[Python] ParquetWriter segfaults in example where passed schema and table schema do not match

[asfimport]

I get a segmentation fault with simply printing a basic table creation from arrays:

 

import pyarrow as pa
import pyarrow.parquet as pq

simple_fields = [
    pa.field('POS', pa.uint32()),
    pa.field('desc', pa.string())
]


simple_schema = pa.schema(simple_fields)

simple_from_array = [pa.Array([1]), pa.Array(['bla'])]
simple_table = pa.Table.from_arrays(simple_from_array, ['POS', 'desc'])
print(simple_table)

Reporter: Tiago Antao
Assignee: Wes McKinney / @wesm

PRs and other links:

• GitHub Pull Request #2339

_{Note: This issue was originally created as ARROW-2926. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
Use pa.array instead of pa.Array to create arrays. Using pa.Array directly in master branch or 0.10.0 raises an exception

[asfimport]

Tiago Antao:
After correcting for that, the segfault resurfaces later:

with pq.ParquetWriter('simple.parquet', simple_schema,
                      version='2.0',
                      compression='snappy', flavor='spark') as w:
    w.write_table(simple_table)

 

[asfimport]

Wes McKinney / @wesm:
Thanks. I reopened the issue and updated the title to reflect this issue, which I confirmed locally. Someone ought to look into while 0.10.0 is still not closed to see if the fix is easy

[asfimport]

Wes McKinney / @wesm:
Issue resolved by pull request 2339
#2339