Metadata is missing

[mrcaseb]

We attach various metadata to the data released to nflverse-data. nflreadr uses those metadata in a print method to inform about data updates and related package versions.

Can we work with the metadata in Python?

[tanho63]

The R S3 attribute metadata no, since that’s serialized with an R function serialize() that would be annoying to unserialize in python. However, we can manually save key-value metadata that isn’t an S3 object attribute and make that available:

https://github.com/nflverse/nflverse-data/blob/2d1ead41b8375d2bbb5a5ab9dc51e493524bf068/R/upload.R#L82-L85

which is then able to be read without unserialization by polars’s read_parquet_metadata

I’ve gone ahead and applied the change to nflversedata::nflverse_save(), and we will need to re-save all of the parquet files with the new function if we want to have it apply consistently (which we probably do)

[mrcaseb]

This sounds good to me. Before we update all parquet files, we should make sure that we have a working print method (I don't know the correct naming in python).

We can use the 2025 pbp data as it should already have been updated with the new approach