Re: [PATCH] proposal for regexp_count, regexp_instr, regexp_substr and regexp_replace

I wrote:
> ... aside from the question of whether
> a too-large subexpression number should be an error or not.

Oh ... poking around some more, I noticed a very nearby precedent.
regexp_replace's replacement string can include \1 to \9 to insert
the substring matching the N'th parenthesized subexpression.  But
if there is no such subexpression, you don't get an error, just
an empty insertion.  So that seems like an argument for not
throwing an error for an out-of-range subexpr parameter.

            regards, tom lane