switch-to-relevant buffer

[jonpither]

Hi I added this function 'nrepl-switch-to-relevant-buffer'. It switches the user to the right nrepl buffer based on the current buffer. So if you have 2 nrepls open for project x and project y, if you open a project x file and run the function (key mapped of course) then you land on the right nrepl buffer. I couldn't see anything that caters for this in nrepl.el hence the pull request.

TBH I'm new to elisp so you might want to make sure it looks sane. Unfortunately since I moved this fn into nrepl.el, I can't seem to use it without eval'ling the nrepl.el in emacs. Not sure why this is? I get:

Symbol's function definition is void: nrepl-switch-to-relevant-buffer. I've tried byte-recompiling my emacs dir to no avail.

Not sure what the issue is.

[bbatsov]

This docstring is missing an ending dot.

[bbatsov]

I think that this behaviour should be integrated into the existing switch to REPL buffer command - it's pretty sensible to assume people would won't to go to proper REPL buffer using that command.

[jonpither]

Thanks - I've made all the changes bar using cl-remove-if-not and merging into the existing switch command.

I couldn't find cl-remove-if-not... I played around but couldn't improve upon remove-if-not.

As commented in http://stackoverflow.com/questions/2234860/lisp-filter-out-results-from-list-not-matching-predicate, I tried using remove-if, but complement isn't present in elisp.

WRT to merging into the existing switch-command I'm happy to do this. Might be worth a second pull-request as it's changing existing behaviour? But whatever I'll work on it.

[bbatsov]

cl-remove-if-not is part of cl-lib for sure. Seems pretty odd that you don't have it. The problem with remove-if is that it's also part of the now deprecated cl module (as is remove-if-not).

The problem with having this functionality in a separate command is that probably nobody will notice and as I already said - I cannot imagine why I'll want to jump to a REPL that's not connected to my current project so I'd rather us change the existing (and widely used command).

@hugoduncan @purcell @technomancy @samaaron Thoughts?

[jonpither]

I've merged the two fns. I'm running with emacs 24.2 which is why I can't seem to get cl-remove-if-not and friends:

http://williamjohnbert.com/2013/05/emacs-cl-lib-madness/

I'll upgrade.

[jonpither]

Might it be an issue with users on older Emacs versions though?

[hugoduncan]

What happens when there are multiple REPLs for a single project? I assume it would just pick the "first", which should be the most recently used, iirc.

Building it into jump to repl should be fine, but should be overridable with a prefix, since it is not impossible that you have REPLs for a project and a subproject open, and want to work with the top level project REPL when changing subproject files. Not sure if checkout dependencies complicate anything further here - I don't think so.

[jonpither]

Have used cl-remove-if-not as discussed (after emacs upgrade)

I moved nrepl-remember-clojure-buffer up as it was being used before it was defined.

@hugoduncan it does pick the first match if you were to have multiple nrepl sessions open in the same project dir.

TBH - not sure how the prefix would function? Like a global flag to turn off the extra behaviour?

[bbatsov]

@jonpither A prefix argument means that when the function is invoked like this C-u C-c C-z it will behave differently from the normal invocation C-c C-z. Have a look at the source of nrepl-jack-in for details. I think @hugoduncan is right and that we should add prefix argument support to C-c C-z. More details can be found here - http://www.emacswiki.org/emacs/PrefixArgument

[jonpither]

Thanks @bbatsov. I understand the prefix behaviour now :-) The same ido-read-directory-name approach could be taken as for nrepl-jack-in.

There is though already an arg taken by nrepl-switch-to-repl-buffer to put the user into the namespace of the file they are on... thoughts?

[bbatsov]

@jonpither Pretty simple - we have C-u C-c C-z behave like it currently does and make C-u C-u C-c C-z trigger the new special behavior. Basically the number of prefix arguments will control the behavior. This is a pretty common practice in Emacs land.

[jonpither]

@bbatsov I've added the second prefix arg that prompts for a directory in the last commit.

The only issue is that if you want the new behaviour of specifying the directory of the repl session you want to jump to through C-u C-u C-c C-z, then the first arg is still also activated, meaning the NS will be changed to whatever the current buffer is (or user if the current buffer does not have a ns open).

This is probably fine, but it may annoy if you want to change to a specific repl but not switch ns.

If there's an easy way of having the second behaviour but not the first, it may be worth doing.

[bbatsov]

You can make this interactive "p" to receive directly a numeric prefix argument. For C-u C-c C-z it would be 4 and for C-u C-c C-z it would be 16.

[bbatsov]

I've commented on the relevant bit of code - basically (interactive "p") and comparisons whether arg is 4 or 16 is what you want to do.

[jonpither]

Have added as per your comment, the two behaviours are triggered separately. Have added to the docs.

[bbatsov]

OK. Now just update the Changelog and squash all the commits into one and we're good to merge this.

[jonpither]

Should be good to go