fix: Raise proper error for mismatching parquet schema instead of panicking

[nameexhaustion]

This has been reported quite a bit lately, so got around to it

Fixes #13436, fixes #17067, fixes #17255, fixes #13438

Maybe also #17311, but I don't have a repro for that one

[etiennebacher]

Hi, just a quick note that in the first post you need to have one "fixes" per issue number to close them, e.g. "fixes #A, #B, #C" will only close #A :)

[codecov]

Codecov Report

Attention: Patch coverage is 81.36646% with 30 lines in your changes missing coverage. Please review.

Project coverage is 80.73%. Comparing base (41610e3) to head (ae742b6).
Report is 4 commits behind head on main.

Files	Patch %	Lines
crates/polars-core/src/schema.rs	65.62%	22 Missing ⚠️
crates/polars-arrow/src/datatypes/schema.rs	71.42%	4 Missing ⚠️
crates/polars-io/src/parquet/read/reader.rs	93.75%	2 Missing ⚠️
crates/polars-plan/src/plans/conversion/scans.rs	0.00%	1 Missing ⚠️
crates/polars-plan/src/plans/functions/count.rs	0.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #17321   +/-   ##
=======================================
  Coverage   80.72%   80.73%           
=======================================
  Files        1475     1475           
  Lines      193333   193383   +50     
  Branches     2760     2760           
=======================================
+ Hits       156077   156119   +42     
- Misses      36746    36754    +8     
  Partials      510      510           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[nameexhaustion]

Hi, just a quick note that in the first post you need to have one "fixes" per issue number to close them, e.g. "fixes #A, #B, #C" will only close #A :)

Very cool tip, thanks 🙂

[ion-elgreco]

@nameexhaustion @ritchie46, but this doesn't add support read parquets with mixed schema's though. This is something pyarrow supports

[stinodego]

Yeah we should be careful not to close issues when a panic has been turned into an error, if it should not throw an error in the first place. In this case, mismatching schemas indicate schema evolution and we may have to actually support this.