pyreadstat.read_sav won't work on databricks

[marcmaxson]

Please read the README, particularly the known limitations section!

When I try this:
df, meta = pyreadstat.read_sav("/Volumes/sandbox/schema/folder/myfile.sav")
I get "Unknown Error".

This happens for versions 1.3.1, 1.3.0, 1.2.9, and 1.2.8 but works fine in 1.2.7.

To Reproduce
databricks (serverless) environment version 4

File example
Tried with many SAV files; any file will reproduce.

Expected behavior
File loads

Setup Information:
How did you install pyreadstat?

• tested with !pip install pyreadstat==1.3.1 in notebook or using the databricks environment GUI (version 4).

Platform: linux
Python Version: Python 3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0] on linux

[ofajardo]

Very strange. If I understand correctly those files would open fine in a local machine but fail in databricks?

If so, unfortunately I cannot reproduce as I don't have databricks.

The error suggests the problem is coming from Readstat, and indeed in version 1.2.8 there was an update on the Readstat sources, so it may be a problem over there, but cannot tell without being able to reproduce. You may want to report also in Readstat in case somebody over there could reproduce.

BTW, have you tried to copy one of those files from /Volumes to your home directory? Maybe it has something to do with the way the Volume is mounted.

[marcmaxson]

I don't think it is related to /Volumes or file locations - as I tried it both in my workspace and in unity catalog. We're freezing at version 1.2.7 (and have been for a while now) until this is resolved.

[ofajardo]

We would need help with somebody having access to databricks in order to debug or report and follow-up on Readstat.

[marcmaxson]

we're happy to test this in databricks and work with you on a fix. Let me know how we can help. Still frozen at 1.2.7 in all our notebooks.

[ofajardo]

thanks @marcmaxson let's see how we can tackle this. I guess the approach should be making a diff of the code between the latest version working and the the first version not working, to see what lines have changed, and then try to identify from which line this could be coming from. Probably the source of the issue is something in the src folder (Readstat source files). Since you get that error "Unknown Error", it has to be something that raises that one.
So the first task would be making the diff to identify candidate lines (also checking for the error), and then probably adding printing statements to the candidate lines to see where the issue is happening.
Do you think you could do that or you need more guidance?
After this process, the output will be writing an issue to Readstat to see if they can fix it, and probably they will also need help for testing the solution.