func()

in analysis/app/trend.go [78:174]

• 88 lines of code
• 15 McCabe index (conditional complexity)

func (a *App) trendQuery(ctx context.Context, q string, opt plotOptions) *trendData {
	d := &trendData{Q: q}
	if q == "" {
		ul := a.StorageClient.ListUploads(ctx, `trend>`, []string{"by", "upload-time", "trend"}, 16)
		defer ul.Close()
		for ul.Next() {
			d.TrendUploads = append(d.TrendUploads, ul.Info())
		}
		if err := ul.Err(); err != nil {
			errorf(ctx, "failed to fetch recent trend uploads: %v", err)
		}
		return d
	}

	// TODO(quentin): Chunk query based on matching upload IDs.
	res := a.StorageClient.Query(ctx, q)
	defer res.Close()
	t, resultCols := queryToTable(res)
	if err := res.Err(); err != nil {
		errorf(ctx, "failed to read query results: %v", err)
		d.Error = fmt.Sprintf("failed to read query results: %v", err)
		return d
	}
	for _, col := range []string{"commit", "commit-time", "branch", "name"} {
		if !hasStringColumn(t, col) {
			d.Error = fmt.Sprintf("results missing %q label", col)
			return d
		}
	}
	if opt.x != "" && !hasStringColumn(t, opt.x) {
		d.Error = fmt.Sprintf("results missing x label %q", opt.x)
		return d
	}
	data := plot(t, resultCols, opt)

	// TODO(quentin): Give the user control over across vs. plotting in separate graphs, instead of only showing one graph with ns/op for each benchmark.

	if opt.raw {
		data = table.MapTables(data, func(_ table.GroupID, t *table.Table) *table.Table {
			// From http://tristen.ca/hcl-picker/#/hlc/9/1.13/F1796F/B3EC6C
			colors := []string{"#F1796F", "#B3EC6C", "#F67E9D", "#6CEB98", "#E392CB", "#0AE4C6", "#B7ABEC", "#16D7E9", "#75C4F7"}
			colorIdx := 0
			partColors := make(map[string]string)
			styles := make([]string, t.Len())
			for i, part := range t.MustColumn("upload-part").([]string) {
				if _, ok := partColors[part]; !ok {
					partColors[part] = colors[colorIdx]
					colorIdx++
					if colorIdx >= len(colors) {
						colorIdx = 0
					}
				}
				styles[i] = "color: " + partColors[part]
			}
			return table.NewBuilder(t).Add("style", styles).Done()
		})
		columns := []column{
			{Name: "commit-index"},
			{Name: "result"},
			{Name: "style", Role: "style"},
			{Name: "commit", Role: "tooltip"},
		}
		d.PlotData = tableToJS(data.Table(data.Tables()[0]), columns)
		d.PlotType = "ScatterChart"
		return d
	}

	// Pivot all of the benchmarks into columns of a single table.
	ar := &aggResults{
		Across: "name",
		Values: []string{"filtered normalized mean result", "normalized mean result", "normalized median result", "normalized min result", "normalized max result"},
	}
	data = ggstat.Agg("commit", "branch", "commit-index")(ar.agg).F(data)

	tables := data.Tables()
	infof(ctx, "tables: %v", tables)
	columns := []column{
		{Name: "commit-index"},
		{Name: "commit", Role: "tooltip"},
	}
	for _, prefix := range ar.Prefixes {
		if len(ar.Prefixes) == 1 {
			columns = append(columns,
				column{Name: prefix + "/normalized mean result"},
				column{Name: prefix + "/normalized min result", Role: "interval"},
				column{Name: prefix + "/normalized max result", Role: "interval"},
				column{Name: prefix + "/normalized median result"},
			)
		}
		columns = append(columns,
			column{Name: prefix + "/filtered normalized mean result"},
		)
	}
	d.PlotData = tableToJS(data.Table(tables[0]), columns)
	d.PlotType = "LineChart"
	return d
}