API sugar for common depend_on_row patterns

[lenaqr]

The idea would be to reduce the amount of lambdas you have to write. Currently there already exists

depend_on_row(Model, 'selector')
->
get_or_create_token(('selector',))
depend_on_row(Model, lambda instance: {'selector': instance})

A useful one would be:

depend_on_row(Model, {'arg1': 'attr1', 'arg2': 'attr2'})
->
get_or_create_token(('arg1', 'arg2'))
depend_on_row(Model, lambda instance: {'arg1': instance.attr1, 'arg2': instance.attr2})

Another one that might be slightly trickier is depend_on_cache:

depend_on_cache(cache, {'arg1': 'arg2', 'arg3': 'arg4'})
->
get_or_create_token(('arg1', 'arg3'))
depend_on_cache(cache, lambda arg2=wildcard, arg4=wildcard, **kwargs: {'arg1': arg2, 'arg3': arg4})

More generally, could also support using lambdas as per-key selectors:

depend_on_row(Model, {'arg1': f1, 'arg2': f2})
->
get_or_create_token(('arg1', 'arg2'))
depend_on_row(Model, lambda instance: {'arg1': f1(instance), 'arg2': f2(instance)})

This would be useful in the case where the per-key selectors are simple functions, like operator.methodcaller('something'), and also has the benefit of being able to create the tokens automatically ahead of time.

[benjaminjkraft]

Another nice one would be to have something for "expire this cache for all arguments if the model matches this predicate". This could probably replace a lot of unnecessarily broad uses of depend_on_model -- e.g. a lot of the uses of depend_on_model(Tag) in the ESP website actually only depend on tags with a specific name.

[lenaqr]

That actually exists, it's the filter parameter to depend_on_row and friends.

[lenaqr]

(But it could also be useful to have shortcuts for defining those without lambdas too -- maybe passing in a QuerySet-like object.)

[benjaminjkraft]

Ah, TIL.