[BEAM-11256] Dataframe writers send pickled methods to workers

• Type: Bug
• Status: Resolved
• Priority: P2
• Resolution: Unresolved
• Affects Version/s: 2.25.0
• Fix Version/s: 2.26.0
• Component/s: sdk-py-core
• Assignee: Brian Hulette
• Reporter: Brian Hulette
• Created: 12/Nov/20 20:16
• Updated: 13/Nov/20 17:25

Description

This can lead to errors when there's a mismatch between the pandas version used at construction time, and the one used in workers at execution time. For example to_csv constructed with 1.1.4, executed with 0.25.3:

  File "apache_beam/runners/common.py", line 1213, in apache_beam.runners.common.DoFnRunner.process
  File "apache_beam/runners/common.py", line 742, in apache_beam.runners.common.PerWindowInvoker.invoke_process
  File "apache_beam/runners/common.py", line 867, in apache_beam.runners.common.PerWindowInvoker._invoke_process_per_window
  File "/usr/local/lib/python3.7/site-packages/apache_beam/io/fileio.py", line 695, in process
    sink.write(record)
  File "/usr/local/google/home/bhulette/working_dir/beam/sdks/python/apache_beam/dataframe/io.py", line 266, in write_record_incremental
    self.empty = self.write_to(value[:0])
  File "/usr/local/lib/python3.7/site-packages/apache_beam/dataframe/io.py", line 260, in write_to
    self.writer(df, non_none_handle, *self.args, **self.kwargs)
  File "/usr/local/google/home/bhulette/.pyenv/versions/beam/lib/python3.7/site-packages/pandas/core/generic.py", line 3168, in to_csv
    .. versionchanged:: 0.24.0
TypeError: __init__() got an unexpected keyword argument 'errors'

Issue Links

• links to: GitHub Pull Request #13322
• links to: GitHub Pull Request #13327