An R wrapper for python-fitparse to read Garmin .fit data files

The python package fitparse seems to be the best-maintained and reasonably current tool for reading the .fit files produced by Garmin devices. Many thanks to the developers/maintainers of that package.

This package was developed to read .fit files from the Edge 820 that replaced my old Edge 800 unit. Several alternative R packages were incapable of dealing with the new format, and quick attempts at patching them were fruitless. In the end it was easiest to spend a couple of evenings learning enough Python to read and format the data, and passing it back to R using the reticulate package. Reticulate had issues passing the resulting dataframes, and the easiest solution was to write then read a JSON tempfile.

All data fields that are single-valued and have an intelligible name are returned. Variables whose names contain "unknown_", "_position", or "phase" are not returned. If your device reports a single-valued field that you want which fits that pattern, or it reports a multi-valued parameter that does not and that throws an error, the code is in readfitffile.py, and you don't need to know much Python to figure out how to make it suit your needs.

You have to have a conda (either anaconda or miniconda) setup that R can find. There are lots of resources to help you do that. Use your search engine. This package requires python 2.X. The environment needs a suitable version on pandas
and numpy. Python 2.7.15, pandas 0.24.1 and numpy 1.15.4 are known to work.
Support for python 2.X ends with reticulate 1.14, so this package may become a hassle to use in the future.

The first call to read a fit file triggers a time-consuming check that the conda environment has the required packages installed, unless a flag to skip it is TRUE.