[FEA] Add File Statistic when writing the ORC file

[wbo4958]

Even with #10041 adding Statistic in RowIndex, CUDF still does not support writing the whole File Statistics into the ORC file.

Why do we need this FEA?

Spark added a new FEA about aggregate pushdown for ORC/Parquet file format as of Spark 3.3.0, which pushes the aggregate to DataSource to calculate the aggregation.

Typically, for ORC, it will read File Statistics from the ORC file and get the column stats and finally do the aggregate(max/min/count) without reading any real data of the ORC file. which can remarkably improve the performance.

But if there is no File Statistic in the ORC file, Spark will throw an exception.

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[amahussein]

There is a standing bug #5826 that points to missing statistics for chunk mode.

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[vuule]

@wbo4958 I think the issue should be fixed by #10694
Is anything else pending here?