quczer/fit-creator

Python (and C#) project for scraping ZWIFT workouts and converting them to FIT Garmin workouts

1. Installation

Using vscode devcontainers

Ctrl+Shift+P > Open Folder in container

Manual

1. Install any C# executable and add dotnet/Dynastream.FIT.Portable.dll dependency
2. Install fit_creator python package e.g. using pip with

   pip install .
   

2. Basic usage

Just shell-execute

run.sh

It does three steps:

1. Download all available html workout pages from whatsonzwift.com/workouts and saves them in data/html/zwift directory (somewhat time consuming - the internet is the bottleneck).
2. Scrapes relevant info from each download page and saves workouts in .wkt format in data/workouts/wkt directory. (for all workouts, fast)
3. Converts workouts from Build Me Up plan from data/workouts/wkt/Build_Me_Up from .wkt to .fit format and saves them in data/workouts/fit directory. This is veeeery time consuming. The work is done via dotnet run and it's C# code executed there - that's why I do this only for one (my favourite) plan. You can just run the last python command in run.sh with your favourite plan in place of Build_Me_Up.

Time: (downloading excluded)
real 2m44.838s
user 2m59.810s
sys 0m11.321s

3. Interesting stuff

I wrote my own serialization decorator.

@serializable
@dataclass
class Workout:
    name: str
    steps: list[WorkoutStep | WorkoutStepRepeat]
    created: datetime = datetime.now()

It solves the problem of flat deserializations performed by @dataclasses_json and maybe other libraries, as they don't seem to keep any information about the origin class. Therefore one can infer if a particular item in steps is WorkoutStep or WorkoutStepRepeat. The decorator adds 4 methods serialize, deserialize, load and save to your dataclass object and is easily extensible to support other data types.