Automate transferring fit-Files from Garmin Connect to ZwiftPower for dual-recording analysis
- Python3
- Python packages: cloudscraper, fitparse, garminconnect, lxml, numpy, pandas, requests
Install requirements via pip3 install -r requirements.txt.
The script is simply run by executing it: ./dual-recording-gc.py or python3 dual-recording-gc.py.
dual-recording-gc.py first connects to ZwiftPower, collecting a list of all Zwift activities ZwiftPower knows about and selects the latest one. Next, the script connects to Garmin Connect and attempts to find and download an activity with the same date as the activity on ZwiftPower. Next, dual-recording-gc.py reads the two fit-files and attempts to align the power curves1 and to identify the power source used. Next, the fit-file downloaded from Garmin Connect is uploaded to ZwiftPower. Both ZwiftPower and Garmin Connect fit files power source names are added. Subsequently, the time offset of the Garmin Connect fit-file is adjusted to the previously calculated alignment value.
A configuration file config.json is used to store login-data and powermeter identification. The file config.json.template can be used as a blueprint.
To automate dual-recording analysis, the script has to log on to ZwiftPower and Garmin Connect. Login data are configured under the garminconnect and zwiftpower entries. Username and password are configured using their respective fields.
The script also tries to identify the power source used by analyzing the fit-files. This is what the powermeters-section in the configuration file is for. Each entry in the powermeters-section contains a name field and a list of matches that must be met in order to identify a powermeter. The name-field contains the name that should appear on ZwiftPower. Each key-value pair in the matches-section has to match entries in a device_info-section of the fit-file. All entries have to match in a single device_info-section for a positive identifcation to occur.
As far as I could tell fit-files from Zwift do not contain any information to identify connected devices. Hence, the powermeter can only be identified by finding the manufacturer-entry with the value zwift in a device_info structure. If a manufacturer-entry with value zwift is found, then the value from the name-entry is used for the fit-file from Zwift.
Fit-files from Garmin have more information. From the file config.json.template the script tries to match manufacturer:garmin, garmin_product:2787 (Product ID for Garmin Vector 3andserial_number:3999631966(redundant, as example only). You can analyze your fit-files usingfitdump-utility from the fitparse` python package or any other online fit-file analyzer to find matching criteria for your powermeter.
Activities might not match correctly. Matching by name only is insufficient, because multiple activities might have the same name, for example by doing the same workout multiple times. To alleviate the situation somewhat, the scipt tries also to match the date of the activity. This might fail as false-negative and false-positive. Issues arise for example when event dates do not exactly match activity dates, for example when groups (A, B, …) start in intervals and therefor exact times do not match, only dates. Matching dates in not exact either, for example when an event start at exactly midnight, but a groups starts a couple of minutes after event start. In this case not even dates match.
These issues arise when trying to determine if an activity has already dual-recoding data uploaded with it and if dual-recording data should be associated with a Zwift event.
- Add fallbacks if passwords are not found in config.json:
- Keychain/password manager
- Console promp
- Automate dual-recording upload for any ZwiftPower entry, not just the latest one
- Add another (fallback) backend to connect to Garmin Connect (for example gcexport).
Footnotes
-
Aligning the offsets can be necessary, for example, when Garmin devices don't have a GPS-lock for an extended period of time. Garmin devices update their time only by GPS-time, never over NTP, even when WiFi or Bluetooth is enabled and connected. ↩