[bug#42146] [PATCH core-updates 1/?] build: substitute: Don't fail silently.

[Maxim Cournoyer]

Hi Ludo,

Ludovic Courtès <ludo@gnu.org> writes:

> Hi,
>
> Maxim Cournoyer <maxim.cournoyer@gmail.com> skribis:
>
>> * etc/teams.scm.in (<regexp*>, make-regexp*, regexp*-exec): Move to...
>> * guix/build/utils.scm: ... here.
>> (list-matches*): New procedure.
>
>
> [...]
>
>> +++ b/guix/build/utils.scm
>
> [...]
>
>> +;;;
>> +;;; Extend regexp objects with a pattern field.
>> +;;;
>> +(define-record-type <regexp*>
>> +  (%make-regexp* pat flag rx)
>> +  regexp*?
>> +  (pat regexp*-pattern)                 ;the regexp pattern, a string
>> +  (flag regexp*-flag)                   ;regexp flags
>> +  (rx regexp*-rx))                      ;the compiled regexp object
>> +
>> +;;; Work around regexp implementation.
>> +;;; This record allows to track the regexp pattern and then display it.
>> +(define* (make-regexp* pat #:optional (flag regexp/extended))
>
> I’m skeptical about the concrete benefits.  I would not include it in
> (guix build utils), or at least not in this patch series.
>
> (I tend to be super conservative about (guix build utils) because we
> rarely get a chance to change it.)

The original users are substitute* and the teams.scm script.  Since
substitute* is from (guix build utils), it makes sense to add it there
as well, since they are coupled.

The benefit is concrete: it makes it possible to show which regexp
pattern failed to match (its textual representation), instead of
something much less useful such as a generic placeholder such as
"<unknown> (regexp)".

It's in actual use if you look at the definition of substitute:

--8<---------------cut here---------------start------------->8---

  (let ((rx+proc (map (match-lambda
                        (((or (? regexp? pattern) (? regexp*? pattern)) . proc)
                         (cons pattern proc))
                        ((pattern . proc)
                         (cons (make-regexp* pattern regexp/extended) proc)))
                      pattern+procs)))
--8<---------------cut here---------------end--------------->8---

The previous version followed a different approach, annotating the
rx+proc list with the raw pattern; I think the approach here is a bit
cleaner, and it should also enable users to pass pre-computed regexp*
objects to substitute* and have useful error messages produced.

-- 
Thanks,
Maxim