Dne 28. února 2012 17:48 Alvaro Herrera <alvherre(at)commandprompt(dot)com> napsal(a):
> I have a few comments about this patch:
> I didn't like the fact that the checker calling infrastructure uses
> SPI instead of just a FunctionCallN to call the checker function. I
> think this should be easily avoidable.
It is not possible - or it has not simple solution (I don't how to do
it). PLpgSQL_checker is SRF function. SPI is used for processing
returned resultset. I looked to pg source code, and I didn't find any
other pattern than using SPI for SRF function call. It is probably
possible, but it means some code duplication too. I invite any ideas.
> Second, I see that functioncmds.c gets a lot into trigger internals just
> to be able to figure out the function starting from a trigger name. I
> think it'd be saner to have a new function in trigger.c that returns the
> required function OID.
> I think CheckFunction would be clearer if the code to check multiple
> objects is split out into a separate subroutine.
> After CheckFunction there is a leftover function comment without any
> following function. There are other spurious hunks that add or remove
> single lines too (once in an otherwise untouched file).
I refreshed patch for current git repository.
> Álvaro Herrera <alvherre(at)commandprompt(dot)com>
> The PostgreSQL Company - Command Prompt, Inc.
> PostgreSQL Replication, Consulting, Custom Development, 24x7 support
In response to
pgsql-hackers by date
|Next:||From: David E. Wheeler||Date: 2012-02-28 19:33:08|
|Subject: Re: pgsql_fdw, FDW for PostgreSQL server|
|Previous:||From: Simon Riggs||Date: 2012-02-28 19:21:41|
|Subject: Re: CLOG contention, part 2|