Allow using QuerySet.in_bulk() with a composed unique field

[Jean-Daniel]

For instance, I have a model 'Data' with an unique_together constraint on (foo, bar) fields.

I would like to use in_bulk() on a query like:

  Data.objects.filter(foo='a').in_bulk(<list of bar values>, field_name='bar').

Obviously, the fact that bar is not unique should not be an issue to run the query, but Django performs a check to make sure the queried field is unique and raise an error if not.

As it may not be simple to check for all cases where field_name would result in a unique constraint, maybe it should be possible to disable this check (using an optional parameter).

[Tim Graham]

In your example, if bar isn't unique, then values may be overwritten in the in_bulk() return value unless the keys become ('foo', 'bar') (but how would in_bulk() know that?). I think this proposal would complicate in_bulk() a bit too much.

[Jean-Daniel]

Replying to Tim Graham:

In your example, if bar isn't unique, then values may be overwritten in the in_bulk() return value unless the keys become ('foo', 'bar') (but how would in_bulk() know that?). I think this proposal would complicate in_bulk() a bit too much.

In my example, bar is guarantee to be unique, as the query already contains an equality filter on foo. And yes, I understand it would make the function too complex to try to handle all possibles cases, that's why I suggested to let the caller choose to disable the check, ie something like this:

in_bulk(<list of bar>, field_name='bar', disable_unique_check=True)

[Simon Charette]

Given in_bulk can be easily replaced with a dict literal I don't think it's worth complexifying it more for the rare cases where this is useful either.

bulk_data = {
    data.bar: data for data Data.objects.filter(foo='a', bar__in=bars)
}