[bug#41164] [PATCH 0/3] Add 'guix graph --path'

[zimoun <zimon.toutoune@gmail.com>]

Hi Ludo,

Awesome!
I remember discussing such feature at FOSDEM. :-)
And a couple of days ago (updating with core-updates), I was annoyed
because a lot of "unexpected" packages were downloaded, I asked myself
"why".  So you removed one item of my feature wishlist. ;-)


On Sun, 10 May 2020 at 01:05, Ludovic Courtès <ludo@gnu.org> wrote:

> --8<---------------cut here---------------start------------->8---
> $ ./pre-inst-env guix graph --path emacs libffi
> --8<---------------cut here---------------end--------------->8---

Well, the command is obviously not symmetric (oriented).  The path is
from 'emacs' to 'libffi'.

It appears to me not clear in '--help'.

--8<---------------cut here---------------start------------->8---
display the shortest path between the given nodes
--8<---------------cut here---------------end-------------->8---

Well, I am already bikeshedding but the CLI "guix graph emacs
--path-to libffi" appears to me clearer (or '--why' as "why" is
mentioned and underlined in the doc :-)).


> There’s a bikeshedding opportunity in the last patch: should it go
> in ‘guix graph’ or elsewhere?  I think ‘guix graph’ is a good home
> for that, and could eventually include more graph queries.  For
> instance, ‘guix refresh -l’ could very well live in ‘guix graph’.

I do not have a strong opinion.  The "graph" subcommand fits well.
But for example:

  guix show emacs --why libffi
  guix search emacs --why libffi

make sense too -- at least to me. ;-)


And bikeshedding again, there is a temptation  to pipe the current CLI:

  guix graph --path emacs libffi | guix show

which does not obviously work because of "show".  From my point of
view, this CLI seems good:

    guix search emacs --why libffi | guix show --format=oneline

emacs@26.3 The extensible, customizable, self-documenting text editor
gnutls@3.6.9 Transport layer security library
guile@3.0.2 Scheme implementation intended especially for extensions
libffi@3.3 Foreign function call interface library

And '--format' could be 'recutils' or whatever.  But that another story. :-)


Last, it is not new but confusing, the node "guile@3.0.2" is returned
but it is reachable with "guile-next".


All the best,
simon