On 04/24/2017 04:33 PM, Daniel Megert
wrote:
> What if a very
valuable processor (returning a 5 minutes of work-worth result)
takes 2
seconds
I'm sorry, I don't get your
point. How
can it take 2 seconds when the work is 5 minutes. LOL
That was indeed not clear, sorry.
The idea was to imagine a content assist processor that takes 2
seconds to compute which operations it can provide at the given
location (2 seconds is too long, it currently gets timed-out by JDT
and proposals are not shown to user), but for completion actions
that are worth 5 minutes of manual work. The net result is that to
save 2 seconds, we hide a 5 minutes-worth feature; whereas it might
be totally acceptable for the user to wait 2 seconds because the
profit is high.
The point I was trying to make is that it's sometimes worth waiting
for the IDE and we shouldn't expect everything to be fast in the
IDE; we need to embrace slowness rather than excluding it, because
slow operations or processing can have results of high value. And in
a fluent UI, embracing slowness is mostly about going asynchronous
as much as possible.
I hope that explains it a bit better.
|