[
https://issues.apache.org/jira/browse/ARROW-1400?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16137747#comment-16137747
]
Wes McKinney commented on ARROW-1400:
-------------------------------------
This would be a very useful feature. The simplest way to do this in the short
term will be to generate the partition scheme from a pandas.DataFrame using
pandas operations to split the object into pieces. We should add a function in
pyarrow.parquet which enables data to be "inserted" into a directory containing
a standard Hive-like partition schema. So you could do something like (just
spitballing here)
{code}
pq.write_table_to_dataset(dataset_path, partition_keys=keys, **options)
{code}
Here dataset_path is a directory, and this will write a new Parquet file in the
appropriate location in the subdirectory structure if partition_keys is not
None.
A patch would be welcome. I will mark this issue for 0.7.0
> [Python] Ability to create partitions when writing to Parquet
> -------------------------------------------------------------
>
> Key: ARROW-1400
> URL: https://issues.apache.org/jira/browse/ARROW-1400
> Project: Apache Arrow
> Issue Type: New Feature
> Components: Python
> Affects Versions: 0.6.0
> Environment: Mac OS Sierra 10.12.6
> Reporter: Safyre Anderson
> Priority: Minor
> Fix For: 0.7.0
>
>
> I'm fairly new to pyarrow so I apologize if this is already a feature, but I
> couldn't find a solution in the documentation nor an existing issue.
> Basically I'm trying to export pandas dataframes to .parquet files with
> partitions. I can see that pyarrow.parquet has a way of reading .parquet
> files with partitions, but there's no indication that it can write with
> partitions. E.g., it would be nice if there was a parameter in
> pyarrow.Table.write_table() that took a list of columns to partition the
> table similar to the pyspark implementation: spark.write.parquet's
> "partitionBy" parameter.
> Referenced links:
> https://arrow.apache.org/docs/python/parquet.html
> https://arrow.apache.org/docs/python/parquet.html?highlight=pyarrow%20parquet%20partition
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
