bindings: Allow raw escape sequence to be bound with bind

[JoeKar]

This will allow to use e.g. bind "\x1b[D" "lua:comment.comment" from within micro directly. The respective sequence will then be implicitly converted into \u notation when it's stored to the bindings.json.

#2950

[dmaluka]

TryBindKey() is also used from Lua, so this escaping might mess up with Lua's own escaping of its strings. So perhaps it's better to do it in BindCmd(), not here? (i.e. do it only on the UI level, specifically for the bind command)

(I see that, although Lua >=5.2 supports \x hex escaping in its strings, it doesn't work in micro. The reason is that gopher-lua implements only Lua 5.1, as it explicitly states. But it may change in the future...)

[dmaluka]

Also, what about the unbind command?

[JoeKar]

TryBindKey() is also used from Lua, so this escaping might mess up with Lua's own escaping of its strings. So perhaps it's better to do it in BindCmd(), not here?

We can do so, but we've to keep in mind, that the Lua plugins will then most probably have the same "issue" that it can't simply forward "\x1b" as a string, because this will be seen by micro as "\\x1b"...at least I expect so.

Also, what about the unbind command?

Again, you right and I should receive the 🤦. I adapted accordingly with the topping to reflect the unbind in tcell too, but this now is an API extension (test build failed by intention) and thus needs zyedidia/tcell#25 merged first.

[dmaluka]

We can do so, but we've to keep in mind, that the Lua plugins will then most probably have the same "issue" that it can't simply forward "\x1b" as a string, because this will be seen by micro as "\x1b"...at least I expect so.

IIUC, not really. I think what happens is:

1. If we don't do escaping in TryBindKey (i.e. as it works now), Lua plugins cannot use the \x notation at all, i.e. neither "\x1b" nor "\\x1b".
2. If we do escaping in TryBindKey, then Lua plugins can use "\\x1b" but not "\x1b", because this escaping is not at the Lua language level, it's an extra level of escaping, while Lua has its own escaping, i.e. \ is a special character in Lua too, so need to escape it too. IMO this extra level of quoting creates mess and confusion.
3. As I said, the reason why Lua plugins cannot currently use \x is that gopher-lua does not implement it. But it is supported in newer Lua versions. So if we care about it so much, I think we'd better add the \x support to gopher-lua too, it should not be too hard. And then plugins would be able to use "\x1b", not "\\x1b", and also it would be consistent with other Lua implementations.
4. In the meantime Lua plugins can use Lua's decimal \ddd notation, i.e. "\27" for escape sequences.

Anyway, I see you've already moved it from TryBindKey to BindCmd. Just a pedantic note: you did it in the next (unrelated) commit, which is confusing.

Regarding unbind: thanks, I didn't realize it requires unregistering raw sequences in tcell as well, I thought we only need to add parsing "\x1b" to UnbindCmd as well.

BTW, side note: unbind is currently buggy: unbinding does not take effect immediately, i.e. it only removes the binding from bindings.json, so it only takes effect after micro restart, when micro reads bindings.json again. (This is a regression, but this regression has been there for like 3 years (probably starting from d33c28e, which made BufUnmap() do nothing). I was always about to fix it, but it has always been low priority...)

[dmaluka]

Shit, there is also the showkey command...

(If fix it too, it's time for a tiny little wrapper function for strings.ReplaceAll(args[0], "\\x1b", "\x1b") to avoid duplicating it in 3 places?..)

[JoeKar]

Anyway, I see you've already moved it from TryBindKey to BindCmd. Just a pedantic note: you did it in the next (unrelated) commit, which is confusing.

Done.

Shit, there is also the showkey command...

Now you're using the right words! 😆
Especially since it was just one function above the changed ones. Never mind, it's fixed too.

BTW, side note: unbind is currently buggy

Hm, sounds like a good moment to fix this as well, but doesn't seem to be a one liner.

[dmaluka]

Hm, sounds like a good moment to fix this as well, but doesn't seem, to be a one liner.

Yep, and I'm afraid that fixing it properly requires a good grasp of the overall micro's keybindings implementation (KeyTree and all that stuff), which is quite intricate. So don't feel obliged to do it.

[dmaluka]

zyedidia/tcell#25 got merged, so now in order to merge this PR, need to bump tcell version in go.mod and go.sum.

[dmaluka]

zyedidia/tcell#25 got merged, so now in order to merge this PR, need to bump tcell version in go.mod and go.sum.

Sorry, you've already done it...

[JoeKar]

Removed the tcell bump commit within this PR, since @zyedidia did it by his own with 68d88b5.