fix: remove foreign key, make trigger handle deletes #485
Conversation
Askir
force-pushed
the
jascha/ai-459-possible-bug-in-event-trigger
branch
2 times, most recently
from
2f3cf22 to
b4ccdc1
Compare
Askir
force-pushed
the
jascha/ai-459-possible-bug-in-event-trigger
branch
from
3e3b196 to
1293d44
Compare
| update ai.vectorizer | ||
| set config = jsonb_set(config, '{version}', format('"%s"', _new_version)::jsonb) | ||
| where id = _vec.id; |
There was a problem hiding this comment.
We need to make sure we don't have any other config upgrades in this release or this may cause the other(s) to fail.
There was a problem hiding this comment.
Only if they rely on the version being pre 0.8.1 right? Or is there another concern, I don't understand?
There was a problem hiding this comment.
i just mean if we have a second upgrade like this in the same release that this one goes out in, then we may have an issue because the other upgrade would also be looking for vectorizers with version < 0.8.1
| @@ -126,7 +126,6 @@ begin | |||
| , chunk text not null | |||
| , embedding @extschema:[email protected](%L) storage main not null | |||
| , unique (%s, chunk_seq) | |||
| , foreign key (%s) references %I.%I (%s) on delete cascade | |||
There was a problem hiding this comment.
source_schema and source_table parameters are no longer required/used.
|
|
||
| elsif (TG_LEVEL = 'STATEMENT') then | ||
| if (TG_OP = 'TRUNCATE') then | ||
| execute format('truncate table %I.%I', '$TARGET_SCHEMA$', '$TARGET_TABLE$'); |
There was a problem hiding this comment.
truncate the queue table too
|
|
||
| return _func_def; | ||
| end; | ||
| $func$ language plpgsql stable security invoker |
There was a problem hiding this comment.
| $func$ language plpgsql stable security invoker | |
| $func$ language plpgsql immutable security invoker |
| ); | ||
|
|
||
| execute format( | ||
| 'create trigger %I after insert or update or delete on %I.%I for each row execute function ai.%I()', |
| execute format( | ||
| 'create trigger %I after truncate on %I.%I for each statement execute function ai.%I()', | ||
| format('%s_truncate',_vec.trigger_name) , _vec.source_schema, _vec.source_table, _vec.trigger_name | ||
| ); |
| execute format( | ||
| 'alter extension ai drop function ai.%I()', | ||
| _vec.trigger_name | ||
| ); |
| update ai.vectorizer | ||
| set config = jsonb_set(config, '{version}', format('"%s"', _new_version)::jsonb) | ||
| where id = _vec.id; |
There was a problem hiding this comment.
i just mean if we have a second upgrade like this in the same release that this one goes out in, then we may have an issue because the other upgrade would also be looking for vectorizers with version < 0.8.1
| FROM ai.vectorizer v | ||
| LOOP | ||
| -- Find the foreign key constraint for this vectorizer's store table | ||
| SELECT conname INTO _constraint_name |
There was a problem hiding this comment.
in theory there could be multiple foreign key constraints here, but this would only happen if the users did so manually, and that would be very weird. so, i'm okay with the assumption
No description provided.