Bindings to keysyms are layout-dependent

[chebykinn]

Hi, I've been using Sway on and off for a while already, but I wanted to understand the reason for making bindings layout-dependent.

Let's say for example a user sets a key binding to open an application menu the same way they did it in i3:

bindsym $mod+d exec rofi -show run

This will only work with English keyboard layout. If they have a setup with multiple layouts like with an additional Cyrillic layout the menu wouldn't be called unless they switch back to English. To avoid this it is suggested to use bindsym --to-code in the man page, but it's not straightforward for the people who come from i3 and I've found multiple related issues here before posting this.

So roughly I have these questions:

• What is the use case for binding keys to layouts? (AFAIK, other OSes, and environments don't do this)
• Can the default be changed?

[chebykinn]

Or maybe this is an incorrect behavior on the i3 side, since they have bindsym and bindcode, the latter is supposed to work independently from the layout, but then I just don't understand the reason for having bindsym.

[emersion]

A keysym (key symbol) doesn't identify a physical key. It identifies the result of translation a key according to a keyboard layout. A keycode identifies a physical key.

One could write a full config with keycodes rather than keysyms, but that's not what bindsym does. But that isn't very user-friendly, e.g. KEY_A is the position of the A key on a US keyboard (second row first key after Tab), and e.g. on a French keyboard it's the Q key.

[chebykinn]

Right, but from my everyday experience on i3 bindsym somehow maps to physical keys, why does it work differently?

I understand now the use case for binding to letters, French layout is a good example, do I understand that you have it as the only layout?

[emersion]

I don't remember what i3 does but probably it does --to-code by default or something. We chose to not do that in Sway because it went against the principle of keysyms.

Some people have French as the only layout, some people have both French and English (personally I only have English).

[chebykinn]

Got it, thanks! But what is the difference between bindsym --to-code vs bindcode?

[emersion]

bindsym --to-code takes keysyms as input and converts them into keycodes according to the default layout. bindcode takes keycodes as input.

bindsym accepts symbol names e.g. "a", but for bindcode you'd need to look up the KEY_* constant in /usr/include/linux/input-event-codes.h and put the number in the config file.

[chebykinn]

I understand now, thanks. I still think that for i3 compatibility it would've been best to keep the same behavior, but I get why you didn't do that. I'll leave this thread open for a while in case someone else would be interested.

[skrattaren]

Respectfully, I don't think it's a good decision. It basically means that default Sway appears* unusable for any non-Latin layout without major sed-style overhaul of config file.

* I mean "appears", because before triaging and investigating the issue keyboard shortcuts simply don't work half of the time.

P.S. I don't think it's i3-compat-related: it's just expected behaviour: "basic shortcuts should work" =)