DataFrameModel vs DataFrameSchema

[lior5654]

Suppose one develops a library full of dataframe formats, and seeks correctness. How should one choose between DataFrameModel and DataFrameSchema?

In my case, being able as a user to alter the requirement&nullability of certain columns of certain dataframes outside the library is critical, and I find that doing so on DataFrameModels is less trivial.

What are the advantages & disadvantages of them over each other?

Note: If this lies in some FAQ section I missed, please refer me to it

[lior5654]

Was suprised I didn't find a similar discussion, excuse me if this was discussed before

[cosmicBboy]

They're logically the same, just a matter of taste for the most part. If you want dataclass-style schemas, use DataFrameModel, if you want object-based schemas, use DataFrameSchema. Under the hood DataFrameModel just converts to a DataFrameSchema.

The main benefit of DataFrameSchema is you can apply dynamic schema transformations: https://pandera.readthedocs.io/en/stable/dataframe_schemas.html#dataframeschema-transformations