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

[Drew Adams <drew.adams@oracle.com>]

> I don't know if I'm misreading it, but from what I read about
> this mode, it seems to only provide completions for minibuffer
> commands. Undoubtedly useful feature, and for this usecase I already
> use ido-mode with fuzzy matches. But in my original text I was rather
> referring to autocompletion when you write code/text, i.e. in buffers
> different from the minibuffer.

You did not misread it.  I should have made that
clear.  But I skipped reading the thread, so I
didn't realize that's all the question is about.

Icicles does support buffer-text completion here
and there, but as you say, most of what's involved
is about completion using the minibuffer -
`completing-read', `read-file-name', etc.

Even the (limited) Icicles support for buffer-text
completion uses the minibuffer as long as there
are multiple matching candidates, however.  So in
those cases you can take advantage of the usual
Icicles features.

You can think of such cases as popping up a list
of autocomplete choices.  The list is shown in
`*Completions*' though, not in your usual menu
form.

This the doc topic that talks about Icicles
buffer-text completion:

https://www.emacswiki.org/emacs/Icicles_-_Completion_in_Other_Buffers

Yes, this is a major lacuna.  I haven't bothered
to provide a general Icicles solution to buffer-text
completion based on, say, `completion-at-point'.

But if I did get around to doing that I'd no doubt
tie in the usual Icicles completion behavior, as
I did for those few special cases, but in a more
general way.

IOW, you'd still use the minibuffer when multiple
candidates match both (a) the buffer text to be
completed and (b) any minibuffer input.  You'd
still be able to progressively complete with
multiple input patterns.  You'd still be able to
sort and change sort orders on the fly.  And so on.