Description
Describe the bug
As pointed out by @jorisvandenbossche on https://issues.apache.org/jira/browse/ARROW-16184 the embedded arrow schema should not be treated as authoritative for a parquet file, and instead should be used in a purely advisory capacity, for example, to aid in inferring list offsets, etc...
In particular
[It should be used as] a description of the original Arrow schema, and not for a description of what is in the Parquet file / for the Parquet schema.
This behavior appears to have been introduced by @carols10cents in apache/arrow#8354 and apache/arrow#8330, as a mechanism to round-trip types correctly. We should ensure the solution preserves this where applicable
To Reproduce
See #1459
Expected behavior
The embedded arrow schema should not be relied upon to be authoritative
Additional context
Working on as part of #1655
Activity
Separate parquet -> arrow conversion logic (apache#1655)
Separate parquet -> arrow conversion logic (apache#1655)
Separate parquet -> arrow conversion logic (apache#1655)
Separate parquet -> arrow conversion logic (apache#1655)
Fix Parquet Arrow Schema Inference (#1682)
1 remaining item