select the *rg* buffer after running a search

[hmelman]

After running an rg search, I'd like an option to have the results buffer (typically *rg*) be selected. I know about using next-error, but this way I could easily navigate and also quit, edit (via wgrep) or rerun the search.

[dajva]

This should be a compilation-mode feature since that is what is used in rg. I have not seen any way of customizing that in compilation-mode itself unfortunately but I think next-error-last-buffer should be set so you can use that in an rg-mode-hook to do what you need.

A naive setup may be:

(defun select-rg-buffer ()
  (switch-to-buffer next-error-last-buffer))
(add-hook 'rg-mode-hook select-rg-buffer)

Not sure how you get that to play well with how compilation buffers are displayed in general but works ok for me that have the default setup.

[hmelman]

This is what I have been using:

(with-eval-after-load 'rg
  (advice-add 'rg-run :after
	      #'(lambda (_pattern _files _dir &optional _literal _confirm _flags) (pop-to-buffer (rg-buffer-name)))))

It's seems reasonable that it should be a compilation-mode feature, though I'd want this for rg but not for M-x compile where next-error would be fine for me.

Your suggestion didn't work for me as next-error-last-buffer wasn't correct, also the add-hook call should be:

(add-hook 'rg-mode-hook #'select-rg-buffer)

[ktetzlaff]

Here's another variant:

(eval-after-load 'rg
  (defun maybe-pop-to-rg-buffer(&optional buffer _)
    "Switch to BUFFER (default: rg buffer) and select first match."
    (with-current-buffer (or buffer (rg-buffer-name))
      (when (string= mode-name "rg")
        (pop-to-buffer (current-buffer))
        (condition-case nil
            (compilation-next-error 1 nil (point-min))
          (error nil)))))
  (add-to-list 'compilation-finish-functions #'maybe-pop-to-rg-buffer))

[dajva]

I guess I can add an rg-finish-functions variable that can be used instead of compilation-finish-functions. In that way the last example should work without finding the rg buffer. Something like this:

(add-to-list 'rg-finish-functions (lambda (buffer _) (pop-to-buffer buffer)))

Would this work for your use cases? I guess one problem for long running searches is that the buffer will not be selected until the search is finished.

[hmelman]

Yeah, with the advice I'm using, it switches to the rg buffer at the start of a long running search, which is nice. Maybe rg-finish-functions could be run at the end of rg-run like my advice? I don't have another use for such a hook so if rg itself could select the buffer via user option I'm fine with that.

[dajva]

Implemented the rg-finish-functions solution for now.

[hmelman]

I see the defvar in the commit but I don't see where they are called. Am I missing something?

[dajva]

Probably the define-compilation-mode inner workings. If defined it's picked up by compilation mode and handled from there. So the defvar isn't strictly needed (any setq would work) but will allow for more flexible usage plus visibility in the package.

[hmelman]

Seems to be working :)

[jeremyf]

I just verified with the rg-finish-functions. @hmelman would you be able to close this issue as resolved?

[hmelman]

I think so. It has been working fine for me though I have one minor issue, but I don't think this has to stay open for it.

I have global-hl-line-mode enabled and the line in the *rg* buffer is not highlighted until I move the cursor in that buffer. Highlighting is run on change-major-mode-hook and post-command-hook so I get why moving point highlights a line, but I hadn't looked into why it's not highlighted initially. Seems like if I put (global-hl-line-highlight) in the rg finish function it works, though if you know of a cleaner approach it would be nice.

[hmelman]

Also, should rg-finish-functions be a defcustom instead of a defvar?

[dajva]

Also, should rg-finish-functions be a defcustom instead of a defvar?

Possibly. I have followed conventions of compilation-mode here which is using defvar for for hooks. Personally I wouldn't use customizing system for this kind of configuration but I don't mind changing this either. Please submit a PR if you want this.

[hmelman]

Also, should rg-finish-functions be a defcustom instead of a defvar?

Possibly. I have followed conventions of compilation-mode here which is using defvar for for hooks. Personally I wouldn't use customizing system for this kind of configuration but I don't mind changing this either. Please submit a PR if you want this.

For me the defcustom is more about discoverability and documentation to see it in customize-group of rg and then setting it in code. Searching the Emacs source base there's a mix of defcustom vs defvar for hooks, though defvar is used more often for -functions, so I think it's your choice and defvar is fine.