Neil Conway <neilc@samurai.com> writes:
> BTW, this is a little off-the-wall, but one interesting idea to help SQL
> and PL/foo integration would be to replace the bison grammar for SQL
> with a hand-written recursive descent parser.
Ick. I gave up hand-written RD parsers twenty-five years ago. They are
tedious to write, error-prone (due to the utter lack of any checking
that what you wrote embodies the syntax you meant), less flexible than
LALR(1), and generally have few redeeming social advantages.
Which is not to say that plpgsql in its current implementation isn't
mighty ugly too. I'd dearly love to find a better solution ... but
"throw away the parser generator" isn't a better solution.
regards, tom lane