>
> Bruce Momjian writes ("Re: [HACKERS] Case sensitivity in identifiers"):
> > > I wouldn't have thought it was a particularly big job; you keep these
> > > identifiers in postgres tables anyway: why not a new batch of char
> > > types and operators using strcasecmp instead of strcmp?
> [snip]
> >
> > It is not that easy. We use indexes and cached keys to look up things.
> > Doing such things case-insensitive is a big job.
>
> OK, well I'm sure you realise I don't really know what I'm talking
> about, but PostgreSQL seems so easily user-extensible and I thought it
> used the same methods internally. I'll get back to you when I do know
> what I'm talking about.
>
> In the mean time to help me get over the incompatibilities the current
> implementation has thrown up, i.e. revert to case-sensitivity, is it
> sufficient to remove the 'tolower' call (loop) in the {identifier}
> chunk in scan.l, and revert to strcasecmp in ScanKeywordLookup() in
> keywords.c ? If not, if you still have the patch you installed to
> implement the current method, could you mail it to me so that I can
> (automatically or manually) uninstall the patch, please? Cheers!
It was done in pieces, not as one patch. The scan.l is the major part
of it.
However, we needed to make changes to libpq where column and database
names are specified, and any hard-coded table name patterns like Inv*.
- --
Bruce Momjian
maillist@candle.pha.pa.us
------------------------------
