JumpForward & JumpBackward triggers defined only during snippet expansion

[mMontu]

When using the default configuration the keys <c-k> and <c-j> are mapped to jump between the placeholders. But as these mappings are always defined, they interfere with the normal function of these keys.

For instance, in insert mode the dictionary completion (:h i_CTRL-X_CTRL-K) doesn't work due to the <c-k> mapping created after a snippet is expanded.

Of course changing the value of g:UltiSnipsJumpBackwardTrigger would solve this problme, but I find these mappings very intuitive so I'd prefer to avoid this approach.

Instead it would be nice if the mappings are defined only when expanding the snippet. I've tried to do this with the following functions:

function! UltiSnips_MapInnerKeys()
    " TODO: use maparg() to save current mapping (if any) to these keys, which
    " can be restored on UltiSnips_RestoreInnerKeys
    if g:UltiSnipsExpandTrigger != g:UltiSnipsJumpForwardTrigger
        exec "inoremap <silent> " . g:UltiSnipsJumpForwardTrigger . " <C-R>=UltiSnips_JumpForwards()<cr>"
        exec "snoremap <silent> " . g:UltiSnipsJumpForwardTrigger . " <Esc>:call UltiSnips_JumpForwards()<cr>"
    endif
    exec "inoremap <silent> " . g:UltiSnipsJumpBackwardTrigger . " <C-R>=UltiSnips_JumpBackwards()<cr>"
    exec "snoremap <silent> " . g:UltiSnipsJumpBackwardTrigger . " <Esc>:call UltiSnips_JumpBackwards()<cr>"
endf

function! UltiSnips_RestoreInnerKeys()
    if g:UltiSnipsExpandTrigger != g:UltiSnipsJumpForwardTrigger
        exec "iunmap " . g:UltiSnipsJumpForwardTrigger
        exec "sunmap " . g:UltiSnipsJumpForwardTrigger
    endif
    exec "iunmap " . g:UltiSnipsJumpBackwardTrigger
    exec "sunmap " . g:UltiSnipsJumpBackwardTrigger
endf

Removing the mappings definition from UltiSnips_MapKeys() and placing calls to function UltiSnips_MapInnerKeys() on the start of UltiSnips_ExpandSnippet()/UltiSnips_ExpandSnippetOrJump() works as expected, but due to my limited knowledge of Python I was unable to find the point to insert the call to remove (and possible restore) these mappings:

      _vim.command("call UltiSnips_RestoreInnerKeys()")

Is it possible? Do you see any downsides on this approach or a better solution?

[SirVer (sirver)]

I think it should be possible to do that. Vim users are also used to modes - so they might take to this idea naturally. However I personally would hate to try to use the feauture my muscle memory expects from this key just to realize that it now jumps forward inside - instead of doing that other thing that it does outside of the snippet.

So in general I am not very fond of the idea. If you provide and implementation including good test coverage I will include it into UltiSnips though - it is optional for users - people like myself will just map to keys that are not overloaded, others might use this feature without even realizing.

[mMontu (mmontu)]

"I personally would hate to try to use the feauture my muscle memory expects from this key just to realize that it now jumps forward inside"

I'm actually going the other way around: my muscle memory expects that hitting Ctrl-x Ctrl-k would trigger the default functionality of Vim, and was used to have Tab and Shift+Tab (in snipmate) move between placeholders. But I like the idea of using Ctrl-j and Ctrl-k to do this, because I think Tab is used by too many plugins and by the recommendation on 98a. of https://github.com/MarcWeber/ultisnips/blob/master/doc/UltiSnips-Quickstart.txt.

"If you provide and implementation" - as I mentioned above, I got a partially working implementation, but lack the necessary knowledge of python to understand the plugin flow and insert the _vim.command("call UltiSnips_RestoreInnerKeys()"). I've spent some time by trial and error , but I'm unsure if is there a point of code that is executed every time the outermost snippet expansion ends.

Thus I'll consider learning python or using a different mapping.
Thanks!

[SirVer (sirver)]

> .... https://github.com/MarcWeber/ultisnips/blob/master/doc/UltiSnips-Quickstart.txt.

Urgl. Note that this inofficial fork of UltiSnips has some bugs that the master version has not - And is unsupported by me. Please do not report bugs against this version, instead ask Marc. Or just use master, please.

> some time by trial and error , but I'm unsure if is there a point of code that is executed every time the outermost snippet expansion ends.

this method: https://github.com/SirVer/ultisnips/blob/master/plugin/UltiSnips/__init__.py#L764 is called whenever a snippet exits, as soon as the list is empty, the other most snippet is done too and you should restore the keymappings. Please also add a test to test.py.

[Launchpad Janitor (janitor)]

[Expired for UltiSnips because there has been no activity for 60 days.]

[SirVer (sirver)]

setting to inprogress. The ball is in my court - i need to review the pull request, but I am currently a bit overloaded.

[SirVer (sirver)]

The pull request has been merged.