GPX to CSV conversion (or to a Python list)

PyPi Project Page

This tool can convert most gpx files to csv format. It preserves trackpoint extensions and trk level tags. It supports multiple <trk> entries per file, multiple <trkseg> segments. It powers gpxrun and my gpx analysis web app.

Advantages / Features

• Preserves most trackpoint data: If a gpx file uses extensions and has hr for heartrate data, it will make a csv with a hr column. No need to ever add specific support for specific new or requested column names.

• Minimal dependencies: Only requires lxml.

• Easy command line usage with wildcards: gpxcsv file.gpx and done. gpxcsv *.gpx just works.

• Preserves and identifies multi-trk or multi-trkseg data: Because trk and trkseg level tags get their own ID columns, multi-track or segment files preserve each segment or track as distinguishable.

• Easily create Pandas dataframe: The gpxtolist function will create a python list for one-line conversation to a dataframe: pd.DataFrame(gpxtolist('myfile.gpx'))

• Supports JSON (even though I named the package gpxcsv): JSON support since it was easy from the list of dictionaries.

Installation and Usage

1. Pip install

pip install gpxcsv

Or clone and install module

git clone https://github.com/astrowonk/gpxcsv.git
cd gpxcsv
pip install .

2. Use directly with command line tool or as a python module. The following examples will create myrun.csv

gpxcsv myrun.gpx
python -m gpxcsv myrun.gpx

Or specify an output file name

gpxcsv myrun.gpx -o myfirstrun.csv

Or, even though I named this gpxcsv, convert to a simple json file:

gpxcsv myrun.gpx --json
python -m gpxcsv myrun.gpx -j
python myrun.gpx -o out.json

3. Use the gpxtolist function to read the gpx file into a python list suitable for conversion into a pandas dataframe in a notebook or iPython.

from gpxcsv import gpxtolist
gpx_list = gpxtolist('myfile.gpx')

#if you have pandas
import pandas as pd
df = pd.DataFrame(gpx_list)

Release Notes

0.3.1

Switch to pyproject.toml. Fix a few minor coverage issues. Use pathlib for suffix substitution.

0.2.15

• Fixes issue with trk level extension data like DisplayColor. This will now be added to the csv file with an identifier along with trk name.

0.2.14

• Fixes issue with gpx files that don't have children of the extension, i.e. no TrackPointExtension inside the Extensions tag.

0.2.11

• Adds support for processing a StringIo object, which was necessary to use this code as the backend for a Dash web app, which encodes all files as base64 strings.

0.2.10

• The _try_to_float function no longer crashes when trying to floatify None. This was happening due to some odd xml in Runalyze exported GPX files.

0.2.9

• Changed the way attribs are pulled from trackpoints. Code will not crash if trackpoint is missing lat or lon. (Obviously this shouldn't happen, but occurs in some exported workouts from Apple Watch that are missing GPS data.)

0.2.8

• Fixed a crashing bug because of an unneccessary import accidentally auto-added.

• Added 0.2.7 fixes, which includes the --silent flag and better handling of missing files (skipping processing them rather than an assert error)

Test GPX files

I tested the conversion against a handful of my own GPX files (exported from Apple Health / Apple Watch via the excellent HealthFit app).

In addition I used several files from this sample-gpx repository, specifically:

• Alt_Portsmouth.gpx
• MoselradwegAusWiki.gpx
• VoieVerteHauteVosges.gpx

as well as all the test files from gpxpy. Many of those intentionally lack any coherent flow of trk -> trkseg -> trkpt so they don't produce a valid csv file.

I also used a bogus_basin file (src) which is a good example of conversion of multiple trk files. Though, the design case was more a single workout tracked with an Apple Watch (as that's the data I'm interested in myself.)

Example Input and Output:

Here the beginning of a tkpt from a HealthFit/Apple HealthKit tracked run, with gps coords altered.

<trk>
<type>running</type>
<trkseg>
<trkpt lat="45.0000" lon="-75.0000">
    <ele>51.0000</ele>
    <time>2021-07-21T12:37:41.000Z</time>
    <extensions>
        <gpxtpx:TrackPointExtension>
            <gpxtpx:atemp>24</gpxtpx:atemp>
            <gpxtpx:cad>72</gpxtpx:cad>
            <gpxtpx:speed>2.147612</gpxtpx:speed>
            <gpxtpx:hAcc>6.406485</gpxtpx:hAcc>
            <gpxtpx:vAcc>5.718293</gpxtpx:vAcc>
        </gpxtpx:TrackPointExtension>
    </extensions>
...

This produces a CSV that begins:

atemp,cad,course,ele,hAcc,hr,lat,lon,speed,time,type,vAcc
24.0,72.0,,51.0,6.406485,,45.0,-75.0,2.147612,2021-07-21T12:37:41.000Z,running,5.718293

Note that the hr column is not in the first trackpoint (or first several) but the header appears in the csv file.