Expose validation errors #402

Activity

1. borchero commented on Sep 10, 2026

   

   Oliver Borchert (borchero)

   on Sep 10, 2026

   Member

   More actions

   If you want to persist the errors, isn't it an option to use Schema.filter instead of Schema.validate?
2. datenzauberai commented on Sep 11, 2026

   

   Thomas Bierhance (datenzauberai)

   on Sep 11, 2026

   Last edited by datenzauberai

   Author

   More actions

   I'm not sure what you mean by persisting the errors. Maybe my description was lacking...

   Right now the errors are not exposed in the public top-level API. Although they are not marked as private, it feels a bit hacky to import them from the exception module.

   Today:

   import dataframely as dy
   from dataframely.exc import SchemaError
   
   try:
       MySchema.validate(df)
   except SchemaError as exc:
       print(exc)

   Technically you can reference the exception module exc at runtime, but linters will complain that exc is an unknown attribute:

   import dataframely as dy
   
   try:
       MySchema.validate(df)
   # works but linter will complain
   except dy.exc.SchemaError as exc:
       print(exc)

   That could be fixed by exposing the exception module, but I think it would be even better to have the exceptions exposed at the top-level (e.g. Patito does that https://patito.readthedocs.io/en/latest/api/patito/DataFrame/validate.html):

   import dataframely as dy
   
   try:
       MySchema.validate(df)
   except dy.SchemaError as exc:
       print(exc)
3. Thomas Bierhance (datenzauberai)

   changed the title [-]Export validation errors[/-] [+]Expose validation errors[/+] on Sep 11, 2026
4. borchero commented on Sep 14, 2026

   

   Oliver Borchert (borchero)

   on Sep 14, 2026

   Member

   More actions

   Got it, sorry! 😅

   In the tests, we commonly do

   from dataframely.exc import ...

   but you could also just do something like

   import dataframely.exc as dyexc

   for now. I think it would be fair to expose these errors on the top-level though.

   ❤️React with ❤️1Reacted by Thomas Bierhance