Re: [ELPA] New package: transient

[João Távora]

On Mon, May 4, 2020 at 2:28 AM Dmitry Gutov <address@hidden> wrote:

> I gave some what I see as suboptimal examples. In this case, it's that,
> for example, rot13-string is sorted before string-to-syntax or
> string-prefix-p. But that's easy to fix: just ding all non-prefix
> matches's scores by 1.

Not horrible, but I'd rather boost by "also mentioned in the
manual" than penalize prefixes/suffixes.  Doing so would
demote rot13-string, lgstring-char, ad-docstring, for example.
Doing what you propose can be seen as yet another
way for the completion system to opine on how things should
be named. And that's fundamentally _not_ the problem we're
trying to solve.

Again, as far as I understand we're trying to solve "how do
discover related things" and "how to quickly reach for things
you know exist but don't can't predict/remember the name".
I think substring/flex completion (with those extra sprinkles)
is the  best way to do the latter.  It _can_ serve decently for
the  former, but I personally prefer the manual for that.

> > into the sorting/grouping of completion results. Newbies prefer
> > API lists by topic? Then let's make those pretty lists from the
> > manual, again.
> I'm curious to see how that would look in code.

I think we would have to enhance the @defun macro to
collect the definition into the a structure the organizes
definitions by node current node.  Then dump that information
in some section of the elisp manual. I thought the
@defun macro was under our control, but it's not.
No idea if Texinfo allows redefinition of macros, and in
fact very little idea of how Texinfo works at all, but I
this kind of things doesn't look very hard to do.  Maybe
Eli would like to comment.

João