Deoplete source

[petobens]

Do you plan to implement (or do you happen to know if there is) a deoplete source for r completion? Thanks!

[kxzk]

Definitely would be nice, especially considering nvim-completion-manager is no longer maintained.

[mschilli87]

Just cross-referencing the relevant issues to avoid the same discussion everywhere:

• Some questions regarding usage jalvesaq/Nvim-R#326
• Completion support for R - nvim-r and/or language server Shougo/deoplete.nvim#745

The second was actually mentioned in the first: jalvesaq/Nvim-R#326 (comment)

The whole discussion started here: jalvesaq/Nvim-R#326 (comment)

And the maintainance situation was already mentioned here: jalvesaq/Nvim-R#326 (comment)

[gaalcaras]

Hi there,

Unfortunately, I had to put ncm-R on the back burner for a while, mainly because I had to switch from R to Python for work, so I rarely used ncm-R myself anymore.

The good news is, I temporarily had to work on a R project last week, and not having proper R completion was... frustrating! So I'm back at it now.

My priority has been to update ncm-R to make it work with ncm2. I'm keeping that on a feature branch for now. Would you mind trying it and telling me if it works for you? If that's the case I'll merge it to the main branch. Please give your feedback in #13.

Now, to your point about deoplete. NCM is pretty cool and hackable, but the abrupt deprecation of ncm in favor of ncm2 obviously broke ncm-R for a while, since I didn't have time to maintain it and to look into ncm2. So I have been thinking about supporting another completion framework. That would allow ncm-R to work even if ncm2 breaks again in the future. Deoplete is obviously larger than ncm-R, but I don't use it myself. I have no idea whether it's actually more stable than ncm2 or whether Shougo sometimes introduces breaking changes. I also don't have a lot of time to study if supporting deoplete will require a lot of work or maintenance, if I should just start something else from scratch to work with Nvim-R and deoplete, or if I should look into some other completion framework.

I could use your input on these topics! What do you think?

[gaalcaras]

FIY, I just merged the feature branch into master, you should use that instead of the now deleted feature branch.

[petobens]

Maybe @Shougo can suggest some way of porting the ncm2 source into deoplete, personally I have no idea how easy/difficult that might be...

[Shougo]

Hm... Current deoplete source interface is pretty simple. You can read the code in deoplete.

Deoplete is obviously larger than ncm-R, but I don't use it myself. I have no idea whether it's actually more stable than ncm2 or whether Shougo sometimes introduces breaking changes.

I keep the compatibility if possible. It is not guaranteed though.

[khughitt]

@Shougo Any suggestions for examples of other custom sources of a similar nature that one might use as a template / to get some insights into interfacing with deoplete?

[Shougo]

You should read deoplete source code before asking me...

https://github.com/Shougo/deoplete.nvim/blob/master/rplugin/python3/deoplete/source/around.py

[khughitt]

Fair enough. I likely won't have time to take a stab at this at least until Fall, but perhaps if no one else has implemented this source by then, I'll take a stab at it.