Batch is essentially a wrapper around a dict. However, iterating over a batch leads to potentially unexpected behavior.

While list({...}) == list({..}.keys()), for batch list(Batch(...)) gives a sequence of Batch objects and list(Batch(...).keys()) gives a list of keys.

This collides with e.g. ruff's recommendation of not using in-dict-keys. I had to disable this rule in #921 because both for b in batch and for key in batch.keys() are currently used with their respective semantics throughout the code.

Fixing this would mean defining __iter__ in Batch to return keys and something like creating a new method in Batch called iter_entries that would replace the current iteration semantics. This would be a breaking change.

@Trinkle23897 what do you think? One can also live with the current status of the "inconsistency", keep ignoring the ruff recommendation, and just document the current behavior somewhere. I would also add dedicated unit tests in that case