[
https://issues.apache.org/jira/browse/ARROW-4432?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16835422#comment-16835422
]
Joris Van den Bossche commented on ARROW-4432:
----------------------------------------------
With latest master, the only difference is the metadata (I suppose that before
the RangeIndex serialization change, the difference was also the index column):
{code}
In [3]: table = pa.Table.from_arrays([])
In [4]: df = table.to_pandas()
In [5]: df
Out[5]:
Empty DataFrame
Columns: []
Index: []
In [6]: table_ = pa.Table.from_pandas(df)
In [7]: table_
Out[7]:
pyarrow.Table
metadata
--------
{b'pandas': b'{"index_columns": [{"kind": "range", "name": null, "start": 0, "'
b'stop": 0, "step": 1}], "column_indexes": [{"name": null, "field_'
b'name": null, "pandas_type": "empty", "numpy_type": "object", "me'
b'tadata": null}], "columns": [], "creator": {"library": "pyarrow"'
b', "version": "0.13.1.dev130+gdd335952"}, "pandas_version": "0.24'
b'.2"}'}
In [8]: table_ = table_.replace_schema_metadata(None)
In [9]: table == table_
Out[9]: True
{code}
> [Python][Hypothesis] Empty table - pandas roundtrip produces inequal tables
> ---------------------------------------------------------------------------
>
> Key: ARROW-4432
> URL: https://issues.apache.org/jira/browse/ARROW-4432
> Project: Apache Arrow
> Issue Type: Bug
> Components: Python
> Reporter: Krisztian Szucs
> Priority: Major
> Labels: hypothesis
> Fix For: 0.14.0
>
>
> The following test case fails for empty tables:
> {code:python}
> import hypothesis as h
> import pyarrow.tests.strategies as past
> @h.given(past.all_tables)
> def test_pandas_roundtrip(table):
> df = table.to_pandas()
> table_ = pa.Table.from_pandas(df)
> assert table == table_
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
