Re: inputting characters by hexadigit

[Ted Zlatanov]

On Thu, 24 Jul 2008 08:55:03 -0700 "Drew Adams" <address@hidden> wrote: 

DA> Speaking only for Icicles, what you type in the minibuffer automatically
DA> (incrementally) filters the candidates (by default, at least), so I don't 
see
DA> the problem you mention.

DA> It sounds like you want to be able to also filter on category, that is, 
filter
DA> on the pair: category + name-in-category. For that, in Icicles, you can use 
a
DA> multi-completion that includes the category.

DA> What you described was actually, IIUC, to *first* filter on category, *then*
DA> filter on name-in-category. That too is possible in Icicles, but filtering 
on
DA> either or both, as I mentioned above, is better - quicker and more general.

I don't know how all that maps to Icicles and your terminology, but what
I'm proposing is really very simple.  I'll explain it differently:

-> user types `M-x TAB'
level = 0 (this is the deepest category level)
candidates = ankle-beta-copy, ankle-boring-call, 200 others with 5 common first 
words
display: ankle(2) right(100) ring(100) rung(100) wrong(100) zeta(100)

-> user types `a TAB'
level = 1 (but we don't need to display categories!)
candidates = ankle-beta-copy, ankle-boring-call
display: ankle-beta-copy, ankle-boring-call (with `a' highlighted)

-> user types `DEL TAB'
(back to first state)

-> user types `r TAB'
level = 1
candidates = 300 beginning with `r'
display: right(100) ring(100) rung(100)

-> user types `ing-r TAB'
level = 2 (we need to display categories, too many candidates)
candidates = 85 beginning with `ring-r'
display: ring-rock(20) ring-ring(25) ring-roar(40)

Basically at any point if there are less than N candidates, just show
the list.  Otherwise, generate the categories for N candidates by
constructing a tree of words in the current candidates' names (split the
names according to the word syntax).  Now, prune the top node (making
its child the top) until the top has more than 1 child.  Then the top's
children are the available categories (but they must be displayed with
the prefixes we pruned earlier).  I think it's a very simple algorithm.

DA> In Icicles, you can also filter progressively, using multiple patterns, so 
you
DA> can easily filter first with category, then with name-in-category, if you 
like,
DA> using separate patterns. But you need not do things in that order, and you 
need
DA> not even treat the two separately at all.

I don't think Icicles is the right way to implement what I'm describing,
because it does a lot of other things which are fairly intrusive to the
user who like the normal completion behavior.

On Thu, 24 Jul 2008 10:00:15 -0700 "Drew Adams" <address@hidden> wrote: 

DA> So now we're going to muck with existing completion mechanisms even
DA> more, trying out new features, just for this Unicode character input
DA> thing?

It's useful, as I've explained, for any list with lots of members,
e.g. the candidates shown for M-x TAB.  I have no opinion on the other
issues you brought up, and I don't need this feature in Emacs itself
unless it requires changes at the C level.  But I do think it would be
useful, especially on a text terminal, where scrolling through the list
can be pretty painful.

Ted