Filter on old_record during deletes
#76
Conversation
| or ( | ||
| action = 'DELETE' | ||
| and realtime.is_visible_through_filters(old_columns, subs.filters) | ||
| ) |
There was a problem hiding this comment.
this is the change
| walrus; | ||
| rec | is_rls_enabled | subscription_ids | errors | ||
| ----------------------------------------------------+----------------+----------------------------------------+-------- | ||
| { +| f | {} | {} |
There was a problem hiding this comment.
no subscription_ids present because there is a filter on body which is not part of the replica identity / not part of the primary key
| }, +| | | | ||
| "commit_timestamp": "2000-01-01T08:01:01.000Z"+| | | | ||
| } | | | | ||
| { +| f | {5211e8ec-8c25-5c7f-9b03-6ff1eac0159e} | {} |
There was a problem hiding this comment.
subscription_ids exists because the filter is on a primary key column
| walrus; | ||
| rec | is_rls_enabled | subscription_ids | errors | ||
| ----------------------------------------------------+----------------+----------------------------------------+-------- | ||
| { +| f | {f4539ebe-c779-5788-bbc1-2421ffaa8954} | {} |
There was a problem hiding this comment.
after setting replica identity full the filter on body = 'bbb' works, so we get subscription_id
| }, +| | | | ||
| "commit_timestamp": "2000-01-01T08:01:01.000Z"+| | | | ||
| } | | | | ||
| { +| f | {5211e8ec-8c25-5c7f-9b03-6ff1eac0159e} | {} |
| ], +| | | | ||
| "commit_timestamp": "2000-01-01T08:01:01.000Z"+| | | | ||
| } | | | | ||
| { +| f | {f4539ebe-c779-5788-bbc1-2421ffaa8954} | {} |
There was a problem hiding this comment.
note that this part of the output changed
| ], +| | | | ||
| "commit_timestamp": "2000-01-01T08:01:01.000Z"+| | | | ||
| } | | | | ||
| { +| f | {f4539ebe-c779-5788-bbc1-2421ffaa8954} | {} |
There was a problem hiding this comment.
note that this part of the output changed
|
@olirice what happens when RLS is enabled on table and there's a delete filter? If client passes RLS they will get the change otherwise they will not. Is that correct? |
this change does not impact RLS, its just about filtering Unfortunately, it is still the case that we can't test RLS for delete |
|
@olirice can the docs be updated? https://supabase.com/docs/guides/realtime/postgres-changes#delete-events-are-not-filterable |
What kind of change does this PR introduce?
Adds support for filters on delete events
What is the current behavior?
Please link any relevant issues here.
resolves #53