* Chordal Hold: restrict what chords settle as hold
* Chordal Hold: docs and further improvements
* Fix formatting.
* Doc rewording and minor edit.
* Support Chordal Hold of multiple tap-hold keys.
* Fix formatting.
* Simplification and additional test.
* Fix formatting.
* Tighten tests.
* Add test two_mod_taps_same_hand_hold_til_timeout.
* Revise handing of pairs of tap-hold keys.
* Generate a default chordal_hold_layout.
* Document chordal_hold_handedness().
* Add license notice to new and branched files in PR.
* Add `tapping.chordal_hold` property for info.json.
* Update docs/reference_info_json.md
* Revise "hand" jsonschema.
* Chordal Hold: Improved layout handedness heuristic.
This commit improves the heuristic used in generate-keyboard-c for
inferring key handedness from keyboard.json geometry data.
Heuristic summary:
1. If the layout is symmetric (e.g. most split keyboards), guess the
handedness based on the sign of (x - layout_x_midpoint).
2. Otherwise, if the layout has a key of >=6u width, it is probably the
spacebar. Form a dividing line through the spacebar, nearly vertical
but with a slight angle to follow typical row stagger.
3. Otherwise, assume handedness based on the widest horizontal
separation.
I have tested this strategy on a couple dozen keyboards and found it to
work reliably.
* Use Optional instead of `| None`.
* Refactor to avoid lambdas.
* Remove trailing comma in chordal_hold_layout.
* Minor docs edits.
* Revise to allow combining multiple same-hand mods.
This commit revises Chordal Hold as described in discussion in
https://github.com/qmk/qmk_firmware/pull/24560#discussion_r1894655238
1. In "RCTL_T(KC_A)↓, RSFT_T(KC_C)↓, RCTL_T(KC_A)↑" before the tapping
term, RCTL_T(KC_A) is settled as tapped.
2. In "RCTL_T(KC_A)↓, RSFT_T(KC_C)↓, RSFT_T(KC_C)↑", both RCTL_T(KC_A)
and RSFT_T(KC_C) are settled as tapped.
3. In "RCTL_T(KC_A)↓, RSFT_T(KC_C)↓, KC_U↓" (all keys on the same side),
both RCTL_T(KC_A) and RSFT_T(KC_C) are settled as tapped.
4. In "RCTL_T(KC_A)↓, RSFT_T(KC_C)↓, LSFT_T(KC_T)↓", with the third key
on the other side, we allow Permissive Hold or Hold On Other Keypress
to decide how/when to settle the keys.
5. In "RCTL_T(KC_A)↓, RSFT_T(KC_C)↓" held until the tapping term, the
keys are settled as held.
1–3 provide same-hand roll protection. 4–5 are for combining multiple
same-hand modifiers.
I've updated the unit tests and have been running it on my keyboard, for
a few hours so far, and all seems good. I really like this scheme. It
allows combining same-side mods, yet it also has roll protection on
streaks. For me, this feels like Achordion, but clearly better streak
handling and improved responsiveness.
* Fix formatting.
* Add a couple tests with LT keys.
* Remove stale use of CHORDAL_HOLD_LAYOUT.
* Fix misspelling lastest -> latest
* Handling tweak for LTs and tests.
* Fix formatting.
* More tests with LT keys.
* Fix formatting.
* Replace Tapping Force Hold feature with Quick Tap Term
* Replace keyboard level TAPPING_FORCE_HOLD with QUICK_TAP_TERM 0
* Deprecate force hold in info_config.json
* Before and after quick tap term unit tests
* Quick tap unit tests iteration
* Keymap config.h correction
* Remove TAPPING_FORCE_HOLD_PER_KEY macros that were missed
* Add two more test cases for quick tap
* Replace TAPPING_FORCE_HOLD with QUICK_TAP_TERM in configs #2
* Replace TAPPING_FORCE_HOLD_PER_KEY with QUICK_TAP_TERM_PER_KEY in configs #2
* Add function declaration for get_quick_tap_term
Co-authored-by: Stefan Kerkmann <karlk90@pm.me>
* Fix Caps Word and Unicode Map
* Tests for Caps Word + Auto Shift and Unicode Map.
* Fix formatting
* Add additional keyboard report expectation macros
This commit defines five test utilities, EXPECT_REPORT, EXPECT_UNICODE,
EXPECT_EMPTY_REPORT, EXPECT_ANY_REPORT and EXPECT_NO_REPORT for use with
TestDriver.
EXPECT_REPORT sets a gmock expectation that a given keyboard report will
be sent. For instance,
EXPECT_REPORT(driver, (KC_LSFT, KC_A));
is shorthand for
EXPECT_CALL(driver,
send_keyboard_mock(KeyboardReport(KC_LSFT, KC_A)));
EXPECT_UNICODE sets a gmock expectation that a given Unicode code point
will be sent using UC_LNX input mode. For instance for U+2013,
EXPECT_UNICODE(driver, 0x2013);
expects the sequence of keys:
"Ctrl+Shift+U, 2, 0, 1, 3, space".
EXPECT_EMPTY_REPORT sets a gmock expectation that a given keyboard
report will be sent. For instance
EXPECT_EMPTY_REPORT(driver);
expects a single report without keypresses or modifiers.
EXPECT_ANY_REPORT sets a gmock expectation that a arbitrary keyboard
report will be sent, without matching its contents. For instance
EXPECT_ANY_REPORT(driver).Times(1);
expects a single arbitrary keyboard report will be sent.
EXPECT_NO_REPORT sets a gmock expectation that no keyboard report will
be sent at all.
* Add tap_key() and tap_keys() to TestFixture.
This commit adds a `tap_key(key)` method to TestFixture that taps a
given KeymapKey, optionally with a specified delay between press and
release.
Similarly, the method `tap_keys(key_a, key_b, key_c)` taps a sequence of
KeymapKeys.
* Use EXPECT_REPORT, tap_keys, etc. in most tests.
This commit uses EXPECT_REPORT, EXPECT_UNICODE, EXPECT_EMPTY_REPORT,
EXPECT_NO_REPORT, tap_key() and tap_keys() test utilities from the
previous two commits in most tests. Particularly the EXPECT_REPORT
macro is frequently useful and makes a nice reduction in boilerplate
needed to express many tests.
Co-authored-by: David Kosorin <david@kosorin.net>
* Add per-test keymaps
* Add better trace and info logs for failed unit-tests
* Add layer state assertion with tracing message
* Use individual test binaries configuration options
* Add basic qmk functionality tests
* Add tap hold configurations tests
* Add auto shift tests
Co-authored-by: Nick Brassel <nick@tzarc.org>
