[bug#50286] [RFC PATCH] Let 'package-location' returns location of surrounding 'let'.

[Sarah Morgensen]

Hi Ludo,

Ludovic Courtès <ludo@gnu.org> writes:

> Example:
>
> scheme@(guile-user)> ,use(gnu packages base)
> scheme@(guile-user)> ,use(gnu packages accessibility)
> scheme@(guile-user)> ,use(guix)
> scheme@(guile-user)> (package-location footswitch)
> $1 = #<<location> file: "gnu/packages/accessibility.scm" line: 257 column: 4>
> scheme@(guile-user)> (package-definition-location footswitch)
> $2 = #<<location> file: "gnu/packages/accessibility.scm" line: 254 column: 0>
> scheme@(guile-user)> (package-location hello)
> $3 = #<<location> file: "gnu/packages/base.scm" line: 79 column: 2>
> scheme@(guile-user)> (package-definition-location hello)
> $4 = #<<location> file: "gnu/packages/base.scm" line: 78 column: 0>
>
> Thoughts?

This is very clever!  Thanks for the work on this.  I'm not very good
with macros, but it *looks* like it should work quite well for our
use-case of adjusting a surrounding 'let' expression.  And it's less
invasive than rewriting 'let'.

However... it doesn't work for unexported packages.  It looks there are
about 200 such packages:

--8<---------------cut here---------------start------------->8---
~/guix$ rg -U '\(define [^\(]+\n.*?\(package' gnu/packages --count 
--no-filename | awk '{a+=$1} END {print a}'
233
--8<---------------cut here---------------end--------------->8---

And, to play the pessimist:

What do we get out of this that couldn't be done by "go to package
location; read backwards one sexp until we reach a defining form"
(like Emacs' 'beginning-of-defun')?

--
Sarah