Re: Better way to make sure external command exists in the system?

[2QdxY4RzWzUUiLuE]

On 2021-03-19 at 14:57:34 +0100,
Daniel Martín <mardani29@yahoo.es> wrote:

> Jean Louis <bugs@gnu.support> writes:
> 
> > I have changed the OS on one computer and noticed that my Emacs Lisp
> > programs use external commands like: mailutils, mail, pandoc,
> > markdown, and others. Sometimes I was hard coding the path names like
> > /usr/local/bin and now some commands changed to /usr/bin
> >
> > And some functions should not even run if external command does not
> > exist. I would even like to stop loading the .el program if necessary
> > external programs do not exist in the system.
> >
> > For that reason I would do some changes:
> >
> > - instead of hard coding the program name within quotes, I will use
> >   variable. Instead of (shell-command "mogrify ...") I will rather use
> >   something like (shell-command (format "%s ..." mogrify-command))
> >
> > - variables for commands I would place at beginning of programs
> >
> 
> Yes, I think that using a variable instead of a hardcoded string is
> recommended.
> 
> > - before variables get defined, I would use `executable-find' but I
> >   would like for the program loading to fail if those external
> >   programs cannot be found.
> 
> Why do you think that checking for the existence of programs on load is
> better? I think it would slow down loading the module, and the benefits
> are not clear to me.  For example, the Elisp module may still be useful
> even if it cannot call external programs.
> 
> Why is it better than simply calling `executable-find' when invoking a
> command that depends on an external program?

Using executable-find at load time is better because modules are only
loaded once, but the functionality inside the modules is executed
repeatedly.  Personally, my emacs.el loads lots of stuff
unconditionally, because I usually only start Emacs when I start my X
session, which is already time consuming.  Anything I don't load at that
time only delays me later.

And if every function in a module is a wrapper around, say, mogrify,
then why shouldn't I verfiy that mogrify exists before I load the
module?  (Obviously, there are a number of user- and application-
interface questions to be answered, but there are definitely modules
that can't do anything (except report failure) without their associated
external dependencies.)