Oops, send to list this time.
___________________________________________________________
Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail
http://uk.messenger.yahoo.com--- Begin Message ---
--- Tuomo Valkonen <[EMAIL PROTECTED]> wrote:
> On 2005-05-25, David Tweed <[EMAIL PROTECTED]>
> wrote:
> > logicalMovementModule.cc
> > logicalMovementModule.hh
> > logicalRoleModule.cc
> > logicalRoleModule.hh
> > probabilisticMovementModule.cc
> > probabilisticMovementModule.hh
> > probabilisticRoleModule.cc
> > probabilisticRoleModule.hh
>
> Seems quite suitable for "flattened submenus" given
> code that can break
> these names into pieces.
I picked the worst files on my hard disc that I'd tend
to have open at the same time for the example; there's
generally some files with some common elements but not
necessarily in the same for. But the more important
point is that with every open buffer having a window
this typically leads to ten--fifteen entries in a tab,
which means 3-4 characters, which means any rule that
disambiguates things really clearly wrt this human
user's cognitive behaviour is going to have to be
incredibly sophisticated. Hence the interest in a
non-tab based way of navigating things.
> > This sounds an interesting approach. Before I
> start
> > too far down this road, a question for Tuomo: in
> an
> > ion text-input the tab-complete can show
> possibilities
> > every time the tab-key is used. Is this behaviour
> > hardwired to respond to the tab key at a level
> > inaccessible from Lua scripts, or would it be
> possible
> > to write a lua command which displays a list of
> > possibilties, with entries being removed as more
> > letters are typed and they don't match any more
> (ie,
> > without pressing tab to redisplay the current
> > possibilities, just incrementally typing more of
> the
> > string)?
>
> In principle you could implement
> auto-show-completions or auto-completion
> (auto-show-completions plus setting the input field
> to the completion with
> the completed part selected) on the Lua side, but I
> recently started working
> on C-side support for auto-show-completions. It's
> mostly working
> (completions automatically shown as the text is
> modified, tab cycles through
> completions), but the last few days I've been busy
> with looking for a flat
> in another town (and will likely be for some time
> with that and related
> matters) and haven't yet decided how directories and
> full command lines
> should be handled. Currently options for file
> completion are:
>
> 1. Require manual Control+Tab (for example) to show
> completions within a
> directory after a directory was selected from
> list of completions.
> 2. Do not complete trailing slash for directories,
> and require it to be
> typed.
Would these rules equally relate to window titles, or
are they something restricted to FS based queries?
> I also haven't decided how command line completion
> cycling should be
> handled. (Everything I know that has
> auto-show-completions or
> auto-completion only works on simple lists of
> possible completions.) Any
> input on these is welcome. An experimental patch can
> be found at
> <http://iki.fi/tuomov/auto-show-completions.darcs>.
> I do not like
> auto-completion as defined above, btw.
Not sure what you mean by above: presumably my
definition? I was not being explicit about a "cunning
plan" I was thinking about, namely the user types
letters and the completion list removes those entries
which do not have those letters occurring in that
order (but not necessarily consecutively), and then an
<enter> when there's only left calls a function with
that single remaining name as an argument. So, for
example "lRMc" (or even "lRc") would be one
combination which would select
"logicalMovementModule.cc", and this seems to have the
benefits of both reducing the amount of typing whilst
also being easy to figure out in your head.
(Also, I dislike using tab for completions because, in
addition to being strictly unnecessary, it also
requires a "stretched-jab by the little finger of the
left hand" which I really don't like on ergonomic
grounds. I know the same problem exists with the
<enter>/<shift> keys, but at least they are doing
something that's necessary.) Whilst this looks at
first glance like a good interface, I don't want to go
too far down this road if it's not something that's
doable purely from the scripting side.
Anyway, thanks for the ideas,
cheers, dave
___________________________________________________________
Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail
http://uk.messenger.yahoo.com
--- End Message ---
