Re: What is the most useful potential feature which Emacs lacks? A: Autocompletion

[Dmitry Gutov]

On 03.06.2020 16:59, Konstantin Kharlamov wrote:

> I am not sure why you say it seems wasteful. Do you mean perhaps, as
> opposed to setting, say, 100ms?

Or 50ms, maybe.

Wasteful of CPU cycles and laptop battery, I'd say. Of course, the exact 
impact depends on the size of your project.

> 100ms I think is the top limit this
> timeout should be set.

The defaults are not set in stone, we can discuss changing them on the 
company issue tracker. But it's a balancing act. Set the value too small 
-- and the result is more likely to annoy people who don't rely on the 
popup as much.

> I just tested how quickly I can type a string
> "prog". I fired up `libinput debug-events` and tried to type
> "prog". The letter "g" says "+0.256s", i.e. it took 256ms. This means
> even if I set to 100ms, there's high chance I won't get any
> autocompletion.

Should I take this to mean the completion request itself took 156ms to 
finish?

> > To be 100% sure, you should try it yourself (I don't do C/C++).
> > Maybe
> > someone else here can testify, though.
>
> Thank you! Indeed, I confirm this does seem to work with an async
> backend. I tested it as follows:
>
> 1. Opened a test.cpp file, enabled company-irony, checked that
>     completion works.
> 2. I set `(setq company-idle-delay 0)`
> 3. I paused the irony-server process with `kill -SIGSTOP $(pidof
>     irony-server)`
> 4. I tried typing a gibberish to see if I get any delay in rendering a
>     text.
>
> I don't see any lags, so I assume using an async backend with the
> timeout set to 0 should work fine.

Indeed, that's what asynchronous means. But the quality of the user 
experience also depends on how quickly the backing server can handle 
those requests.

> This is great news! I wonder if
> company mode should default to zero or so timeout, and print a warning
> if somebody tries to connect a non-async backend?

The majority of backends are synchronous. And the "standard" completion 
API for Emacs (which we want to integrate with) still only supports the 
synchronous convention.