This is an automated email from the ASF dual-hosted git repository.
etseidl pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/arrow-rs.git
The following commit(s) were added to refs/heads/main by this push:
new 4443bd4579 [Parquet] Add row group distinct counts to
`StatisticsConverter` (#10652)
4443bd4579 is described below
commit 4443bd45790ba44b81cd19a2d3b25e448a60d98d
Author: RIchard Baah <[email protected]>
AuthorDate: Tue Aug 18 15:58:48 2026 -0400
[Parquet] Add row group distinct counts to `StatisticsConverter` (#10652)
# Which issue does this PR close?
- works towards #10650.
- works towards [#24114 &
](https://github.com/apache/datafusion/issues/24114) &
https://github.com/apache/datafusion/issues/22891
# Rationale for this change
see #10650
<!--
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.
-->
# What changes are included in this PR?
Added StatisticsConverter::row_group_distinct_counts() to the parquet
crate, which reads the distinct_count field from row group statistics
and returns a UInt64Array with one entry per row group.
<!--
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?
yes this PR includes two test that assert that row group stats are being
propogated
<!--
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)?
If this PR claims a performance improvement, please include evidence
such as benchmark results.
-->
# Are there any user-facing changes?
yes, adds new public method for reading row group distinct stats
<!--
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 call them out.
-->
---
parquet/src/arrow/arrow_reader/statistics.rs | 22 +++++
parquet/tests/arrow_reader/statistics.rs | 132 ++++++++++++++++++++++++++-
2 files changed, 153 insertions(+), 1 deletion(-)
diff --git a/parquet/src/arrow/arrow_reader/statistics.rs
b/parquet/src/arrow/arrow_reader/statistics.rs
index 8436a70745..7c628e9688 100644
--- a/parquet/src/arrow/arrow_reader/statistics.rs
+++ b/parquet/src/arrow/arrow_reader/statistics.rs
@@ -1820,6 +1820,28 @@ impl<'a> StatisticsConverter<'a> {
Ok(UInt64Array::from_iter(nan_counts))
}
+ /// Extract the distinct counts from row group statistics in
[`RowGroupMetaData`]
+ ///
+ /// See docs on [`Self::row_group_mins`] for details
+ pub fn row_group_distinct_counts<I>(&self, metadatas: I) ->
Result<UInt64Array>
+ where
+ I: IntoIterator<Item = &'a RowGroupMetaData>,
+ {
+ let Some(parquet_index) = self.parquet_column_index else {
+ let num_row_groups = metadatas.into_iter().count();
+ return Ok(UInt64Array::from_iter(std::iter::repeat_n(
+ None,
+ num_row_groups,
+ )));
+ };
+
+ let distinct_counts = metadatas
+ .into_iter()
+ .map(|x| x.column(parquet_index).statistics())
+ .map(|s| s.and_then(|s| s.distinct_count_opt()));
+ Ok(UInt64Array::from_iter(distinct_counts))
+ }
+
/// Extract the minimum values from Data Page statistics.
///
/// In Parquet files, in addition to the Column Chunk level statistics
diff --git a/parquet/tests/arrow_reader/statistics.rs
b/parquet/tests/arrow_reader/statistics.rs
index 0bc61ba108..e0740d38e9 100644
--- a/parquet/tests/arrow_reader/statistics.rs
+++ b/parquet/tests/arrow_reader/statistics.rs
@@ -2630,10 +2630,13 @@ mod test {
Int32Array, Int64Array, RecordBatch, StringArray, StructArray,
TimestampNanosecondArray,
new_empty_array,
};
- use arrow_schema::{DataType, SchemaRef, TimeUnit};
+ use arrow_schema::{DataType, Field, SchemaRef, TimeUnit};
use bytes::Bytes;
use parquet::arrow::parquet_column;
+ use parquet::data_type::{ByteArray, ByteArrayType, Int32Type};
use parquet::file::metadata::{ParquetMetaData, RowGroupMetaData};
+ use parquet::file::writer::SerializedFileWriter;
+ use parquet::schema::parser::parse_message_type;
use std::path::PathBuf;
use std::sync::Arc;
// TODO error cases (with parquet statistics that are mismatched in
expected type)
@@ -3222,4 +3225,131 @@ mod test {
.collect();
Arc::new(array)
}
+
+ // Verifies that distinct_count is correctly read back from UTF-8 column
statistics.
+ // Uses the low-level writer to inject a known distinct_count into the
parquet footer
+ // since ArrowWriter does not yet write this field.
+ #[test]
+ fn test_row_group_distinct_counts_utf8_roundtrip() {
+ let unique_string_values: Vec<ByteArray> = (0..10)
+ .map(|index|
ByteArray::from(format!("value_{index}").into_bytes()))
+ .collect();
+
+ let parquet_schema = Arc::new(
+ parse_message_type("message schema { REQUIRED BYTE_ARRAY col
(UTF8); }").unwrap(),
+ );
+ let writer_properties = Arc::new(
+ WriterProperties::builder()
+ .set_statistics_enabled(EnabledStatistics::Chunk)
+ .build(),
+ );
+
+ let mut file_buffer: Vec<u8> = Vec::new();
+ let mut file_writer =
+ SerializedFileWriter::new(&mut file_buffer, parquet_schema,
writer_properties).unwrap();
+
+ let mut row_group_writer = file_writer.next_row_group().unwrap();
+ let mut column_writer =
row_group_writer.next_column().unwrap().unwrap();
+ let min_value = unique_string_values.first().unwrap().clone();
+ let max_value = unique_string_values.last().unwrap().clone();
+ let expected_distinct_count = unique_string_values.len() as u64;
+ column_writer
+ .typed::<ByteArrayType>()
+ .write_batch_with_statistics(
+ &unique_string_values,
+ None,
+ None,
+ Some(&min_value),
+ Some(&max_value),
+ Some(expected_distinct_count),
+ )
+ .unwrap();
+ column_writer.close().unwrap();
+ row_group_writer.close().unwrap();
+ file_writer.close().unwrap();
+
+ let parquet_bytes = Bytes::from(file_buffer);
+ let reader_builder =
ParquetRecordBatchReaderBuilder::try_new(parquet_bytes).unwrap();
+ let file_metadata = reader_builder.metadata().clone();
+ let arrow_schema = reader_builder.schema().clone();
+ let parquet_schema_descriptor =
file_metadata.file_metadata().schema_descr();
+
+ let statistics_converter =
+ StatisticsConverter::try_new("col", &arrow_schema,
parquet_schema_descriptor).unwrap();
+ let distinct_counts = statistics_converter
+ .row_group_distinct_counts(file_metadata.row_groups().iter())
+ .unwrap();
+
+ assert_eq!(
+ distinct_counts,
+ UInt64Array::from(vec![Some(expected_distinct_count)]),
+ "expected distinct_count of 10 unique string values"
+ );
+ }
+
+ // Verifies that a missing distinct_count in one row group does not affect
the others.
+ #[test]
+ fn test_row_group_distinct_counts_absent() {
+ let parquet_schema =
+ Arc::new(parse_message_type("message schema { REQUIRED INT32 col;
}").unwrap());
+ let writer_properties = Arc::new(
+ WriterProperties::builder()
+ .set_statistics_enabled(EnabledStatistics::Chunk)
+ .build(),
+ );
+
+ let mut file_buffer: Vec<u8> = Vec::new();
+ let mut file_writer =
+ SerializedFileWriter::new(&mut file_buffer, parquet_schema,
writer_properties).unwrap();
+
+ // row group 0: distinct_count present
+ let mut row_group_writer = file_writer.next_row_group().unwrap();
+ let mut column_writer =
row_group_writer.next_column().unwrap().unwrap();
+ column_writer
+ .typed::<Int32Type>()
+ .write_batch_with_statistics(&[1, 2], None, None, Some(&1),
Some(&2), Some(2))
+ .unwrap();
+ column_writer.close().unwrap();
+ row_group_writer.close().unwrap();
+
+ // row group 1: distinct_count absent — should appear as null in output
+ let mut row_group_writer = file_writer.next_row_group().unwrap();
+ let mut column_writer =
row_group_writer.next_column().unwrap().unwrap();
+ column_writer
+ .typed::<Int32Type>()
+ .write_batch_with_statistics(&[3, 4], None, None, Some(&3),
Some(&4), None)
+ .unwrap();
+ column_writer.close().unwrap();
+ row_group_writer.close().unwrap();
+
+ // row group 2: distinct_count present — iteration must reach here
despite row group 1
+ let mut row_group_writer = file_writer.next_row_group().unwrap();
+ let mut column_writer =
row_group_writer.next_column().unwrap().unwrap();
+ column_writer
+ .typed::<Int32Type>()
+ .write_batch_with_statistics(&[5, 6, 7, 8, 9], None, None,
Some(&5), Some(&9), Some(5))
+ .unwrap();
+ column_writer.close().unwrap();
+ row_group_writer.close().unwrap();
+
+ file_writer.close().unwrap();
+
+ let parquet_bytes = Bytes::from(file_buffer);
+ let reader_builder =
ParquetRecordBatchReaderBuilder::try_new(parquet_bytes).unwrap();
+ let file_metadata = reader_builder.metadata().clone();
+ let arrow_schema = reader_builder.schema().clone();
+ let parquet_schema_descriptor =
file_metadata.file_metadata().schema_descr();
+
+ let statistics_converter =
+ StatisticsConverter::try_new("col", &arrow_schema,
parquet_schema_descriptor).unwrap();
+ let distinct_counts = statistics_converter
+ .row_group_distinct_counts(file_metadata.row_groups().iter())
+ .unwrap();
+
+ assert_eq!(
+ distinct_counts,
+ UInt64Array::from(vec![Some(2), None, Some(5)]),
+ "a missing distinct_count in one row group should not affect the
others"
+ );
+ }
}
