This is my zmk-config generator, it is written in TS and is located under ./yanConfig/index.ts
Basic layers, are ok to write by hand, but tap dances are just horrible in both ZMK and QMK, my config generator of ~500 lines in JavaScript generates .keymap ~5000 lines long. How am I supposed to maintain it manually?
- Mirror layer
- Create layer from another, but with modifier
- Easily define macros, tap-dance etc
- Combos are defined per layer and are human readable.
- layers defined using their names, and if I ever need to rearrange layer I would not have to kill myself
I.e. tap-dance for key like 'E,LG(E),LS(LG(E))' means emit E on press, LG(E) on hold and LS(LG(E
8973
)) on tap-and-hold.
The declaration '+L' means emit L on tap and LG(L) on hold.
Overall it allows to keep full layer worth of keys with tap-dances on a single laptop screen
| Button like this means it fires E on tap, CMD-E on hold, and SFT-CMD-E on tap-and-hold. Allows to easily fire modifiers with one finger. | |
This means that if I tap simultaneously R and X the cut command will fire. |
You can see full keymap here
Tap-dances on both ZMK and QMK are a mess. I'm not blaming developers, they did an awesome job, I just state the fact that the API is super convoluted and misleading. Especially timings. Consider this: You have a button with 4 functions.
- tap emits 1
- hold emits 2
- tap-n-hold emits 3
- double-tap emits 4
And you have a timeout set to 100ms to trigger from one state to another You'd expect that:
- press and release less than 100ms will emit 1
- press and hold longer than 100ms will emit 2
- press wait less than 100ms, release, press again for longer than 100ms will emit 3
- press wait less than 100ms, release, press wait less than 100ms, release will emit 4
Your expectations are wrong. Due to the nested nature of tap-dance and tap-n-hold functions their timeouts overlap and subtract and I encourage you to set timeouts to some ridiculous high number like 5000ms and observe for yourself actual timings with a stopwatch.
This is what I saw:
- press and release less than 200ms will emit 1
- press and hold longer than 200ms will emit 2
- press wait less than 200ms, release, press again for longer than 100ms will emit 3
- press wait less than 200ms, release, press wait less than 100ms, release will emit 4
That makes setup counter intuitive and adjusting timings very hard.
One more wired (but sensible for developers) default is that the hold key is repeated and not clicked just once.
That makes tap-dance for hotkeys scary to use. Imagine I have a tap to Q and hold to CMD+Q button.
Hold it a bit too long and it will start firing CMD+Qs like crazy, closing all the apps.
What I actually want is to emit CMD+Q only once regardless of how long I keep the key pressed.
Same for opening new tabs, etc.
Essentially hold keys are usually never need to be repeated, but they are hard to not repeat without explicit fix.
var keymap = {
default: {
keys: [
['Q,LG(SLASH),LA(LG(Q))', 'W,LG(W),LG(Q)', 'F,LG(F),LA(LG(F))', 'P,LG(P),LS(LG(P))', '&mo config'],
['+A', '+R', '+S', '+T', 'G,LG(G),LG(LS(G))'],
['Z', 'X', 'C', '+D', 'B,LG(B),LG(LS(B))'],
['&mo windows', '&mo arrows', '&mo numbers'],
['&mo symbols', '&mo default_mirror', '&mo colemak_shift'],
['&mo config', '+L', '+U', '+Y', 'N1,N2,N3'],
['+M', 'N,LG(N),LG(LS(N))', 'E,LG(E),LS(LG(E))', 'I,LG(I),LG(LA(I))', '+O'],
['+J', '+H', '+V', '+K', '&none'],
['SPACE', '&mo symbols', '&mo colemak_shift'],
['&mo numbersf', '&mo numbers', '&globeCaps'],
],
sensor: '&yan_encoder',
combos: [
{ keys: [8, 13], binding: 'LG(V),LG(LS(V))' },
{ keys: [7, 12], binding: 'LG(C)' },
{ keys: [6, 11], binding: 'LG(X)' },
],
},
...