Opened 4 years ago

Closed 4 years ago

#32281 closed New feature (wontfix)

Prefetech Object - to_attr has ambiguous behavior when using a related lookup

Reported by: Tal Perry Owned by: nobody
Component: Database layer (models, ORM) Version: 3.1
Severity: Normal Keywords:
Cc: Triage Stage: Unreviewed
Has patch: no Needs documentation: no
Needs tests: no Patch needs improvement: no
Easy pickings: no UI/UX: no

Description (last modified by Mariusz Felisiak)

Code Example

from django.db.models import Prefetch

p = Prefetch("a_set", to_attr="my_a")
assert p.prefetch_to == "my_a"

p = Prefetch("b__a_set", to_attr="my_a")
assert p.prefetch_to == "my_a", p.prefetch_to

That last assertion fails because p.prefetch_to == "b__my_a"

Description and use case
We have a models X, Y and Z
Z has a foreign key to Y and Y has a foreign key to Z
We'd like to fetch X with all of it's Z's using prefetch related and to be able to set the attribute to "z"

But as above, this doesn't seem to work.

The relevant Django code seems to ignore this use case (or discourage it?) and instead puts to "z" attribute on each "Y"

            self.prefetch_to = LOOKUP_SEP.join(lookup.split(LOOKUP_SEP)[:-1] + [to_attr])


I assume "fixing this" is a breaking change, but could we do an explicit parameter for prefetch_to ?At the moment we're setting it manually after creating the prefetch object

Change History (3)

comment:1 by Mariusz Felisiak, 4 years ago

Component: UncategorizedDatabase layer (models, ORM)
Description: modified (diff)
Resolution: wontfix
Status: newclosed
Type: UncategorizedNew feature

Thanks for this ticket, however IMO this is an expected behavior. In the second example a_set is a field lookup on b so we can change its name to my_a (in b) but we cannot assign it directly to the main object, it will be misleading.

comment:2 by Tal Perry, 4 years ago

Resolution: wontfix
Status: closednew

I understand that we can't assign to the main object, but am reopening because I think significant confusion remains.

In our use case, we're assigning a filtered queryset to the prefetch object .

annotations = Prefetch(
        "tokens__tagged_tokens", queryset=tagged_token_qs, to_attr="annotations",
    )

In the above example it's not clear whether the resulting prefetch will bring all tokens, or only those tokens who have a tagged_token in the supplied queryset.

Perhaps the solution is as simple as making this explicit in the docs. Happy to contribute if that's the right way and someone can shed light on the behavior and any edge cases.

in reply to:  2 comment:3 by Mariusz Felisiak, 4 years ago

Resolution: wontfix
Status: newclosed

In our use case, we're assigning a filtered queryset to the prefetch object .

annotations = Prefetch(
        "tokens__tagged_tokens", queryset=tagged_token_qs, to_attr="annotations",
    )

In the above example it's not clear whether the resulting prefetch will bring all tokens, or only those tokens who have a tagged_token in the supplied queryset.

Perhaps the solution is as simple as making this explicit in the docs. Happy to contribute if that's the right way and someone can shed light on the behavior and any edge cases.

I'm not sure how this question is related with the ticket. It's explicitly stated in docs:

The queryset argument supplies a base QuerySet for the given lookup.

Please follow triaging guidelines with regards to wontfix tickets.

Note: See TracTickets for help on using tickets.
Back to Top