Modify mode-line lighter customization

[sellout]

This adds a new customizable variable, envrc-lighter-function that sits between envrc-lighter and envrc-STATUS-lighter. This allows overriding envrc-lighter without relying on internal definitions (envrc--status).

The other benefit of this is that any customization is likely to want to set all three envrc-STATUS-lighter, with them being mostly the same, with a small conditional part (as you can see from both envrc--default-lighter and my lambda below). While this variable is a risky-local-variable, I believe that envrc-STATUS-lighter are also risky, because they can also contain :eval, no?

If envrc-lighter-function is not customized, the existing envrc-STATUS-lighter variables should continue to work as they have.

In future, I think this could be the only -lighter customizable variable.

I like my minor modes to be pretty minimal, so I use this like

(setq envrc-lighter-function
      (lambda (status)
        `(:propertize "🗁" ; looks like GH’s font doesn’t have a glyph for “open folder” 
                      face
                      ,(pcase status
                         ('on 'envrc-mode-line-on-face)
                         ('error 'envrc-mode-line-error-face)
                         (_ 'envrc-mode-line-none-face)))))

[purcell]

Thanks! I figured people would just write their custom function and set envrc-lighter to (:eval (my-lighter-function). Part of me thinks there's no point changing custom vars in order to "unwrap" that function, but in practical terms I don't think anyone is going to set the lighter to a constant string, for example, so I'm +1 on this PR. If we're doing that, I'd also happy to just go ahead and remove envrc-lighter and mark it obsolete, plus remove the various envrc-STATUS-lighter vars. Would you like to amend the PR to do that too?

[purcell]

P.S. Sorry that CI is a bit broken currently, I still haven't figured out what's wrong with the failing tests.

[sellout]

tl;dr: This is basically just rambling, no real information here other than “I’ll obsolete the oher vars”.

Thanks! I figured people would just write their custom function and set envrc-lighter to (:eval (my-lighter-function). Part of me thinks there's no point changing custom vars in order to "unwrap" that function, but in practical terms I don't think anyone is going to set the lighter to a constant string, for example, so I'm +1 on this PR.

An alternative I meant to mention in the description is just making envrc--status non-internal (envrc-status). And then setting envrc-lighter directly makes more sense. I also feel like exposing envrc--status is an API improvement, perhaps with an envrc-status-change-hook so users can make other decisions based on envrc (although I have no immediate use case for this).

If we're doing that, I'd also happy to just go ahead and remove envrc-lighter and mark it obsolete, plus remove the various envrc-STATUS-lighter vars.

I think removing envrc-STATUS-lighter is the real goal of this PR. Historically, I had configured the lighter like

(defun sellout--envrc-lighter (status)
  "Return a lighter for the provided envrc STATUS."
  `(:propertize "🗁"
                face
                ,(pcase status
                   ('error 'envrc-mode-line-error-face)
                   ('on 'envrc-mode-line-on-face)
                   (_ 'envrc-mode-line-none-face))))

(use-package envrc
  :custom
  (envrc-error-lighter (sellout--envrc-lighter 'error))
  (envrc-none-lighter (sellout--envrc-lighter 'none))
  (envrc-on-lighter (sellout--envrc-lighter 'on))
  :hook (after-init . envrc-global-mode))

Which I prefer over

(use-package envrc
  :custom
  (envrc-lighter '(:eval (sellout--envrc-lighter envrc--status)))
  :hook (after-init . envrc-global-mode))

because … I guess just because of envrc--status being internal. But, despite me doing that, I don’t see much of a case for wanting to set the envrc-STATUS-lighters individually. I guess I feel like by setting envrc-lighter directly, I would be bypassing part of system, making envrc-STATUS-lighter useless. So getting rid of envrc-STATUS-lighter just reduces my idiosyncratic mental stress 😅

Another possible benefit of envrc-lighter-function is that instead of having envrc-lighter use :eval, envrc--apply (or something around there) could (setq envrc-lighter (envrc-lighter-function envrc--status)), so that the lighter only gets updated when the environment does, rather than on every mode-line redraw.

Would you like to amend the PR to do that too?

Yeah, I’m happy to.

[purcell]

Yeah, I guess I'm not super averse to making envrc--status public, ie. envrc-status. Or – potentially safer for the long run – exposing it via a corresponding function.

Regarding :eval or not, I'd probably be equally happy with the envrc-lighter-function path.