bug#41531: 27.0.91; Better handle asynchronous eldoc backends

[Dmitry Gutov]

Hi!

On 26.05.2020 17:53, Stefan Monnier wrote:
> > But really: now we have deadlock too?  I just want to solve this
> > problem: please let's commit something, and move on to the next bug.
> Can you use the sample `eldoc-future-*` code I sent earlier?

How about the attached file for a rough, but a largely feature complete 
first version of futures?

To remind from previous discussions, we wanted futures:

- To be cancelable, so that the issuer could abort their calculations, 
if they so desire.
- Force-able, meaning the consumer should be able to "realize" the 
future synchronously, with the future's creator being able to support 
their, most optimal, version of that logic. The default needs to be 
useful and reliable enough, though, that callers could use it in 99% of 
cases anyway.

The error callback probably wasn't mentioned, but it seems logical to 
have it anyway.

For the first two features, I also considered using cl-generic, but 
result might turn out to be clunkier, and we need overridability only 
for two of these functions. But suggestions welcome.

TBD:

- Probably rename to "promises" in the end.
- It would be nice to have a certain degree of compatibility with 
Christopher Wellons's emacs-aio, so that its promises could be accepted 
by code that expects "our" futures/promises. Not sure if we can do that 
without making the API more complex (aio-resolve's signature comes to 
mind), or if we adopt its approach, importing the package wholesale 
might make more sense.

future.el
Description: Text Data