On 2024/09/09 11:38:04 +0200, Stefan Sperling <stsp@stsp.name> wrote:
> On Mon, Sep 09, 2024 at 11:12:48AM +0200, Omar Polo wrote:
> > Here's the diff.  It supports quoting with both single and double
> > quotes, as well as escaping.  plus, there's also a new regress.
> > 
> > One more thing that I'd like to do after this maybe is rename the "auth"
> > keyword to "http".

actually I think it's not a good idea.  we may want to reuse this syntax
also for smtp auth, so "auth" is fine as general username/password pair.

> > ok?
> 
> Thanks, ok by me.

committed, thanks.

> Abiding by the principle of "never write your own lexer" I wonder if
> we shouldn't be trying to use a parse.y-based parser for this instead
> of a custom oner. But what you have works, and has tests, and the input
> is created by the sysadmin. So I'm not too worried.

I guess the motto is more like "never write your own parser" since all
the parser we have are using a custom lexer (since lex(1) is probably the
more counter-intuitive program ever written) :P

I don't know, I like yacc in general so I won't feel bad if someone
wants to rewrite it in yacc.  I've been told there are preprocessor
tricks to don't have symbol clashes.

> There are some typos and the new test output could be tweaked to
> match existing tests. But those aren't blocking problems.

do you mean asserting the error messages too?