rx of (any SET...)

rx of (any SET...)

[Michael Albinus]

Hi,

I have the regexp

--8<---------------cut here---------------start------------->8---
(concat "[^" (bound-and-true-p mm-7bit-chars) "]")
--8<---------------cut here---------------end--------------->8---

I want to write it in rx notation, but I fail. I know how to use xr for
reverse engineering, but I'd like to keep the variable mm-7bit-chars in
the rx notation, and not to replace it by the literal string.

How would this look like?

Best regards, Michael.

RE: [External] : rx of (any SET...)

[Drew Adams]

> I have the regexp
> (concat "[^" (bound-and-true-p mm-7bit-chars) "]")

Dunno about rx.  But what happens when `mm-7bit-chars'
isn't bound and true?  Your string is then "[^]", not
a valid regexp.  (Maybe you take care of that elsewhere?)

Re: [External] : rx of (any SET...)

[Michael Albinus]

Drew Adams <drew.adams@oracle.com> writes:

Hi Drew,

> Dunno about rx.  But what happens when `mm-7bit-chars'
> isn't bound and true?  Your string is then "[^]", not
> a valid regexp.  (Maybe you take care of that elsewhere?)

In practice, it always works. This is because this check is used in
tramp-bug, while composing the message. And while composing the message,
package mm-bodies is already loaded. Using bound-and-true-p is just for
pacifying the byte compiler.

This code snippet is 20 years old. Time to replace bound-and-true-p by
(defvar mm-7bit-chars)

But the question remains: how to express (concat "[^" mm-7bit-chars "]")
in rx notation?

Best regards, Michael.

Re: rx of (any SET...)

[Michael Heerdegen]

Michael Albinus <michael.albinus@gmx.de> writes:

> (concat "[^" (bound-and-true-p mm-7bit-chars) "]")
>
> I want to write it in rx notation, but I fail. I know how to use xr for
> reverse engineering, but I'd like to keep the variable mm-7bit-chars in
> the rx notation, and not to replace it by the literal string.

C-h f rx RET:
| (regexp EXPR) Match the string regexp from evaluating EXPR at run time.

That's exactly what you are looking for...correct?


(the other) Michael

Re: rx of (any SET...)

[Harald Jörg]

Michael Albinus writes:

> Hi,
>
> I have the regexp
>
> (concat "[^" (bound-and-true-p mm-7bit-chars) "]")
>
> I want to write it in rx notation, but I fail. I know how to use xr for
> reverse engineering, but I'd like to keep the variable mm-7bit-chars in
> the rx notation, and not to replace it by the literal string.
>
> How would this look like?

I found that variables can be included in rx forms with some form of
eval, or by using rx-to-string, which is a function and not a macro.

With the value of mm-7bit-chars at macro expansion time:
(rx (eval `(not (any ,mm-7bit-chars))))

With the value of mm-7bit-chars at runtime:
(eval `(rx (not (any ,mm-7bit-chars))))
(rx-to-string `(not (any ,mm-7bit-chars)))

--
Cheers,
haj

Re: rx of (any SET...)

[Michael Heerdegen]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> | (regexp EXPR) Match the string regexp from evaluating EXPR at run time.
>
> That's exactly what you are looking for...correct?

But ok, you have a string that contains a list of chars, so I too have
no better idea than

  (rx (regexp (concat "[^" (bound-and-true-p mm-7bit-chars) "]")))

or to use the function instead:

  (rx-to-string `(not (any ,@mm-7bit-chars)))

What makes sense depends on the context.

Michael.

Re: rx of (any SET...)

[Michael Heerdegen]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> (rx (regexp (concat "[^" (bound-and-true-p mm-7bit-chars) "]")))

Even using `rx-define' is not so helpful (its expanding mechanism is
different from e.g. defmacro):

#+begin_src emacs-lisp
  (rx-define any-char (&rest string) (regex (concat "[" string "]")))
  (rx-define not-any-char (&rest string) (regex (concat "[^" string "]")))
#+end_src

Will not work inside `not'.  Thus two definitions.

Ok, in the end I could not really help at all...

Michael.

Re: rx of (any SET...)

[Michael Albinus]

Harald Jörg <haj@posteo.de> writes:

Hi Harald,

> I found that variables can be included in rx forms with some form of
> eval, or by using rx-to-string, which is a function and not a macro.
>
> With the value of mm-7bit-chars at macro expansion time:
> (rx (eval `(not (any ,mm-7bit-chars))))
>
> With the value of mm-7bit-chars at runtime:
> (eval `(rx (not (any ,mm-7bit-chars))))
> (rx-to-string `(not (any ,mm-7bit-chars)))

Thanks, that does the trick!

> Cheers,
> haj

Best regards, Michael.