Re: extending icomplete -- help?

[Stefan Monnier]

> I'm trying to modify icomplete such as to allow the typing of
> arbitrary substrings for completion, not just beginnings of words.  At
> the current stage I need some more help from people that understand
> minibuffers better than I do.

I don't understand.  icomplete does not change the way completion works.
It only displays extra info about what completion would do if it was invoked.
You code doesn't seem to change this (since it doesn't change any of the
bindings in minibuffer-local-completion-map).  What am I missing ?

For what it's worth, I wrote a complete replacement for the current
completion functions (replacing the C functions with elisp ones)
which includes partial-completion-mode functionality plus substring-matching
as well (and I updated icomplete to know about it).

More specifically, if you type

        foo*bar-baz<TAB>

it will first try to see if "foo*bar?baz" is a prefix of one of the
completions (i.e. it first tries the current completion system,
as opposed to partial-completion-mode which immediately would try
too look for something that matches "\\`foo.*bar[^-]*-baz") if
that fails, it tries to see if something matches "\\`foo.*bar.*-baz",
if that still fails, it tries "foo.*bar.*-baz".

The intention is to replace the current basic completion
functions with this new code (just like I did for newcomment.el).

The code has been working fine for me for a while, but it needs cleaning
up and discussion of how exactly to integrate it.  If anybody is
interested in helping out...


        Stefan