[
https://issues.apache.org/jira/browse/ARROW-4311?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16800850#comment-16800850
]
Antoine Pitrou commented on ARROW-4311:
---------------------------------------
I think this works now:
{code:python}
>>> pq.write_table(table, "t/level1#level2.parquet")
>>>
>>> !ls t
>>>
level1#level2.parquet
>>> pq.read_pandas("t/level1#level2.parquet")
>>>
pyarrow.Table
ints: int64
{code}
> [Python] Regression on pq.ParquetWriter incorrectly handling source string
> --------------------------------------------------------------------------
>
> Key: ARROW-4311
> URL: https://issues.apache.org/jira/browse/ARROW-4311
> Project: Apache Arrow
> Issue Type: Bug
> Components: Python
> Affects Versions: 0.12.0
> Reporter: Francisco Sanchez
> Priority: Major
> Fix For: 0.14.0
>
>
> In the latest changes to filesystem.py some new functions have been added to
> check the source string when calling pq.ParquetWriter. With the current
> implementation some assumptions are done about the format of the string which
> means that if the string is provided following some of these patterns it will
> be automatically split/formatted and changed to something else.
> To give you a specific example, if I provide a string like
> {{directory/level1#level2.parquet}} it will be written to disk as
> {{directory/level1}}. The behaviour has changed on 0.12.0 from 0.11.1 and
> nothing is stated in the documentation.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
