Improve / disable smart import mapping

[blueyed]

Given (at least) #524, #559 and #567 we should consider disabling this by default and/or improving it (also with regard to finding it).

You can quite easily find it by searching for "import" in :h jedi-vim though.

So, calling out to the reporters (@hjwp @tinruufu @chkur) of the issues mentioned above:
What do you dislike about it? (it it too slow? are you keeping to type ahead?)

I could imagine to skip/abort the process in case the user keeps typing and/or only trigger it after e.g. 50ms after the space has been typed.
What do you think?

I've got used to it myself, but can understand that it's surprising - but would still not say to disable it by default (but without knowing how many like this feature).
Let's have a vote in the next comment.

/cc @aliev as the original inventor of this.. :) (in #454).

[blueyed]

Please vote up if you like the feature in general, down if not.

[blueyed]

Do you have it enabled currently?

[blueyed]

Would you like to have it triggered after a configurable timeout of e.g. 50ms by default, so that it would get skipped when typing the i directly after the space?

[blueyed]

I think it will be easier to implement using a callback in Neovim (since we're using a hack via 'updatetime' for g:jedi#show_call_signatures_delay already), so here's another question for up/down voting:
Are you using Neovim?

[blueyed]

Is having the same timeout as with g:jedi#show_call_signatures_delay OK (which is 500ms) by default?
This setting could be renamed then and be applied for both things.

[hjwp]

Hi there, to answer the first question, the reason I wanted to switch it off is I couldnt get used to it -- I did try (not very hard admittedly) to train myself, but I always found myself typing the "import" keyword, and finding it appear twice.

[hjwp]

regarding the idea of a short delay before the completion starts. As someone who doesn't want the feature, as a default behaviour i think that would solve my problem, but: wouldn't that make the experience worse for the people who do like the feature?

In that respect, maybe it's unlike call signatures? Because that's something I sometimes want to see and sometimes don't want, so I don't mind waiting a bit if, as I'm coding, just now is one of the times i do want to see it. but the import keyword autocompletion, i think (maybe?) that's something you either want or don't want, forever.

[hjwp]

In any case, I'm very happy now that I've got the feature switched off. My docs suggestion for improvement would be: rename that section in the docs? and/or the name of the feature? to me, smart_auto_mappings doesn't immediately bring to mind "ah yes, this is the feature that autocompletes the word "import""

[davidhalter]

Thanks for bringing this up! I'm actually also undecided if it's a great idea. I've gotten used to it, but it still has that kind of strange feeling where when you write import after it, it's just a duplicate. Of course you can train to not use it, but it's just not the same behavior that a normal IDE has.

[tinruufu]

i'm okay with the feature, i just think it's a bad default and not in keeping with the rest of the stuff this project does for me

[blueyed]

For reference: it's also problematic with iabbr: #495 (comment): if <space> is mapped it seems to not expand abbreviations anymore.
Not sure if we could somehow detect and skip/simulate it.

[davidhalter]

Well, I would also be fine with another default. That would probably cause less issues.

I have realized that smart_auto_mappings will not contain as much stuff as I would have liked to. There's only very few examples where auto completion allows you to add a keyword just right after a space (and most of them are difficult to handle without a lot of gain). The thing that would be really cool is adding () to function calls, but that won't work either, because you also need to ignore the closing ) and opening ( if the user types it. That's just a nightmare to do in VIM.