On 3 Jul 1999, Lars Gullik Bjønnes wrote:
> I know/realise that pseudoactions has served us well, but IMHO they
> are still a hack. Pseudoactions makes us f.ex. not be as type safe as
> we want to.
???!
[...]
> So entities that need to store ac kb_action and (maybe) arguments
> could store a LyXCommand instead of a pseudoaction number.
So all advantages on having pseudo actions will disappear.
> How Pseduactions is better:
>
> Not more that one pseudoaction is generated for each action-args pair.
> However that to can easily be obtained with LyXCommand if we store a
> reference/pointer/iterator instead of the object itself. (the
> LyXCommand objects could be stored in a multi_map or a multi_set)
>
> By using LyXCommand we can get rid of all the special cases that has
> to deal with pseudoactions, and we can change paramters that really
> are kb_actions from int to kb_action. The LyXCommand container will be
> completely separated from the kb_action.
ah, that's what you mean with type safe. Lyxfunc would not accept
but kb_action.
> Opinions?
Your suggestion doesn't fill the advantages of pseudoactions (in space and
speed). It would be simpler to just get ride of all the stuff and always
store a string with full lyx commands (including arguments), and use a
fast lyx command generator (we have it already but could be faster).
Anyway, as the gui independence advances, maybe pseudoactions will not be
used anymore.
Greets,
Alejandro
