Re: Why fido, icycles, ido, icomplete

[Ergus]

Hi Joao:

Thanks for the answer it is very clarifying for me now. Maybe you should
add all this information somewhere in the documentation.

On Wed, Nov 06, 2019 at 10:03:20PM +0000, Jo�o T�vora wrote:
> Ergus <address@hidden> writes:
>
> > 1) Do we really need another confusing mode (fido-mode)? instead of
> > improve one of the already existing alternatives with just an option to
> > use flex?
>
> I'm sorry you find the name confusing.  I find it appealing (it's "fake
> ido"), but we can change it (Stefan proposed icomplete-ido-mode).
The name is not important for me actually :p if it is documented
properly and in a logical place in the manual with links (or in the same
place) of the ido section and icomplete... so easily to find.  

> Now, I wish I could just put 'flex' (and many other things) in ido-mode.
> Actually the matter is much more complicated.  Ido mode is a completely
> separate completion system that doesn't respect Emacs completion
> interfaces (completion-in-region-functions,
> completion-at-point-functions, completion-styles, etc etc etc ).  This
> means it doesn't work nicely for M-x, C-h f, and many many other
> completion situations.  So, believe me, I tried to change it... and gave
> up immediately (but go ahead, open lisp/ido.el hehehe).
I actually have very strong feelings behind ido in 2019 (I know I am a
sort of apostate for this). But I think it is something that needs to be
removed/deprecated/substituted for the good of newer alternatives like
icomplete. So new users will try more maintained/ modern/ powerful/
better integrated alternatives: like icomplete/ivy/helm.

The intention is to move the users to the newer functionalities so they
can get the best possible first impression.

From the software point of view it is "complex" to keep such a big piece
of code that nobody wants to touch anymore... specially if we already
have alternatives for it.

I think Abo-abo actually tried to modify ido to improve it and he
finally ended implementing ivy... was easier that way.

> Recently, we worked together with the Helm developer to help Helm to
> abide by a greater number of Emacs completion interfaces.  It should
> have, among other benefits, the interesting result that you can use Helm
> matching styles in icomplete.el's frontend.  Or company's.  Or Emacs's
> flex in Helm.

I will pray you to do the same for ivy... please please... Actually I
think ivy is now much better integrated than helm before, but for sure
there will be things missing you could help to improve.

> So I don't think it's true there's no cooperation.
Sorry, maybe I explained wrong... I know there is some cooperation
(actually some contributors here are active contributors of ivy for
example). I just think it is not enough. The argument is actually that
spacemacs is so big just because they need to hack or re-implement many
functionalities that sometimes we can provide with an extra parameter in
vanilla emacs.

> Maybe you can convince someone to do that for ido.el, idk.
Again, I am not an ido fan... I would try to convince someone to do the
same for avy-counsel actually.

> I was a heavy Ido addict for many years because it has the best
> interface, hands down, much better than Ivy, or Helm, IMO of course.  I
> also know many people in my company that swear by it.  But the
> implementation was impossibly hard to hack on.  Then I found
> icomplete.el, with very lean implementation (around 8x less lines) and
> the closest gap to ido-mode.  I improved a decent part of icomplete.el,
> so Icomplete mode is now better by itself.  But that gap to ido-mode is
> still annoyingly (and legitimately) there, and we can't just change
> icomplete-mode's defaults like that.
I have never used icomplete... so I don't know what ido provides that
icomplete can't. So where is the gap? Is a part of the gap fixed in helm
or ivy for example?

> So fido-mode.  It's really just icomplete with slightly different
> defaults that emulate ido mode as close as possible.  Indeed the
> specific contract of fido-mode is to emulate ido-mode as much as
> possible, to use it as a reference, so if you find something in the
> emulation that can be improved, please share.

Maybe this paragraph should go in the manual in the ido section
suggesting to switch to fido in order to improve fido as much as
possible and deprecate the actual ido implementation in the future... (I
have a dream, please don't burn me for this "A man can dream... a man
can dream")

> Regarding your other points, <shrug> that's just the way Emacs is.  I
> like lean implementations that play along with the rest of Emacs, so I
> chose icomplete.el to use and to hack on.
Me too, now I get why we need fido... 

It should say explicitly in the documentation that "Fido is a modern
icomplete emulating ido with better integration with all the emacs
infrastructure" etc etc etc... and it. 

> Jo�o
Best,
Ergus