| Commit message (Collapse) | Author | Age |
|---|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
An official climb matched on endpoint coordinates alone. On a loop ridden
the other way the route meets the climb's top before its bottom, so pairing
each start with the first end inside its lap window credited a 5.9 km
passage as the 850 m Chemin de la Tour Pey Blanc, and a 49 km one on
another ride.
Climbs now keep the sample of the segment they were created from: the
recorded length and elevation gain of the selected route points, captured
by the ride detail and preview handlers and stored NOT NULL. Matching takes
the whole route and drops any candidate outside ±50% of that sample on
either axis, and attempts are refreshed by upsert and prune so re-analyzing
a ride keeps the recorded-at stamp of the passages that survive.
The backfill leaves the web package: backfill.Run recomputes ride values
then official climb attempts as a single `go run . backfill` command
(climbs-backfill is gone), rideanalysis owns the attempt snapshot, and the
mountain-pass loading margin is shared. Applied to the stored rides it
stamps 78 samples, drops the four bogus attempts and leaves ride 800 with
its three Pont Rout passages.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
Passes are extracted from OSM PBF nodes (mountain_pass=yes) and linked to
containing ways; highway road values map to road, track to track,
path/footway to path, plus hiking for rock/climbing. All passes are kept,
coloured distinctly (road/track/path/hiking/unknown) for road-bike validation
on ride 397. Single squashed migration adds category column, cheap to re-extract.
|
| | |
|
| |
|
|
|
|
|
|
| |
Context / Problem: Strava titles can be corrected after import. The sync path only handled new rides; existing rides were marked skipped without checking the cheap list metadata, so corrections like ride #789 stayed stale.
What: In syncRides reuse the existing Save write path for rare title resyncs (pay GPX parse cost) when existing.Name != list.Name; track progress.Updated and surface it in SSE/TS UI; add tests for resync without stream fetch. Remove the now-obsolete TODO.
QA: go test/ vet, tsc, biome/stylelint pass.
|
| |
|
|
|
|
|
|
|
| |
Add a docker-compose service running the Grafana Pyroscope backend on
port 4040 with a persistent data volume, and instrument the Go app via
github.com/grafana/pyroscope-go. The app pushes CPU, memory, and
goroutine profiles every 10s when pyroscope.enabled is set in
config.yaml. Config is extended with application_name and server_address
(defaults to http://localhost:4040). Document usage in the README.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Move climb detection, slope, and Cotacol scoring out of ride into a new
climb package that operates on a geometry-only Route (distances, elevations,
coords). Ride embeds climb.Route and keeps its timing and telemetry columns.
- climb.Detect returns []DetectedClimb (Climb + ride-absolute Start).
- Climb owns its sliced geometry; indices are local, with Start reported at
the detection boundary for absolute position.
- Detected bounds use core.Interval; freeRegions claim suppression fixed.
- ride no longer builds slices: ClimbFromDist/ClimbFromIndexes/climbFrom and
ride.Climb/Interval/Range are removed; rideanalysis, mountain_pass MatchClimb,
and web views are rewired to climb types.
- New internal/routetest builds climb.Route from the ride DSL; ridetest reuses
it and adds timestamps and telemetry.
- Tests moved to climb (synthetic detection and difficulty) and kept in ride
for GPX-example integration; ride.comparison detection coverage retained.
- mountain_pass.nearestCrossing now reuses geo.Nearest.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
extract-osm)
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Expose pprof at /debug/pprof via -pprof flag, tag each handler with runtime/pprof route label, add handleRoute helper to avoid repeating patterns, and document UI workflow in README.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Official climb matching now returns all non-overlapping occurrences
per ride instead of only the globally nearest pair. Each lap is
detected by clustering contiguous points within the endpoint radius
and pairing each start cluster with the first end cluster inside its
window, preserving direction. The ride analysis aggregates all matches
and sorts claimed ranges before excluding them from auto-detection.
Fixes the case where a looping ride only showed one instance of a
saved climb while the user expected one per repetition.
|
| |
|
|
|
|
|
| |
Extend analytics with a second scatter plotting Cotacol per 100 km against
moving hours, keep speed versus Cotacol as first chart, render both
canvases from shared points, open ride detail in a new tab on dot click,
and keep all website copy in English.
|
| |
|
|
|
|
|
|
| |
Consolidate persistence (Save, List, Backfill) into ride/store.go alongside
ride.go/climb.go/moving_time.go, rename rides.Ride to ride.StoredRide to
avoid collision with parsed ride.Ride, update all web/cli consumers and
RideView to use named Ride field, adjust templates and regenerate templ files,
and remove rides directory.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
A detected climb must now average at least 3.5% slope with re-segmentation:
gentle drags are split into their qualifying steeper parts instead of being
reported wholesale or dropped. Qualifying climbs keep their previous
boundaries exactly; only regions whose classic best climb averages below the
threshold get re-segmented (full-scan for the best qualifying sub-segment,
found=false when none exists).
QA: A/B across data/rides (362 rides) went from 137 to 133 climbs, 23 rides
changed; the 7 fully-dropped rides were audited to contain zero qualifying
sub-segments. Example-ride tests updated for the new behavior; the resolved
"minimum slope" TODO was removed.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
Replace the centcols CSV download/enrichment pipeline with OSM PBF
extraction as the primary (only) source: merge osmpass into mountain_pass,
drop the centcols table and model fields, filter to named passes with
elevation, and remove the centcols CLI flags.
|
| | |
|
| |
|
|
|
|
|
|
| |
Add ride.MovingTime, a speed-threshold algorithm over track timestamps.
Add a -backtest CLI command comparing computed values against Strava's
reference across all stored rides, with a percentile distribution summary.
Baseline: mean err 3.9%, p50 1.6, p75 4.6, p90 8.4, p95 12.1, p99 44.9.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
