>>>>> "Lars" == Lars Gullik Bjønnes <[EMAIL PROTECTED]> writes:
Lars> handling, and | saw several things: | | - the bindings are not
Lars> saved in a map. Lars did you do this just for
Lars> s/not/now?
Indeed.
Lars> O(log n) instead of O(n).
That's a good reason.
Lars> Ok this can be fixed by providing a more clever comparison
Lars> operator (functor). Was this any different earlier? (before
Lars> kbmap rewrite)
What worries me is that you do not know in which order the binding
will be done (am I wrong?). Why not have a simple queue of
(kmap,function)?
Lars> | - the bindings member creates duplicates templates, for no
Lars> apparent | gain (and I do not see how to `pimpl' it cleanly)
Lars> What template? What do you want to pimpl? kb_keymap?
The map<string,int> in lyxrc. lyxrc.h is included in a lot of
places and most of these places could not care less of this bindings
stuff. In fact, I do not understand why the bindings map (assuming it
actually is useful) is not local to defaultKeyBindings(), which could
do the bind loop by itself. The bindings map does not mean to stay
around forever.
Lars> | - this adds some ugliness in lyx_main.C, that I'd rather see
Lars> hidden in | lyxrc or keymaps.
Lars> What where? Some of this is historical baggage. The problem is
Lars> the global vairables.
I was referring to the loop doing the actual bindings in LyX::LyX.
Lars> | - you do the bindings as early as possible. Is there a reason
Lars> for | that? Can I move at least the deadkey bindings until after
Lars> lyxrc has | been read?
Lars> Why? We want the hardcoded bindings to go in first.
I'd like to render the bindings of X deadkeys to LyX deadkeys optional
(remember the messages from a while ago). This means that I have to
defer this part of the bindings until after lyxrc has been read.
Another soltion would be to do all the default bindings after lyxrc
has been read, but only for keys that have not yet been bound.
JMarc
