Thread: bug: pg_dump use strange tag for trigger
Hi,
When I played with regression tests for pg_restore, I tested -T filtering triggers too. I had problems with restoring triggers. I found that the name for trigger uses the pattern "tablename triggername" (not just (and correct) triggername).
I propose to generate tag just like trigger name
proposed patch attached
regards
Pavel
Attachment
Pavel Stehule <pavel.stehule@gmail.com> writes:
> When I played with regression tests for pg_restore, I tested -T filtering
> triggers too. I had problems with restoring triggers. I found that the name
> for trigger uses the pattern "tablename triggername" (not just (and
> correct) triggername).
> I propose to generate tag just like trigger name
Trigger names by themselves aren't even a little bit unique, so that
doesn't seem like a great idea to me. There's backwards compatibility
to worry about, too. Maybe we need a documentation adjustment, instead?
regards, tom lane
ne 6. 11. 2022 v 15:52 odesílatel Tom Lane <tgl@sss.pgh.pa.us> napsal:
Pavel Stehule <pavel.stehule@gmail.com> writes:
> When I played with regression tests for pg_restore, I tested -T filtering
> triggers too. I had problems with restoring triggers. I found that the name
> for trigger uses the pattern "tablename triggername" (not just (and
> correct) triggername).
> I propose to generate tag just like trigger name
Trigger names by themselves aren't even a little bit unique, so that
doesn't seem like a great idea to me. There's backwards compatibility
to worry about, too. Maybe we need a documentation adjustment, instead?
I understand, but the space is a little bit non intuitive. Maybe use dot there and better documentation.
Regards
Pavel
regards, tom lane