[C++] Null values in partitioning field for FilenamePartitioning

[asfimport]

While using FilenamePartitioning, currently when we read a dataset, say from PyArrow, the partitioning field only has Null values.

 

The issue can be reproduced with the following code

 

table = pa.table([
            pa.array(range(20)), pa.array(np.random.randn(20)),
            pa.array(np.repeat(['a', 'b'], 10))],
            names=["f1", "f2", "part"]
        ) 

part = ds.partitioning(pa.schema([("part", pa.string())]), flavor="filename")

# test is the directory where partitions are written
ds.write_dataset(
            table, "test",
            format="parquet",  partitioning=part
        )

result = ds.dataset(
             "test",  format="parquet", partitioning=part,
         ).to_table()

This results in something this:

pyarrow.Table
f1: int64
f2: double
part: string
----
f1: [[0,1,2,3,4,5,6,7,8,9],[10,11,12,13,14,15,16,17,18,19]]
f2: [[-1.1753280347394899,-0.9640239222827617,0.7907247451009602,1.3667778347936321,0.005079832420686733,0.9024313772071855,-1.01618656608383,-1.1459911861999188,-0.7407261867306765,-0.012823499364722428],[0.29893685698088185,1.3907720928021299,-0.48826416913435605,-1.3436821154932153,-0.5492388164165941,-0.07093280675027104,0.009918818541272493,-1.05561750529359,-2.0209000426858927,-0.28081085330210676]]
part: [[null,null,null,null,null,null,null,null,null,null],[null,null,null,null,null,null,null,null,null,null]] 

 

 

 

Reporter: Sanjiban Sengupta / @sanjibansg
Assignee: Sanjiban Sengupta / @sanjibansg

Related issues:

• [C++] FileSystemDataset FilenamePartitioning error - fsspec filesystem (fixes)

PRs and other links:

• GitHub Pull Request #12977

_{Note: This issue was originally created as ARROW-16302. Please see the migration documentation for further details.}

[asfimport]

Weston Pace / @westonpace:
Issue resolved by pull request 12977
#12977