913539 - Rocketfuel API returns spurious uniqueness errors on PATCH

Status: RESOLVED DUPLICATE of bug 914874

(Marketplace Graveyard :: API, defect)

Description

[Matt Basta [:basta]]

PATCH https://marketplace-dev.allizom.org/api/v1/rocketfuel/collections/225/

  is_public=true

{
"collection_uniqueness":"You can not have more than one Featured Apps/Operator Shelf collection for the same category/carrier/region combination."
}

Note that the collection *is* unique (at least so far as I can see).

Comment 1

[Mathieu Pillard [:mat]]

I'll take it back on monday if nobody jumped on it in the meantime.

Comment 2

[Chuck Harmston [:chuck]]

It is a dupe with

https://marketplace-dev.allizom.org/api/v1/rocketfuel/collections/165/

Both have null carriers and categories and the `us` region.

Comment 3

[Matt Basta [:basta]]

REOPENing because the error doesn't make any sense for updating the is_public field. The error should happen 1) at POST, though it didn't or 2) on update of region/carrier/category. We shouldn't prevent you from updating other fields because the collection is in a weird state.

Comment 4

[Chuck Harmston [:chuck]]

This happens on CollectionSerializer().full_clean(), which seems like the appropriate place for this validation to take place.

Can you devise a test case where you're able to show that it doesn't happen on POST? There's a chance this was just bad data that got in before the validation was in place. If there's a bug, that's where it is.

Also, go home. You're on PTO.

Comment 5

[Matt Basta [:basta]]

I'm not on PTO, I'm just OOTO. :P

I'll build a test case after I get bug 913003 fixed.

Comment 6

[Mathieu Pillard [:mat]]

Is it possible the 2 collections were created before we started enforcing this ? (about to go home, don't want to fire up VPN to check) - we can't express properly this constraint using just SQL, so it doesn't apply for the old collections

Comment 7

[Matt Basta [:basta]]

It's possible.

It might be a smart move for us--since we've made so many changes--to wipe all of the collections on -dev just to be sure that we're not chasing ghosts. Who do we need to talk to about that?

Comment 8

[Matt Basta [:basta]]

Hmmm, yeah. Now that bug 913003 is fixed, I can't reproduce [the bug] on any new collections. Let's wipe the DB and call it a day, yeah?

Comment 9

[Mathieu Pillard [:mat]]

See discussion in bug 914874 this is essentially the same thing.