stackoverflow when doing diagonal concat of 17k CSVs #13915
Labels
A-streaming
Related to the streaming engine
accepted
Ready for implementation
bug
Something isn't working
P-medium
Priority: medium
python
Related to Python Polars
Checks
Reproducible example
(Change
source_dir
anddest_path
accordingly.)Log output
Issue description
My context:
scan_csv
diagonal
concat. (I'm not sure if there's some way to tellscan_csv
what columns I expect, such that it returns them present but Null.)Note that when using a
vertical
concat instead ofdiagonal
, or usingscan_csv(csv_paths)
, I instead get:If I halve the number of files, I don't get this error.
Expected behavior
Polars should be smart enough to only open a reasonable number of files at a time. (Closing and re-openning if required.)
Even if this is not possible, I expect a clear error not a seg fault. (I thought it was theoretically impossible to get a seg fault with Rust.)
Installed versions
I'm on Ubuntu.
The text was updated successfully, but these errors were encountered: