Command builder integration

[gchp]

I merged @Crespyl's command builder stuff into a branch on this repo. There's still some work in getting it fully integrated, so I'm opening this ticket to track progress in doing that.

The first thing I see is changing Editor::handle_command to work with command::Command rather than the command enum.

After that, standardmode will need to be updated to use the Command struct rather than enum.

There's a good bit more than that, but I think doing those two things will get us to the place where it is functional. Any improvements & optimisations can be working in then.

[crespyl]

The first thing I see is changing Editor::handle_command to work with command::Command rather than the command enum.

As a transitional step, we can create a shim Editor::Command::TextObjectCommand(command::Command) variant to work with a Editor::handle_textobject_command function. Modes can then be gradually converted to the new system while still working through the Editor::Command interface.

Central to moving to the new system is being able to turn TextObject instances into buffer indexes. This should be slightly easier now that we have the Buffer::chars, Buffer::chars_from(Mark) iterators.

I've been considering making TextObject::Kind a trait, rather than an enum. Doing so would mean we'd have different structs for each kind;

struct Char;
struct Word(Anchor);
struct Line(Anchor);

trait Kind {
    pub fn with_anchor(&self, anchor: Anchor) -> Self;
    pub fn get_anchor(&self) -> Anchor;
    pub fn find_first_index(&self, iter: &mut Iterator<Item=char>) -> usize;
}

Doing this would help to separate matching logic for each kind, as opposed to having something like:

enum Kind { ... }
impl Kind {
    ...
    pub fn find_first_index(&self, iter:&mut Iterator<Item=char>) -> usize {
        match *self {
            Char => ...
            Word(anchor) => ...
            Line(anchor) => ...
            ...
        }
    }
}

I'd hoped to make some progress in this area, but frequent build breakage and a severe lack of spare time has kept me moving a good deal slower than I'd hoped.

[crespyl]

See also #95 and the cmd-builder branch for a rough proof of concept.