feat: Support Parquet writer options #1123
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
timsaucer
requested changes
Jun 13, 2025
Comment on lines
+697
to
+717
| data_pagesize_limit: int = 1024 * 1024, | ||
| write_batch_size: int = 1024, | ||
| writer_version: str = "1.0", | ||
| skip_arrow_metadata: bool = False, | ||
| compression: Optional[str] = "zstd(3)", | ||
| dictionary_enabled: Optional[bool] = True, | ||
| dictionary_page_size_limit: int = 1024 * 1024, | ||
| statistics_enabled: Optional[str] = "page", | ||
| max_row_group_size: int = 1024 * 1024, | ||
| created_by: str = "datafusion-python", | ||
| column_index_truncate_length: Optional[int] = 64, | ||
| statistics_truncate_length: Optional[int] = None, | ||
| data_page_row_count_limit: int = 20_000, | ||
| encoding: Optional[str] = None, | ||
| bloom_filter_on_write: bool = False, | ||
| bloom_filter_fpp: Optional[float] = None, | ||
| bloom_filter_ndv: Optional[int] = None, | ||
| allow_single_file_parallelism: bool = True, | ||
| maximum_parallel_row_group_writers: int = 1, | ||
| maximum_buffered_record_batches_per_stream: int = 2, | ||
| column_specific_options: Optional[dict[str, ParquetColumnOptions]] = None, |
There was a problem hiding this comment.
I think the full parquet writer options are going to be less commonly used for most users. I would suggest we want to have two different versions of the write_parquet signature, one that is the simple existing compression parameters and on that looks like
def write_parquet(self, path: str | pathlib.Path, options: ParquetWriterOptions)
And then we have a ParquetWriterOptions class that has this large initialize options.
Comment on lines
+1766
to
+1775
| for data_type in data_types: | ||
| match data_type: | ||
| case "int": | ||
| data["int"] = [1, 2, 3] | ||
| case "float": | ||
| data["float"] = [1.01, 2.02, 3.03] | ||
| case "str": | ||
| data["str"] = ["a", "b", "c"] | ||
| case "bool": | ||
| data["bool"] = [True, False, True] |
There was a problem hiding this comment.
This code failed for me because our minimum supported python is 3.9 and match was introduced in 3.10. 3.9 doesn't reach end of life until Oct 2025.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Which issue does this PR close?
N/A.
Rationale for this change
Supporting all Parquet writer options allows us more flexibility when creating data directly from
datafusion-python.For consistency, it supports all writer options defined by
ParquetOptionsindatafusion, using the same defaults: https://github.com/apache/datafusion/blob/555fc2e24dd669e44ac23a9a1d8406f4ac58a9ed/datafusion/common/src/config.rs#L423.What changes are included in this PR?
write_parquetwith all writer options, including column-specific options.pyarrowdoes not expose page-level information, some options could not be directly tested, like enabling bloom-filters (an external tool confirmed that this option works). For this specific case, there is a test that compares the file sizes.)Are there any user-facing changes?
The main difference relates to the existing
compressionfield, which now uses astrlikedatafusion, instead of a custom enum. The main advantage is that future algorithms will not require updating the Python-side code.Additionally, the default compression was changed from
zstd(4)tozstd(3), the same asdatafusion.