Joerg Schneider created ARROW-13048:
---------------------------------------
Summary: [Python] S3FileSystem fails moving filepaths containing =
or +
Key: ARROW-13048
URL: https://issues.apache.org/jira/browse/ARROW-13048
Project: Apache Arrow
Issue Type: Bug
Affects Versions: 4.0.1
Reporter: Joerg Schneider
Hi Arrow team,
we have the very common use-case of having partitioned parquet tables on S3,
written by Spark. These include equals (=) to denote the partition value per
folder.
In trying to use PyArrows S3FileSystem `move` function, it's not possible to
move these objects in the bucket underneath a path which contains `=`
somewhere:
{code:java}
OSError: When copying key
'table/date=202007/part-00000-e39069c2-0ea6-4a62-85ea-8011047cd4f4.c000.snappy.parquet'
in bucket 'bucket' to key
'table2/date=202007/part-00000-e39069c2-0ea6-4a62-85ea-8011047cd4f4.c000.snappy.parquet'
in bucket 'bucket': AWS Error [code 133]: The specified key does not
exist.{code}
It is also not possible to move, using preemptively URL-quoted paths, like
these:
{code:java}
OSError: When copying key
'table/date%3D202007/part-00000-e39069c2-0ea6-4a62-85ea-8011047cd4f4.c000.snappy.parquet'
in bucket 'bucket' to key
'table2/date%3D202007/part-00000-e39069c2-0ea6-4a62-85ea-8011047cd4f4.c000.snappy.parquet'
in bucket 'bucket': AWS Error [code 133]: The specified key does not
exist.{code}
The source object does definitely exist, it has in fact been returned by a
FileSelector from PyArrow itself and is just passed to move.
Is there any configuration option to be set, or special quoting to be used?
Thanks in advance.
Joerg
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
