[jira] [Created] (ARROW-10146) [Python] Parquet metadata to_dict raises attribute error

["Florian Jetter (Jira)" <ji...@apache.org>]

Florian Jetter created ARROW-10146:
--------------------------------------

             Summary: [Python] Parquet metadata to_dict raises attribute error
                 Key: ARROW-10146
                 URL: https://issues.apache.org/jira/browse/ARROW-10146
             Project: Apache Arrow
          Issue Type: Bug
          Components: Python
            Reporter: Florian Jetter


h2. Description

When accessing rowgroup metadata and trying to convert it to a dict using the method {{to_dict}} I sometimes receive an Attribute error.

This can be consistently produced with an empty dataframe (see example below) but I have also seen it already for non empty dataframes. I couldn't track down what makes the non-empty Attribute errors special, therefore the example below.
h2. Expected behaviour

I would expect the to_dict to always, consistently return a dictionary with the appropriate metadata and statistics irregardless of the file content.
h2. Minimal Example

 
{code:python}
import pandas as pd
import pyarrow as pa
import pyarrow.parquet as pq

df = pd.DataFrame({"col": [1]}).head(0)
table = pa.Table.from_pandas(df)
buf = pa.BufferOutputStream()
pq.write_table(table, buf)
reader = pa.BufferReader(buf.getvalue())
parquet_file = pq.ParquetFile(reader)
# Raises Attribute Error
parquet_file.metadata.to_dict()
{code}
h3. Traceback
{code:java}
~/miniconda3/envs/kartothek-dev/lib/python3.7/site-packages/pyarrow/_parquet.pyx in pyarrow._parquet.FileMetaData.to_dict()

~/miniconda3/envs/kartothek-dev/lib/python3.7/site-packages/pyarrow/_parquet.pyx in pyarrow._parquet.RowGroupMetaData.to_dict()

~/miniconda3/envs/kartothek-dev/lib/python3.7/site-packages/pyarrow/_parquet.pyx in pyarrow._parquet.ColumnChunkMetaData.to_dict()

AttributeError: 'NoneType' object has no attribute 'to_dict'
{code}
h3. Versions
{code:java}
In [28]: pa.__version__
Out[28]: '1.0.1'

In [29]: pd.__version__
Out[29]: '1.0.5'
{code}



--
This message was sent by Atlassian Jira
(v8.3.4#803005)