The GPX data of my hiking trails, which I recorded with the help of geotrackers and apps like Komoot [1], hold some potential for statistical analysis. On which days was I on the move, and when was I lazy? Which regions were my favorites for hiking, and in which regions on the world map did I cover the most miles?

No matter where the GPX files come from – whether recorded by a Garmin tracker or by an app like Komoot that lets you download the data from its website [2] – the recorded data just screams to be put through more or less intelligent analysis programs. For each hike or bike ride, the tours/ directory (Figure 1) contains one file in XML format (Figure 2). Each of these GPX files consists of a series of geodata recorded with timestamps. In each case, the data shows the longitude and latitude determined using GPS, from which, in turn, you can determine a point on the Earth's surface, visited at a given time.

Figure 1: Homegrown analysis tools extract motion data from a collection of GPX files.

Figure 2: The XML data in the GPX file represents track points on a hiking trail.

Nabbed from GitHub

It would be a tedious task to read the XML data manually with Go because its internal structure, with separate tracks, segments, and points, dictates that you have matching structures in Go. Fortunately, someone already solved that problem with the gpxgo project, which is available on GitHub. The program from Listing 1 [3] retrieves it in line 4 and completes the job in one fell swoop using ParseFile() in line 26. Any analysis program presented later on just needs to call gpxPoints() from Listing 1 with the name of a GPX file to retrieve Go structures with all the geopoints in the file and the matching timestamps.

01 package main
02
03 import (
04   "github.com/tkrajina/gpxgo/gpx"
05   "os"
06   "path/filepath"
07 )
08
09 func gpxFiles() []string {
10   tourDir := "tours"
11   files := []string{}
12
13   entries, err := os.ReadDir(tourDir)
14   if err != nil {
15     panic(err)
16   }
17
18   for _, entry := range entries {
19     gpxPath := filepath.Join(tourDir, entry.Name())
20     files = append(files, gpxPath)
21   }
22   return files
23 }
24
25 func gpxPoints(path string) []gpx.GPXPoint {
26   gpxData, err := gpx.ParseFile(path)
27   points := []gpx.GPXPoint{}
28
29   if err != nil {
30     panic(err)
31   }
32
33   for _, trk := range gpxData.Tracks {
34     for _, seg := range trk.Segments {
35       for _, pt := range seg.Points {
36         points = append(points, pt)
37       }
38     }
39   }
40   return points
41 }
42
43 func gpxAvg(path string)(float64, float64, int) {
44     nofPoints := 0
45     latSum,longSum := 0.0, 0.0
46     for _, pt := range gpxPoints(path) {
47       latSum += pt.Latitude
48       longSum += pt.Longitude
49       nofPoints++
50     }
51     return latSum/float64(nofPoints),
52       longSum/float64(nofPoints), nofPoints
53  }

To calculate the average of all geopoints in a GPX file, say, to determine where the whole trail is located, gpxAvg() first calls gpxPoints() starting in line 43, uses pt.Longitude and pt.Latitude from the Point structure to pick up the values for longitude and latitude, and adds them up to create two float64 sums. Also, for each geopoint processed, the nofPoints counter is incremented by one, and the averaging function only has to divide the total by the number of points at the end to return the mean value.

Time for an Overview

To get a brief overview of the contents of all collected GPX files, Listing 2 walks through all the files in the tours/ directory using gpxFiles() from Listing 1. It reads the files' XML data and uses gpxPoints() to return a list of all the geopoints it contains along with matching timestamps.

01 package main
02
03 import ("fmt")
04
05 func main() {
06   for _, path := range gpxFiles() {
07     lat, lon, pts := gpxAvg(path)
08     fmt.Printf("%s %.2f,%.2f (%d points)\n",
09       path, lat, lon, pts)
10   }
11 }

The output in Figure 3 shows that the trails were recorded all over the place. For example, the intersection of the latitude of 37 degrees north and the longitude of -122 degrees west is my adopted home of San Francisco. On the other hand, the latitude of 48 degrees north and the longitude of 10 degrees east represents my former home of Augsburg, Germany, which I visited as an American tourist last summer.

Figure 3: A collection of GPX files and their analysis by tourstats.go.

Out and About or Lazy?

A recording's GPX points also come with timestamps; in other words, a collection of GPX files reveals the calendar days on which I recorded walks. From this data, Listing 3 generates a time-based activity curve. To accumulate the number of all track points recorded during a given calendar day, it sets the hour, minute, and second values of all timestamps it finds to zero and uses time.Date() to set the recording date, valid for all points sampled during a specific calendar day. In the perday hash map, line 17 then increments the respective day entry by one with each matching timestamp it finds. All that remains to do then is to sort the keys of the hash map (i.e., the date values) in ascending order and to draw them on a chart with values corresponding to the assigned counters.

01 package main
02
03 import (
04   "fmt"
05   "github.com/wcharczuk/go-chart/v2"
06   "os"
07   "sort"
08   "time"
09 )
10
11 func main() {
12   perday := map[time.Time]int{}
13
14   for _, path := range gpxFiles() {
15     for _, pt := range gpxPoints(path) {
16       t := time.Date(pt.Timestamp.Year(), pt.Timestamp.Month(), pt.Timestamp.Day(), 0, 0, 0, 0, time.Local)
17       perday[t]++
18     }
19   }
20
21   keys := []time.Time{}
22   for day, _ := range perday {
23     keys = append(keys, day)
24   }
25   sort.Slice(keys, func(i, j int) bool {
26     return keys[i].Before(keys[j])
27   })
28
29   xVals := []time.Time{}
30   yVals := []float64{}
31   for _, key := range keys {
32     xVals = append(xVals, key)
33     yVals = append(yVals, float64(perday[key]))
34   }
35
36   mainSeries := chart.TimeSeries{
37     Name: "GPS Activity",
38     Style: chart.Style{
39       StrokeColor: chart.ColorBlue,
40       FillColor: chart.ColorBlue.WithAlpha(100),
41     },
42     XValues: xVals,
43     YValues: yVals,
44   }
45
46   graph := chart.Chart{
47     Width:  1280,
48     Height: 720,
49     Series: []chart.Series{mainSeries},
50   }
51
52   f, _ := os.Create("activity.png")
53   defer f.Close()
54
55   graph.Render(chart.PNG, f)
56 }