On Wed, Jun 24, 2020 at 10:27 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> More generally, though, why would we want to change this policy only
> here? I believe we're reasonably consistent about letting the parser
> do any required down-casing and then just checking keyword matches
> with strcmp.
I've had the feeling in the past that our use of pg_strcasecmp() was a
bit random. Looking through the output of 'git grep pg_strcasecmp', it
seems like we don't typically don't use it on option names, but
sometimes we use it on option values. For instance, DefineCollation()
uses pg_strcasecmp() on the collproviderstr, and DefineType() uses it
on storageEl; and also, not to be overlooked, defGetBoolean() uses it
when matching true/false/on/off, which probably affects a lot of
places. On the other hand, ExplainQuery() matches the format using
plain old strcmp(), despite indirectly using pg_strcasecmp() for the
Boolean parameters. So I guess I'm not really convinced that there is
all that much consistency here.
Mind you, I'm not sure whether or not anything really needs to be
changed, or exactly what ought to be done. I'm just making the
observation that it might not be as consistent as you may think.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company