Add bulk_update() support for unique fields instead of only primary key

[Ebram Shehata]

Currently, bulk_update() function from django.db.models.query only performs the update using the primary key field..

I think we can generalize it more to use a unique field..
Example: MyModel.objects.bulk_update(objs, fields=["name"], unique_field="national_id")
This will use MyModel.national_id to identify objects since national_id is unique.
I wrote the code for it and I thought to contribute it to Django..

I also have a question about that function, we can't use it to update primary keys.. I'm wondering why ? I want to also add that support to it.

Note: I'm not really sure what to do to reserve the implementation for me. I want to be the one to implement it, actually, I already wrote the code.

Here's a PR I created: ​https://github.com/django/django/pull/15764

[Carlton Gibson]

Hi Ebram. Welcome. Thanks!

Can I ask you to send this to the DevelopersMailingList to get more eyes on it, and see if it's a change worth adding?
(I'll close the ticket now, but we'll reopen if there's a consensus to add the feature.)

When you post, can you add a bit more as to why you think this is useful? I can't immediately see why it matters, given that I have the objects in hand 🤔

I also have a question about that function, we can't use it to update primary keys.. I'm wondering why ? I want to also add that support to it.

The mailing list is a better place to discuss this as well.

...what to do to reserve the implementation for me...

If there's agreement to add, you can assign the ticket to yourself. (There's not often contention in these areas 🙂)

Ref the PR: I appreciate this is just a first proof-of-concept but, all changes need test coverage and docs.

[Ebram Shehata]

Replying to Carlton Gibson:
Hi Carlton, thank you for replying!
Here's a discussion I started ​https://groups.google.com/g/django-developers/c/SKvpdIN-NE0
I mentioned the use case in which I needed such feature and demonstrated why this could be a useful addition. Please have a look. Thank you.

[Carlton Gibson]

Hi Ebram.

Yes, I'm following the thread there.

Jerch's reply seems positive: "Overall this sounds like a valuable API addition to bulk_update…" but he raises the complexity of the MTI case, and also whether it's needed:

NB: Btw fetching proper pks upfront from some other unique field is
typically very cheap compared to bulk_update runtime itself, given you
have indexed those columns.

Short of further input, if you're keen to keep working on this, I'd suggest adding some tests for the MTI case, and making that work. From there — assuming it's not too complex -- you have a decent case for a quality of life improvement.