bug#36237: Support (rx (and (regexp EXPR) (regexp-quote EXPR)))

[Noam Postavsky]

> [ BTW, please use "--" for rx-compile-to-lisp since I believe it's
>   internal.  ]

I'm not sure that it is, e.g., python-rx might bind it too (if it
weren't for the fact that python.el needs to maintain backwards
compatibility with older Emacs).

> What do regexp and regexp-quote do in rx-to-string?

regexp does exactly what it did before, i.e., it accepts only a constant
string.  Likewise regexp-quote accepts only a constant string, which
makes it pointless to use in rx-to-string (just use a plain STRING
directly), but I didn't disallow it.

Hmm, I think I had meant to update rx-to-string's docstring, but forgot
about it.

>> - A more suitable name is needed; `regexp-quote' makes no sense at all in
>> the context of rx. What about `literal'? For simplicity, I'm proceeding with
>> that name below.
>
> I'll let others figure that one out.

Probably `literal' makes sense.  I originally used regexp-quote, just
because I was thinking of it as a short form of

    (rx (regexp (regexp-quote EXPR)))

>> - What is now the correct way of including a compile-time regexp expression,
>> such as a defconst? (regexp (eval-when-compile EXPR))? Still a mouthful, but
>> perhaps outside the scope of this bug.

Oh, hmm.  That might be a reason to keep using `eval'.

> FWIW, I have the impression that in most cases where this could be
> useful, a better solution would be to provide something like
> `rx-defmacro` and/or `rx-macrolet`.

I guess that could replace the "old crusty" rx-constituents thing too.

>> I have: I'm sitting on a full rewrite, code-named `ry'. It's shorter,
>> much cleaner, and about twice as fast (usually more). The only thing
>> still missing is compatibility with the old crusty `rx-constituents'
>> extension mechanism.
>> 
>> The plan was to replace rx with ry entirely when complete.

How far away is this?  Would it make sense to delay this bug until "ry"
comes in?