Stand alone reproducer for Error "Not all children array length are the same!"

[alamb]

Related to

• Fix skip_records over-counting when partial record precedes num_rows page skip #9374
• Error "Not all children array length are the same!" when decoding rows spanning across page boundaries in parquet file when using RowSelection #9370

I wanted to get an "end to end" reproducer (aka with a parquet file) that triggers the error reported in from #9370 @jonded94

This is what I have so far (though it requires creating a column chunk with both V1 and V2 page headers)

I could not simplify the test more

Here is the file produced:
v1_v2_mixed.parquet.zip

Fascinatingly, it seems to work just fine without the row selection:

(venv) andrewlamb@Andrews-MacBook-Pro-3:~/Software/arrow-rs$ datafusion-cli
selecDataFusion CLI v52.1.0
> select * from '/tmp/v1_v2_mixed.parquet';
+---------------------+
| s                   |
+---------------------+
| {a: [10, 20], b: 1} |
| {a: [30, 40], b: 2} |
| {a: [50, 60], b: 3} |
| {a: [70, 80], b: 4} |
+---------------------+
4 row(s) fetched.
Elapsed 0.010 seconds.

And indeed https://parquet-viewer.xiangpeng.systems/ shows the different headers

[jonded94]

@alamb please have a look at #9399.

I incorrectly stated that this was about mixing v1 and v2 data pages. Apparently, there also can be other bugs, as your test in this PR discovers (?), but what my original issue of #9370 was about actually can be reproduced with v2 data pages only.

[alamb]

#9399 is better, let's continue there instead