Allow colon : for symbols in binding of evil-ex-grammar

[Alexander-Shukaev]

No description provided.

[Alexander-Shukaev]

Just a friendly ping to merge this change.

[TheBB]

Sorry for the delay. Is there an example ex command where this is needed?

[Alexander-Shukaev]

I defined a number of custom ex commands for myself, which begin with :, so that I can quickly type ::command as evil-ex is itself :.

[TheBB]

But if you define them yourself then you can choose different names. The ex parser is somewhat fragile and I'm reluctant to change it just for that.

I tried adding a test for this and as you can see it fails to parse properly. If you experience differently I'd like to see the command(s) you've added.

[TheBB]

I guess it's the colon in the first branch of the regexp that's greedily matching just the first colon of :command.

[Alexander-Shukaev]

Apologies, I think you misunderstood that I meant custom ex commands (i.e. defined with evil-ex-define-cmd), not the regular Emacs interactive functions. In this case, the test should be against (evil-ex-parse-command ":command").

[TheBB]

It still doesn't work as expected.

ELISP> (defun :command () (interactive))
:command
ELISP> (evil-ex-parse-command ":command")
(":" . "command")

ELISP> (defun command () (interactive))
command
ELISP> (evil-ex-parse-command "command")
("command" . "")

ELISP> (defun comm:and () (interactive))
comm:and
ELISP> (evil-ex-parse-command "comm:and")
("comm:and" . "")

Do you see other behaviour?

[Alexander-Shukaev]

Once again, you have to use evil-ex-define-cmd rather than plain interactive function.

[TheBB]

Sorry. I misunderstood in which direction I was supposed to have misunderstood. Then my original comment stands: you can substitute the defuns with evil-ex-define-cmd in my previous comment and the behaviour is identical.

[Alexander-Shukaev]

I get

ELISP> (evil-ex-parse-command ":command")
(":command" . "")

[TheBB]

Okay. I rebased this branch on top of master, which also has #1203, so maybe that interferes (if you haven't updated the PR branch since). If not then I'm not sure why the difference.

[Alexander-Shukaev]

Right, since I was rather customizing the variable, I had the following

 (binding "[[:alpha:]-:]+\\|[~&*@<>=:]+\\|[[:alpha:]-]+\\|!")

which works but is not what we want to release of course. Sorry.