Fix database error that means UI was not able to write.
Simplification and maintenance - removed support for downloading HR data from Garmin (too difficult to maintain for the reward) and updated to the latest React (hopefully removing a bunch of npm vulnerabilities).
Lots of small fixes and an updated FIT file parser (generated from the latest ANT specs). Simplified and updated install / usage instructions.
Adding some graphics to web. Maps + plots. Also, easy definition of sectors, which are carried across versions.
Started exploiting PostGIS: climbs are recognised across activities; common sectors are identified via clustering. Display of climbs and sectors includes sparklines and thumbnails. Some consolidation of Docker scripts so that we focus on common use cases.
Restored some user support thanks to Chris Kelly. Database now using PostGIS, but no new user-visible features. Continuing with simplification of the system (dead code pruned).
Giving up (for a time) on user support as I transition to some kind of web-hosted solution. See updated README. Changes to code include more simplification (especially in loading data) and more consistent use of Docker images (including Jupyter and correct user IDs).
Docker and PostgreSQL! Plus bugfixes and web UI improvements.
Big effort to simplify and clean (10% reduction in LOC). Activity group moved to source and non-null constraints added to statistics. Names and titles are separate for statistics; naming in general has been regularized (eg in search, constants). API for querying statistics improved (simpler and faster). Also many small improvements and fixes.
Schema for statistic names changed from textual 'constraint' to foreign key relation on activity group. This cleans up the database, but has some downsides (mainly, that diary topics must have unique names). It also helps the new search work.
Moving towards a more integrated system, building on the modular
components that already exist. A new command, upload copies data to
the permanent subdirectory of the base directory (a new concept,
can be provided to most commands via --base). The activities
command can read from this directory. This allows the user to upload
data via the (extended) web interface and formalizes 'best practice'
use. Another new command, upgrade copies data from a previous
install and replaces the migration scripts.
Warning - versions 0.31.0 to 0.31.8 (inclusive) will NOT correctly update activity diary entries across versions. Please use an earlier or later release.
Changed interval handling so that possibly incorrect values are marked 'dirty' rather than deleted immediately. Added separate activity groups for road and MTB (see default config). Added achievements. Fixed a bunch of bugs introduced in the big changes in the previous release.
Rewrote diary, decoupling model and view, and added web based (React)
view. See ch2 web start. Re-did activity topics (so old data lost
on upgrade, sorry). Added support for 'extended' intervals which
allows statistics on 'all time'.
Major database changes to allow activity topics and decouple model and view in diary (which has also been rewritten). Coverage statistics and scaling of SHRIMP model when HR data incomplete. System data in separate database to allow for commits that don't affect main data.
Small change to database schema (removing activity type from segments). Fixed power calculations. Improved docs.
Small change to database schema. Not user-visible, but requires a new database / minor release. Important: v0.26.0 has incorrect database name. Please use 0.26.1. Sorry :(
Equipment Tracking. Some database changes (timestamp statistics).
Improved support for multiple activities. FF statistics are calculated for all activities (together). Clearer upgrade instructions.
Dropped old notebooks. Use new templates instead (ch2 jupyter list).
HR calculations via data frames. New pipeline class names - use dev/reset.sh (modify to get the import paths correct) to rebuild the database.
New Jupyter template some_activities that allows filtering by statistics.
Modified ch2 fit to use sub-command syntax (so ch2 fit records
instead of ch2 fit --records). This gives more feedback and better
structured help from argparse. Also, recent code has a bunch of fixes
from user feedback, so it seemed worth making a new minor release.
Added composite sources to simplify dependencies between data. Fixes
problem where multiple statistics pipeline runs were needed to get a
stable system. Also changed (much of) Jupyter access to data to
construct DataFrames directly from SQL queries - this should give
faster data access, but required that Statistic type was stored in the
StatisticName table (ie that statistics are of a single type, which
wasn't actually required until now). As always, the database update
script can be found in the migraine directory.
Major rewrite to import pipelines. Now use multiple processes where
possible. Rebuild time for my data, on my machine, cut in half (and
now including power calculations). Experimented with power
calculations, but "advanced" approach, fitting for wind direction and
speed, didn't give good results (and was slow, hence the rewrite).
Migrate existing databases using migraine/sqlp2sqlq.sh (edit the
file, run, and then reload data from FIT files).
Added timestamp to database which improves / extends logic to avoid
un-needed work re-calcualting statistics. Migrate existing databases
using migraine/sqlo2sqlp.sh (edit the file, run, and then reload
data from FIT files).
Diary plots are generated via Jupyter (running an embedded Jupyter server, generating a notebook, and pushing it to the browser) rather than Bokeh. This works round some Bokeh server bugs and serves as a nice intro to Jupyter. Cleaned up a lot of the plotting code, too.
Modified database schema (serial in statistic_journal which makes
time-series logic simpler). Migrate existing databases using
migraine/sqln2sqlo.sh (you will then need to reload data from FIT
files).
Zoom in summary plots (embedded Bokeh server while diary is used).
Automatic addition of elevation and detection of climbs.
Extend ch2 fix-fit functionality (can scan a directory and print
file names of god or bad files). Required a change in parameters -
now you must explicitly add --fix-header and --fix-checksum if you
want to do that.
Parsing of "accumulated" fields in FIT files plus a bunch more fixes thanks to test data from python-fitparse.
Contains a tool to fix corrupt FIT files.
Choochoo has a GUI!!!
Nearby activities and simplified / improved data access in Jupyter.
Impulse calculations. Faster importing and statistics. See Scaled Heart Rate Impulse - SHRIMP
More readable database (using text instead of opaque numerical hashes in a couple of places). Faster database loading of activity and monitor data. Time is now directly present in the statistic journal table, along with all activity and monitor data (no separate data tables). This enables TSS calculation (next version).
Tidied lots of rough corners, improved docs, added examples, download from Garmin Connect. This could probably be used by 3rd parties.
Diary now uses dates (rather than datetimes) and is timezone aware (Previously all times were UTC datetimes; now data related to the diary - like statistics calculated on daily intervals - use the date and the local timezone to convert to time. So, for example, stats based on monitor data are from your local "day" (midnight to midnight)).
Monitor data from FIT files can be imported.
Major rewrite to generalize the database schema. Moved a lot of configuration into the database. Now much more flexible, but less interactive.