleanArray::from_iter([None, Some(false), Some(true)]))
+            as ArrayRef;
+
+        let rows = converter.convert_columns(&[Arc::clone(&col)]).unwrap();
         assert!(rows.row(2) > rows.row(1));
         assert!(rows.row(2) > rows.row(0));
         assert!(rows.row(1) > rows.row(0));
 
+        let cols = converter.convert_rows(&rows).unwrap();
+        assert_eq!(&cols[0], &col);
+
         let mut converter = RowConverter::new(vec![SortField::new_with_options(
             DataType::Boolean,
             SortOptions {
@@ -546,11 +834,21 @@ mod tests {
             },
         )]);
 
-        let col = Arc::new(BooleanArray::from_iter([None, Some(false), Some(true)]));
-        let rows = converter.convert_columns(&[col]).unwrap();
+        let rows = converter.convert_columns(&[Arc::clone(&col)]).unwrap();
         assert!(rows.row(2) < rows.row(1));
         assert!(rows.row(2) < rows.row(0));
         assert!(rows.row(1) < rows.row(0));
+        let cols = converter.convert_rows(&rows).unwrap();
+        assert_eq!(&cols[0], &col);
+    }
+
+    #[test]
+    fn test_null_encoding() {
+        let col = Arc::new(NullArray::new(10));
+        let mut converter = RowConverter::new(vec![SortField::new(DataType::Null)]);
+        let rows = converter.convert_columns(&[col]).unwrap();
+        assert_eq!(rows.num_rows(), 10);
+        assert_eq!(rows.row(1).data.len(), 0);
     }
 
     #[test]
@@ -561,16 +859,19 @@ mod tests {
             None,
             Some("foo"),
             Some(""),
-        ]));
+        ])) as ArrayRef;
 
         let mut converter = RowConverter::new(vec![SortField::new(DataType::Utf8)]);
-        let rows = converter.convert_columns(&[col]).unwrap();
+        let rows = converter.convert_columns(&[Arc::clone(&col)]).unwrap();
 
         assert!(rows.row(1) < rows.row(0));
         assert!(rows.row(2) < rows.row(4));
         assert!(rows.row(3) < rows.row(0));
         assert!(rows.row(3) < rows.row(1));
 
+        let cols = converter.convert_rows(&rows).unwrap();
+        assert_eq!(&cols[0], &col);
+
         let col = Arc::new(BinaryArray::from_iter([
             None,
             Some(vec![0_u8; 0]),
@@ -601,6 +902,9 @@ mod tests {
             }
         }
 
+        let cols = converter.convert_rows(&rows).unwrap();
+        assert_eq!(&cols[0], &col);
+
         let mut converter = RowConverter::new(vec![SortField::new_with_options(
             DataType::Binary,
             SortOptions {
@@ -608,7 +912,7 @@ mod tests {
                 nulls_first: false,
             },
         )]);
-        let rows = converter.convert_columns(&[col]).unwrap();
+        let rows = converter.convert_columns(&[Arc::clone(&col)]).unwrap();
 
         for i in 0..rows.num_rows() {
             for j in i + 1..rows.num_rows() {
@@ -622,6 +926,9 @@ mod tests {
                 );
             }
         }
+
+        let cols = converter.convert_rows(&rows).unwrap();
+        assert_eq!(&cols[0], &col);
     }
 
     #[test]
@@ -650,17 +957,23 @@ mod tests {
         assert_eq!(rows_a.row(1), rows_a.row(6));
         assert_eq!(rows_a.row(1), rows_a.row(7));
 
+        let cols = converter.convert_rows(&rows_a).unwrap();
+        assert_eq!(&cols[0], &a);
+
         let b = Arc::new(DictionaryArray::<Int32Type>::from_iter([
             Some("hello"),
             None,
             Some("cupcakes"),
-        ]));
+        ])) as ArrayRef;
 
-        let rows_b = converter.convert_columns(&[b]).unwrap();
+        let rows_b = converter.convert_columns(&[Arc::clone(&b)]).unwrap();
         assert_eq!(rows_a.row(1), rows_b.row(0));
         assert_eq!(rows_a.row(3), rows_b.row(1));
         assert!(rows_b.row(2) < rows_a.row(0));
 
+        let cols = converter.convert_rows(&rows_b).unwrap();
+        assert_eq!(&cols[0], &b);
+
         let mut converter = RowConverter::new(vec![SortField::new_with_options(
             a.data_type().clone(),
             SortOptions {
@@ -669,11 +982,14 @@ mod tests {
             },
         )]);
 
-        let rows_c = converter.convert_columns(&[a]).unwrap();
+        let rows_c = converter.convert_columns(&[Arc::clone(&a)]).unwrap();
         assert!(rows_c.row(3) > rows_c.row(5));
         assert!(rows_c.row(2) > rows_c.row(1));
         assert!(rows_c.row(0) > rows_c.row(1));
         assert!(rows_c.row(3) > rows_c.row(0));
+
+        let cols = converter.convert_rows(&rows_c).unwrap();
+        assert_eq!(&cols[0], &a);
     }
 
     #[test]
@@ -727,6 +1043,17 @@ mod tests {
         assert!(rows.row(3) < rows.row(0));
     }
 
+    #[test]
+    #[should_panic(expected = "rows were not produced by this RowConverter")]
+    fn test_different_converter() {
+        let values = Arc::new(Int32Array::from_iter([Some(1), Some(-1)]));
+        let mut converter = RowConverter::new(vec![SortField::new(DataType::Int32)]);
+        let rows = converter.convert_columns(&[values]).unwrap();
+
+        let converter = RowConverter::new(vec![SortField::new(DataType::Int32)]);
+        let _ = converter.convert_rows(&rows);
+    }
+
     fn generate_primitive_array<K>(len: usize, valid_percent: f64) -> PrimitiveArray<K>
     where
         K: ArrowPrimitiveType,
@@ -888,6 +1215,11 @@ mod tests {
                     );
                 }
             }
+
+            let back = converter.convert_rows(&rows).unwrap();
+            for (actual, expected) in back.iter().zip(&arrays) {
+                assert_eq!(actual, expected)
+            }
         }
     }
 }
diff --git a/arrow/src/row/variable.rs b/arrow/src/row/variable.rs
index 2213dad9e..36f337e65 100644
--- a/arrow/src/row/variable.rs
+++ b/arrow/src/row/variable.rs
@@ -16,12 +16,26 @@
 // under the License.
 
 use crate::compute::SortOptions;
-use crate::row::Rows;
+use crate::row::{null_sentinel, Rows};
 use crate::util::bit_util::ceil;
+use arrow_array::builder::BufferBuilder;
+use arrow_array::{Array, GenericBinaryArray, GenericStringArray, OffsetSizeTrait};
+use arrow_buffer::MutableBuffer;
+use arrow_data::ArrayDataBuilder;
+use arrow_schema::DataType;
 
 /// The block size of the variable length encoding
 pub const BLOCK_SIZE: usize = 32;
 
+/// The continuation token
+pub const BLOCK_CONTINUATION: u8 = 0xFF;
+
+/// Indicates an empty string
+pub const EMPTY_SENTINEL: u8 = 1;
+
+/// Indicates a non-empty string
+pub const NON_EMPTY_SENTINEL: u8 = 2;
+
 /// Returns the length of the encoded representation of a byte array, including the null byte
 pub fn encoded_len(a: Option<&[u8]>) -> usize {
     match a {
@@ -50,8 +64,8 @@ pub fn encode<'a, I: Iterator<Item = Option<&'a [u8]>>>(
         match maybe_val {
             Some(val) if val.is_empty() => {
                 out.buffer[*offset] = match opts.descending {
-                    true => !1,
-                    false => 1,
+                    true => !EMPTY_SENTINEL,
+                    false => EMPTY_SENTINEL,
                 };
                 *offset += 1;
             }
@@ -61,7 +75,7 @@ pub fn encode<'a, I: Iterator<Item = Option<&'a [u8]>>>(
                 let to_write = &mut out.buffer[*offset..end_offset];
 
                 // Write `2_u8` to demarcate as non-empty, non-null string
-                to_write[0] = 2;
+                to_write[0] = NON_EMPTY_SENTINEL;
 
                 let chunks = val.chunks_exact(BLOCK_SIZE);
                 let remainder = chunks.remainder();
@@ -76,7 +90,7 @@ pub fn encode<'a, I: Iterator<Item = Option<&'a [u8]>>>(
                     *out_block = *input;
 
                     // Indicate that there are further blocks to follow
-                    output[BLOCK_SIZE] = u8::MAX;
+                    output[BLOCK_SIZE] = BLOCK_CONTINUATION;
                 }
 
                 if !remainder.is_empty() {
@@ -97,11 +111,121 @@ pub fn encode<'a, I: Iterator<Item = Option<&'a [u8]>>>(
                 }
             }
             None => {
-                if !opts.nulls_first {
-                    out.buffer[*offset] = 0xFF;
-                }
+                out.buffer[*offset] = null_sentinel(opts);
                 *offset += 1;
             }
         }
     }
 }
+
+/// Returns the number of bytes of encoded data
+fn decoded_len(row: &[u8], options: SortOptions) -> usize {
+    let (non_empty_sentinel, continuation) = match options.descending {
+        true => (!NON_EMPTY_SENTINEL, !BLOCK_CONTINUATION),
+        false => (NON_EMPTY_SENTINEL, BLOCK_CONTINUATION),
+    };
+
+    if row[0] != non_empty_sentinel {
+        // Empty or null string
+        return 0;
+    }
+
+    let mut str_len = 0;
+    let mut idx = 1;
+    loop {
+        let sentinel = row[idx + BLOCK_SIZE];
+        if sentinel == continuation {
+            idx += BLOCK_SIZE + 1;
+            str_len += BLOCK_SIZE;
+            continue;
+        }
+        let block_len = match options.descending {
+            true => !sentinel,
+            false => sentinel,
+        };
+        return str_len + block_len as usize;
+    }
+}
+
+/// Decodes a binary array from `rows` with the provided `options`
+pub fn decode_binary<I: OffsetSizeTrait>(
+    rows: &mut [&[u8]],
+    options: SortOptions,
+) -> GenericBinaryArray<I> {
+    let len = rows.len();
+    let mut null_count = 0;
+    let nulls = MutableBuffer::collect_bool(len, |x| {
+        let valid = rows[x][0] != null_sentinel(options);
+        null_count += !valid as usize;
+        valid
+    });
+
+    let values_capacity = rows.iter().map(|row| decoded_len(row, options)).sum();
+    let mut offsets = BufferBuilder::<I>::new(len + 1);
+    offsets.append(I::zero());
+    let mut values = MutableBuffer::new(values_capacity);
+
+    for row in rows {
+        let str_length = decoded_len(row, options);
+        let mut to_read = str_length;
+        let mut offset = 1;
+        while to_read >= BLOCK_SIZE {
+            to_read -= BLOCK_SIZE;
+
+            values.extend_from_slice(&row[offset..offset + BLOCK_SIZE]);
+            offset += BLOCK_SIZE + 1;
+        }
+
+        if to_read != 0 {
+            values.extend_from_slice(&row[offset..offset + to_read]);
+            offset += BLOCK_SIZE + 1;
+        }
+        *row = &row[offset..];
+
+        offsets.append(I::from_usize(values.len()).expect("offset overflow"))
+    }
+
+    if options.descending {
+        values.as_slice_mut().iter_mut().for_each(|o| *o = !*o)
+    }
+
+    let d = match I::IS_LARGE {
+        true => DataType::LargeBinary,
+        false => DataType::Binary,
+    };
+
+    let builder = ArrayDataBuilder::new(d)
+        .len(len)
+        .null_count(null_count)
+        .null_bit_buffer(Some(nulls.into()))
+        .add_buffer(offsets.finish())
+        .add_buffer(values.into());
+
+    // SAFETY:
+    // Valid by construction above
+    unsafe { GenericBinaryArray::from(builder.build_unchecked()) }
+}
+
+/// Decodes a string array from `rows` with the provided `options`
+///
+/// # Safety
+///
+/// The row must contain valid UTF-8 data
+pub unsafe fn decode_string<I: OffsetSizeTrait>(
+    rows: &mut [&[u8]],
+    options: SortOptions,
+) -> GenericStringArray<I> {
+    let d = match I::IS_LARGE {
+        true => DataType::LargeUtf8,
+        false => DataType::Utf8,
+    };
+
+    let builder = decode_binary::<I>(rows, options)
+        .into_data()
+        .into_builder()
+        .data_type(d);
+
+    // SAFETY:
+    // Row data must have come from a valid UTF-8 array
+    GenericStringArray::from(builder.build_unchecked())
+}