It sounds like it should be possible, could you perhaps open a JIRA with a
more complete example including schema?
Thanks,
Micah
On Thu, Jun 25, 2020 at 4:10 PM Rafael Ladeira <rwlade...@gmail.com> wrote:
> Hi,
>
> Is it possible to read just selected columns from a dataframe with
> hierarchical levels in the columns, passing tuple to 'columns' argument?
>
> Example:
>
> pd.read_parquet('file.parquet', engine='pyarrow', columns=[('level_0_key',
> 'level_1_key')])
>
> Trying to accomplish this in version '0.17.1' of pyarrow package , it
> raises the following error:
>
> "
> AttributeError Traceback (most recent call last)
> in
> ----> 1 pd.read_parquet('test.parquet', columns=[('IBOV Index',
> 'PX_LAST')])
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pandas\io\parquet.py
> in read_parquet(path, engine, columns, **kwargs)
> 308
> 309 impl = get_engine(engine)
> --> 310 return impl.read(path, columns=columns, **kwargs)
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pandas\io\parquet.py
> in read(self, path, columns, **kwargs)
> 122
> 123 kwargs["use_pandas_metadata"] = True
> --> 124 result = self.api.parquet.read_table(
> 125 path, columns=columns, **kwargs
> 126 ).to_pandas()
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pyarrow\array.pxi
> in pyarrow.lib._PandasConvertible.to_pandas()
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pyarrow\table.pxi
> in pyarrow.lib.Table._to_pandas()
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pyarrow\pandas_compat.py
> in table_to_blockmanager(options, table, categories, ignore_metadata,
> types_mapper)
> 763
> 764 _check_data_column_metadata_consistency(all_columns)
> --> 765 columns = _deserialize_column_index(table, all_columns,
> column_indexes)
> 766 blocks = _table_to_blocks(options, table, categories,
> ext_columns_dtypes)
> 767
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pyarrow\pandas_compat.py
> in _deserialize_column_index(block_table, all_columns, column_indexes)
> 875 # if we're reconstructing the index
> 876 if len(column_indexes) > 0:
> --> 877 columns = _reconstruct_columns_from_metadata(columns,
> column_indexes)
> 878
> 879 # ARROW-1751: flatten a single level column MultiIndex for
> pandas 0.21.0
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pyarrow\pandas_compat.py
> in _reconstruct_columns_from_metadata(columns, column_indexes)
> 1068
> 1069 # Convert each level to the dtype provided in the metadata
> -> 1070 levels_dtypes = [
> 1071 (level, col_index.get('pandas_type', str(level.dtype)))
> 1072 for level, col_index in zip_longest(
>
>
> ~\AppData\Local\Continuum\anaconda3\envs\vam_trading\lib\site-packages\pyarrow\pandas_compat.py
> in (.0)
> 1069 # Convert each level to the dtype provided in the metadata
> 1070 levels_dtypes = [
> -> 1071 (level, col_index.get('pandas_type', str(level.dtype)))
> 1072 for level, col_index in zip_longest(
> 1073 levels, column_indexes, fillvalue={}
>
> AttributeError: 'dict' object has no attribute 'dtype'
> "
>
> Thanks in advance,
>
> Rafael.
>
