You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "mbutrovich (via GitHub)" <gi...@apache.org> on 2025/03/31 21:48:39 UTC
[PR] Test int96 Parquet file from Spark [arrow-rs]
mbutrovich opened a new pull request, #7367:
URL: https://github.com/apache/arrow-rs/pull/7367
# 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.
-->
I think we can close out #7220 after this test merges.
# 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.
-->
We would like to enforce testing on a challenging int96 file generated by Spark with its microsecond timestamps. It is challenging because it includes dates that cannot be represented in a nanosecond timestamp.
# 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.
-->
Add a test that is dependent on https://github.com/apache/parquet-testing/pull/73 merging first.
# Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be updated before approving the PR.
-->
No.
<!---
If there are any breaking changes to public APIs, please call them out.
-->
--
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] Test int96 Parquet file from Spark [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb merged PR #7367:
URL: https://github.com/apache/arrow-rs/pull/7367
--
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] Test int96 Parquet file from Spark [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #7367:
URL: https://github.com/apache/arrow-rs/pull/7367#discussion_r2028639282
##########
parquet/src/arrow/arrow_reader/mod.rs:
##########
@@ -1563,6 +1562,106 @@ mod tests {
})
}
+ #[test]
+ fn test_int96_from_spark_file_with_provided_schema() {
+ // int96_from_spark.parquet was written based on Spark's microsecond timestamps which trade
+ // range for resolution compared to a nanosecond timestamp. We must provide a schema with
+ // microsecond resolution for the Parquet reader to interpret these values correctly.
+ use arrow_schema::DataType::Timestamp;
+ let test_data = arrow::util::test_util::parquet_test_data();
+ let path = format!("{test_data}/int96_from_spark.parquet");
+ let file = File::open(path).unwrap();
+
+ let supplied_schema = Arc::new(Schema::new(vec![Field::new(
+ "a",
+ Timestamp(TimeUnit::Microsecond, None),
+ true,
+ )]));
+ let options = ArrowReaderOptions::new().with_schema(supplied_schema.clone());
+
+ let mut record_reader =
+ ParquetRecordBatchReaderBuilder::try_new_with_options(file, options)
+ .unwrap()
+ .build()
+ .unwrap();
+
+ let batch = record_reader.next().unwrap().unwrap();
+ assert_eq!(batch.num_columns(), 1);
+ let column = batch.column(0);
+ assert_eq!(column.data_type(), &Timestamp(TimeUnit::Microsecond, None));
+
+ let expected = Arc::new(Int64Array::from(vec![
+ Some(1704141296123456),
+ Some(1704070800000000),
+ Some(253402225200000000),
+ Some(1735599600000000),
+ None,
+ Some(9089380393200000000),
+ ]));
+
+ // arrow-rs relies on the chrono library to convert between timestamps and strings, so
+ // instead compare as Int64. The underlying type should be a PrimitiveArray of Int64
+ // anyway, so this should be a zero-copy non-modifying cast.
+
+ let binding = arrow_cast::cast(batch.column(0), &arrow_schema::DataType::Int64).unwrap();
+ let casted_timestamps = binding.as_primitive::<types::Int64Type>();
+
+ assert_eq!(casted_timestamps.len(), expected.len());
+
+ casted_timestamps
+ .iter()
+ .zip(expected.iter())
+ .for_each(|(lhs, rhs)| {
+ assert_eq!(lhs, rhs);
+ });
+ }
+
+ #[test]
+ fn test_int96_from_spark_file_without_provided_schema() {
+ // int96_from_spark.parquet was written based on Spark's microsecond timestamps which trade
+ // range for resolution compared to a nanosecond timestamp. Without a provided schema, some
+ // values when read as nanosecond resolution overflow and result in garbage values.
+ use arrow_schema::DataType::Timestamp;
+ let test_data = arrow::util::test_util::parquet_test_data();
+ let path = format!("{test_data}/int96_from_spark.parquet");
+ let file = File::open(path).unwrap();
+
+ let mut record_reader = ParquetRecordBatchReaderBuilder::try_new(file)
+ .unwrap()
+ .build()
+ .unwrap();
+
+ let batch = record_reader.next().unwrap().unwrap();
+ assert_eq!(batch.num_columns(), 1);
+ let column = batch.column(0);
+ assert_eq!(column.data_type(), &Timestamp(TimeUnit::Nanosecond, None));
+
+ let expected = Arc::new(Int64Array::from(vec![
+ Some(1704141296123456000), // Reads as nanosecond fine (note 3 extra 0s)
+ Some(1704070800000000000), // Reads as nanosecond fine (note 3 extra 0s)
+ Some(-4852191831933722624), // Cannot be represented with nanos timestamp (year 9999)
Review Comment:
nice
--
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] Test int96 Parquet file from Spark [arrow-rs]
Posted by "mbutrovich (via GitHub)" <gi...@apache.org>.
mbutrovich commented on code in PR #7367:
URL: https://github.com/apache/arrow-rs/pull/7367#discussion_r2021824338
##########
parquet/src/arrow/arrow_reader/mod.rs:
##########
@@ -978,6 +972,11 @@ mod tests {
ArrowError, DataType as ArrowDataType, Field, Fields, Schema, SchemaRef, TimeUnit,
};
use arrow_select::concat::concat_batches;
+ use bytes::Bytes;
Review Comment:
`cargo fmt` decided to move these for me.
--
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] Test int96 Parquet file from Spark [arrow-rs]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #7367:
URL: https://github.com/apache/arrow-rs/pull/7367#discussion_r2027707092
##########
parquet/src/arrow/arrow_reader/mod.rs:
##########
@@ -1563,6 +1562,57 @@ mod tests {
})
}
+ #[test]
Review Comment:
It might also be worth a test showing what happens when a schema is not supplied for this file (that the data is read out as nanosecond precision)
--
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] Test int96 Parquet file from Spark [arrow-rs]
Posted by "mbutrovich (via GitHub)" <gi...@apache.org>.
mbutrovich commented on PR #7367:
URL: https://github.com/apache/arrow-rs/pull/7367#issuecomment-2776092255
https://github.com/apache/parquet-testing/pull/73 merged so I think this is ready for review. I am now realizing that I probably need to bump the commit in the submodule, though.
--
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