Lap function #46

New issue

Closed

opened 2026-08-31 17:14:54 +02:00 by robert · 1 comment

robert commented

2026-08-31 17:14:54 +02:00

(Migrated from git.butzei.de)

Goal

Mark segments during a ride.

Acceptance criteria

A button gesture records a lap without disturbing the ride state
Lap distance, time and average speed captured
Brief lap summary shown on the watch, then back to the ride view
Laps sent to the phone and recorded in the GPX
Lap list viewable on the phone after the ride

Files

watchapp/src/c/state.c
companion/.../ride/

## Goal Mark segments during a ride. ## Acceptance criteria - [ ] A button gesture records a lap without disturbing the ride state - [ ] Lap distance, time and average speed captured - [ ] Brief lap summary shown on the watch, then back to the ride view - [ ] Laps sent to the phone and recorded in the GPX - [ ] Lap list viewable on the phone after the ride ## Files - `watchapp/src/c/state.c` - `companion/.../ride/`

robert added a new dependency

2026-09-03 19:01:23 +02:00

#11 Ride state machine: idle / running / paused / stopped

robert added a new dependency

2026-09-03 19:01:23 +02:00

#58 Field value store and page descriptor model

robert referenced this issue from a commit

2026-09-05 18:07:45 +02:00

Watch side of the lap function: double-click Select, CMD_LAP, brief summary (#46)

robert referenced this issue

2026-09-05 18:20:39 +02:00

Lap function: double-click Select, CMD_LAP, brief summary, phone-side in progress (#46) #115

robert referenced this issue from a commit

2026-09-05 18:23:58 +02:00

Companion side of the lap function: split computation, GPX marker, Room laps table (#46)

robert referenced this issue from a commit

2026-09-05 22:16:59 +02:00

Merge pull request 'Lap function: double-click Select, CMD_LAP, brief summary, phone-side in progress (#46)' (#115) from area/lap-function into main

robert commented

2026-09-05 22:17:23 +02:00

Owner

Closed by PR #115 (area/lap-function), built as a genuinely cross-platform split: Vela did the watch side, then handed the phone side to a Kestrel sub-agent on the same branch — a good pattern, kept for future cross-cutting issues.

All five acceptance criteria addressed:

A button gesture records a lap without disturbing the ride state: double-click Select (window_multi_click_subscribe, 300ms window) — chosen after auditing every existing click binding in carousel.c (#60) and finding double-click the one gesture type unclaimed. ride_state_lap_mark() has exactly one observable side effect (queuing RIDE_CMD_LAP on #11's existing dropout-survival queue) and touches none of RideStateSnapshot's four persisted fields or checkpoint_dirty — verified both by a real emulator double-click test and by 7 new host C tests (test_state.c, all passing, including no-op cases for idle/paused/stopped/stop-pending and the queue-full drop-newest contract).
Lap distance, time and average speed captured: LapSplitCalculator (phone-side, host-tested) slices the same cumulative moving-distance/time figures DISTANCE_M/ELAPSED_S already report ride-wide — a genuinely new average-speed formula since none existed anywhere in this codebase to reuse (checked and stated plainly, not silently claimed as a reuse).
Brief lap summary on the watch: a 4-second timed overlay (QUAD template, the four FIELD_LAP_* ids #7 already reserved), dismissible early by any button press, skipped (not queued) if a confirmation prompt happens to be showing.
Laps sent to the phone and recorded in the GPX: wire keys LAP_NUMBER/LAP_DISTANCE_M/LAP_TIME_S/LAP_AVG_SPEED_MMS already existed from #7's original codegen (verified directly in proto.h/Proto.kt/message_keys.json/PROTOCOL.md §2.3 before merging — nothing needed regenerating). GPX representation is a point-level <pp:lap> extension (own reserved-per-RFC-2606 .invalid namespace), not a new <trkseg> — keeps segment boundaries meaning only pause/resume. Honest, tested deviation from D65's literal "closest point" wording: GpxWriter's incremental-flush design means a marker can only attach to the next point written, not retroactively to one already on disk — documented in GpxRecorder's KDoc with a dedicated test.
Lap list viewable on the phone after the ride: new laps Room table (RideHistoryDatabase 1→2, real MIGRATION_1_2, schema JSON hand-diffed against the SQL), expandable per-ride in RideHistoryScreen.

DECISIONS.md D65 records the four real design choices (gesture, split-computation split between watch/phone, summary-as-overlay-not-page, GPX-marker-not-segment) plus the known #4 (real PebbleKit transport) scope gap — nothing calls RideService.recordLap() or a real inbound CMD=LAP handler yet, stated openly rather than faked.

Real verification, not hand-review: ./gradlew :companion:core:test :companion:pebble:testDebugUnitTest :companion:ride:assembleDebug :companion:assembleDebug --rerun-tasks — full genuine rebuild, all green (LapSplitCalculatorTest 8/8, LapHistoryTest 5/5, GpxWriterTest 20/20, GpxRecorderTest 11/11, LapWireEncodingTest 12/12). pebble build clean on gabbro/emery/basalt after a full pebble clean. Host C test_state.c hand-compiled (no cmake in this sandbox) and run directly: 55/55 passing including all 7 new lap tests, -Wall -Wextra -Werror clean. Plus a real emulator double-click verification during design (see D65).

40 issues closed.

Closed by PR #115 (`area/lap-function`), built as a genuinely cross-platform split: Vela did the watch side, then handed the phone side to a Kestrel sub-agent on the same branch — a good pattern, kept for future cross-cutting issues. All five acceptance criteria addressed: - **A button gesture records a lap without disturbing the ride state**: double-click Select (`window_multi_click_subscribe`, 300ms window) — chosen after auditing every existing click binding in `carousel.c` (#60) and finding double-click the one gesture type unclaimed. `ride_state_lap_mark()` has exactly one observable side effect (queuing `RIDE_CMD_LAP` on #11's existing dropout-survival queue) and touches none of `RideStateSnapshot`'s four persisted fields or `checkpoint_dirty` — verified both by a real emulator double-click test and by 7 new host C tests (`test_state.c`, all passing, including no-op cases for idle/paused/stopped/stop-pending and the queue-full drop-newest contract). - **Lap distance, time and average speed captured**: `LapSplitCalculator` (phone-side, host-tested) slices the same cumulative moving-distance/time figures `DISTANCE_M`/`ELAPSED_S` already report ride-wide — a genuinely new average-speed formula since none existed anywhere in this codebase to reuse (checked and stated plainly, not silently claimed as a reuse). - **Brief lap summary on the watch**: a 4-second timed overlay (QUAD template, the four `FIELD_LAP_*` ids #7 already reserved), dismissible early by any button press, skipped (not queued) if a confirmation prompt happens to be showing. - **Laps sent to the phone and recorded in the GPX**: wire keys `LAP_NUMBER`/`LAP_DISTANCE_M`/`LAP_TIME_S`/`LAP_AVG_SPEED_MMS` **already existed from #7's original codegen** (verified directly in `proto.h`/`Proto.kt`/`message_keys.json`/`PROTOCOL.md` §2.3 before merging — nothing needed regenerating). GPX representation is a point-level `<pp:lap>` extension (own reserved-per-RFC-2606 `.invalid` namespace), not a new `<trkseg>` — keeps segment boundaries meaning only pause/resume. Honest, tested deviation from D65's literal "closest point" wording: `GpxWriter`'s incremental-flush design means a marker can only attach to the *next* point written, not retroactively to one already on disk — documented in `GpxRecorder`'s KDoc with a dedicated test. - **Lap list viewable on the phone after the ride**: new `laps` Room table (`RideHistoryDatabase` 1→2, real `MIGRATION_1_2`, schema JSON hand-diffed against the SQL), expandable per-ride in `RideHistoryScreen`. DECISIONS.md D65 records the four real design choices (gesture, split-computation split between watch/phone, summary-as-overlay-not-page, GPX-marker-not-segment) plus the known #4 (real PebbleKit transport) scope gap — nothing calls `RideService.recordLap()` or a real inbound `CMD=LAP` handler yet, stated openly rather than faked. **Real verification, not hand-review**: `./gradlew :companion:core:test :companion:pebble:testDebugUnitTest :companion:ride:assembleDebug :companion:assembleDebug --rerun-tasks` — full genuine rebuild, all green (LapSplitCalculatorTest 8/8, LapHistoryTest 5/5, GpxWriterTest 20/20, GpxRecorderTest 11/11, LapWireEncodingTest 12/12). `pebble build` clean on gabbro/emery/basalt after a full `pebble clean`. Host C `test_state.c` hand-compiled (no cmake in this sandbox) and run directly: 55/55 passing including all 7 new lap tests, `-Wall -Wextra -Werror` clean. Plus a real emulator double-click verification during design (see D65). 40 issues closed.

robert closed this issue

2026-09-05 22:17:23 +02:00

robert referenced this issue

2026-09-05 23:30:19 +02:00

Cue-sheet review screen before riding (issue #55) #118

robert referenced this issue

2026-09-05 23:32:48 +02:00

Cue-sheet review on the phone before riding #55