"David G. Johnston" <david.g.johnston@gmail.com> writes:
> So I can see an argument for the existing behavior. It doesn't seem worth
> changing in any case. And I don't really see the documentation being
> improved by covering this corner case in detail when the current behavior
> is at least intuitive.
quote_ident is a good bit more conservative than the core lexer
about what is an "identifier character" --- it considers all
non-ASCII characters as requiring quoting, too.
For typical uses of quote_ident, I think this is good future-proofing:
it makes it very much less likely that something quote_ident decides
not to quote would be rejected by some future PG version (not to mention
non-PG SQL databases). So I'm not really in a hurry to change the code.
Maybe we should tweak the docs a bit.
regards, tom lane