Re: review: CHECK FUNCTION statement

2012/2/28 Alvaro Herrera <alvherre(at)commandprompt(dot)com>:
>
>
> In gram.y we have a new check_option_list nonterminal.  This is mostly
> identical to explain_option_list, except that the option args do not
> take a NumericOnly (only opt_boolean_or_string and empty).  I wonder if
> it's really worthwhile having a bunch of separate productions for this;
> how about we just use the existing explain_option_list instead and get
> rid of those extra productions?

I have to look on it

>
> elog() is used in many user-facing messages (errors and notices).  Full
> ereport() calls should be used there, so that messages are marked for
> translations and so on.

yes, I'll fix it

>
> Does the patched pg_dump work with older servers?
>

yes, It should to do

> I don't like CheckFunction being declared in defrem.h.  It seems
> completely out of place there.  I don't see any better place though, so
> I'm thinking maybe we should have a new header file for it (say
> commands/functions.h; but we already have executor/functions.h so
> perhaps it's better to find another name).  This addition means that
> there's a distressingly large number of .c files that are now getting
> dest.h, which was previously pretty confined.
>

you have much better knowledge about headers then me, so, please,
propose solution.

Regards

Pavel

> --
> Álvaro Herrera <alvherre(at)commandprompt(dot)com>
> The PostgreSQL Company - Command Prompt, Inc.
> PostgreSQL Replication, Consulting, Custom Development, 24x7 support