Currently the command i3-msg -t get_tree provides the field scratchpad_state that is useful in identifying windows that are currently shown but still considered to be in the scratchpad. This field is not present/unmarshalled into i3.Node.

I believe adding this line to the Node struct definition would suffice.

ScratchpadState    string           `json:"scratchpad_state"`

It looks like only https://github.com/i3/go-i3/blob/master/socket.go#L32 would need to be abstracted to an environment var with a sane default to i3 if you wanted to support swaywm.

More than happy to PR this if there is any interest.

I'm using this pkg for supporting both i3 and sway in my https://github.com/quite/i3gopher

I just found out that I'd need to get hold of the app_id of a node, which is only available under sway (approx like the window instance in X11). I'm using this pkg because it is so good, and I did from the beginning only support i3. And thanks to #4 it just works out of the box with sway. Only now did I run into the missing app_id for i3gopher's exclude functionality.

Having read the #4 issue I'm thinking that you might be wary of making such a sway-specific addition to this pkg @stapelberg . What do you think? I have not come across the fork of this pkg that @dlasky mentioned, which blanked-renames it to go-sway. But I also think that is not such a great idea.

See e.g.:
https://travis-ci.org/i3/go-i3/builds/530377615?utm_medium=notification&utm_source=email
https://travis-ci.org/i3/go-i3/builds/528495890?utm_medium=notification&utm_source=email
https://travis-ci.org/i3/go-i3/builds/526529017?utm_medium=notification&utm_source=email

It would be nice if the api did not log the detection of the version Variant. Is there a need to communicate that this is unsupported?

I'm writing a small tool with I'd like to be usable both with i3 and sway. https://github.com/quite/i3gopher

Though the tool is supposed to be run in the background, this frequently logged string (on every messaged roundtripped!) will drown any other relevant logging (error or otherwise) that it does.

I see that we can't really return non-nil since that'd stop the variant from being accepted. What to do...

@dlasky what do you think?