[
https://issues.apache.org/jira/browse/BEAM-11282?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17266832#comment-17266832
]
Beam JIRA Bot commented on BEAM-11282:
--------------------------------------
This issue is P2 but has been unassigned without any comment for 60 days so it
has been labeled "stale-P2". If this issue is still affecting you, we care!
Please comment and remove the label. Otherwise, in 14 days the issue will be
moved to P3.
Please see https://beam.apache.org/contribute/jira-priorities/ for a detailed
explanation of what these priorities mean.
> Cannot set compression level when writing compressed files
> ----------------------------------------------------------
>
> Key: BEAM-11282
> URL: https://issues.apache.org/jira/browse/BEAM-11282
> Project: Beam
> Issue Type: Improvement
> Components: sdk-py-core
> Affects Versions: 2.25.0
> Reporter: Jack Whelpton
> Priority: P2
> Labels: stale-P2
>
> CompressedFile._initialize_compressor hardcodes the compression level used
> when writing:
>
> self._compressor = zlib.compressobj(
> zlib.Z_DEFAULT_COMPRESSION, zlib.DEFLATED, self._gzip_mask)
>
> It would be good to be able to control this, as I have a large set of GZIP
> compressed files that are creating output 10x larger then the input size when
> writing the same data back.
>
> I've tried various monkeypatching approaches: these seem to work with the
> local runner, but failed when using DataflowRunner. For example:
>
> class WriteData(beam.PTransform):
> def __init__(self, dst):
> import zlib
> self._dst = dst
> def _initialize_compressor(self):
> self._compressor = zlib.compressobj(
> zlib.Z_BEST_COMPRESSION, zlib.DEFLATED, self._gzip_mask
> )
> CompressedFile._initialize_compressor = _initialize_compressor
> def expand(self, p):
> return p | WriteToText(
> file_path_prefix=self._dst,
> file_name_suffix=".tsv.gz",
> compression_type="gzip",
> )
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
