VI mode <M-...> bindings

[mkoskar]

I would like to have VI insert mode bindings for <M-...> where by <M-> I don't mean traditional 8bit meta but escape prefixed sequence. E.g., <M-h> would be ^]h that is Escape + h.

So I did:

registry.add_binding(Keys.Escape, 'h', filter=insert_mode)(...)

Now that works ok, but it seems to me prompt_toolkit doesn't have a notion of timeout for when there are multiple bindings for so far ambiguous input, i.e., with above in place I can't go to command mode just by pressing Escape. And even if there is no conflicting binding, there is delay (actually as I've figured out it's the eventloop's INPUT_TIMEOUT of 1s by default).

I think we need at least one timeout (see keyseq-timeout in readline) that can be set to something relatively low, in my case I set it to 10ms. You might also have a look how vim does it, it has ttimeoutlen for keycodes, and then separate timeoutlen for mappings.

[jonathanslenders]

Hi @mkoskar,

Thanks for opening this issue! This is interesting.

I didn't know about ttimeoutlen. On my system it's 100ms by default. Prompt_toolkit 2.0 (see 2.0 branch) uses 500ms. Maybe we can decrease this. Maybe 200ms or 300ms, but not too much, because Python can be a little slower than C, and we don't want to call it too early.

Further, prompt_toolkit indeed doesn't support timeoutlen yet. I had no idea that Vim has this option. In my setup it's 1000ms for this one. I think we can do the same thing here.

This feature request makes sense. I can probably implement this in the 2.0 branch.

Jonathan

[jonathanslenders]

Hi @mkoskar,

Here you go: faa202f
It's on the 2.0 branch. Hope that works for you.

FYI: we would like to release 2.0 in the coming months.

[jonathanslenders]

Closing this, given that it's implemented. Both timeoutlen and ttimeoutlen are now attributes of the Application object. Feel free to open again.