Michael Paquier <michael.paquier@gmail.com> writes:
> On Sat, Jan 2, 2016 at 4:06 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> The grammar fixes seem like a good thing to do in the long run, too,
>> but there's little need to risk back-patching it since accepting
>> col_name_keywords without quoting would be mostly a convenience issue.
> A different angle of attack is to flatten the argument quotes directly
> in reloptions.c:
> http://www.postgresql.org/message-id/CAB7nPqTpdGLqLTxuGhBC2GabGNiFRAtLjFbxu=aGy1rX_DgwUg@mail.gmail.com
> But you did not like that :p
It seemed pretty messy. There is nothing very wrong with the convention
that pg_class.reloptions is an array of "name=value" entries with both
name and value being taken literally. The only thing that rule excludes
is that the option name cannot include an "=", which is a restriction that
bothers me not at all.
The dumped form of reloptions needs to meet the grammar restrictions on
what can be in WITH, but that's really a separate question.
The bug we had was that pg_dump and ruleutils.c weren't considering that
the rules might be different for the two representations. Yeah, you could
fix it by insisting that the rules be identical, but I don't really find
that cleaner (and it could not be a back-patchable fix for existing
databases, anyway).
regards, tom lane
