v0.29.0

[0.29.0] - 2026-04-04

Added

• WorkoutData.merge_many() class method to merge a list of separately recorded workouts into a single continuous WorkoutData object. Workouts are sorted by start timestamp; overlapping timestamps favour the later workout; time gaps are preserved as NaN rows after resampling (when interpolate=False). Accepts drop_gaps=True to shift workouts back-to-back (no gap rows), which condenses elapsed time — note that performance metrics will not account for rest time in this mode.
• WorkoutData.merge() instance method as a convenience wrapper around merge_many(), also accepts drop_gaps.
• distance is forward-filled across gap rows in merged workouts (it is cumulative and should not appear as NaN during rest periods).
• WorkoutData.set_start_time() method to shift the datetime index of a workout to a specified start time, recomputing the time column. Useful when a file (e.g. a Strava activity stream) contains no absolute timestamp.
• Three new Strava activity-stream example files covering a single session split across three recordings: strava_17497731832_warm_up.json, strava_17497955000_track_workout.json, strava_17498168651_warm_down.json.
• Added lab/merge_workouts.ipynb notebook demonstrating merge_many(), merge(), and set_start_time() with the new example files.

Changed

• Updated pyproject.toml build backend from poetry.masonry.api to poetry.core.masonry.api to support editable installs (pip install -e .).