minstrel/internal/library/loudness_test.go

package library

import (
	"context"
	"errors"
	"fmt"
	"math"
	"os"
	"os/exec"
	"slices"
	"strings"
	"testing"
	"time"
)

// parseEbur128 runs the parser over a whole log, as computeLoudness does over
// ffmpeg's stderr.
func parseEbur128(log string) *ebur128Parser {
	p := newEbur128Parser()
	p.consume(strings.NewReader(log))
	return p
}

// The fixture is real ffmpeg 6.1 output for 12 s of stereo tone whose first
// second is digital silence (testdata/ebur128_fixture.txt). Pinning the parser
// to captured output, not to a hand-written imitation of it, is the point: the
// per-block lines are where a format drift would hide.
func TestEbur128Parser_RealOutput(t *testing.T) {
	raw, err := os.ReadFile("testdata/ebur128_fixture.txt")
	if err != nil {
		t.Fatalf("read fixture: %v", err)
	}
	r := parseEbur128(string(raw)).result()
	if r.err != nil {
		t.Fatalf("result err = %v", r.err)
	}
	for name, c := range map[string]struct {
		got  *float32
		want float32
	}{
		"integrated": {r.integratedLUFS, -10.7},
		"true peak":  {r.truePeakDBTP, -5.6},
		"range":      {r.rangeLU, 2.0},
	} {
		if c.got == nil || *c.got != c.want {
			t.Errorf("%s = %v, want %v", name, c.got, c.want)
		}
	}

	// 120 blocks, of which the 10 covering the silent second are below the
	// absolute gate and not counted. The loudest is -8.1 LUFS, the quietest
	// that passed -22.6.
	var blocks int32
	for _, c := range r.hist.counts {
		blocks += c
	}
	if blocks != 110 {
		t.Errorf("histogram holds %d blocks, want 110", blocks)
	}
	if r.hist.start != 474 || len(r.hist.counts) != 146 {
		t.Errorf("histogram spans bins %d..%d, want 474..619 (-22.6..-8.1 LUFS)",
			r.hist.start, int(r.hist.start)+len(r.hist.counts)-1)
	}
	if r.hist.counts[0] == 0 || r.hist.counts[len(r.hist.counts)-1] == 0 {
		t.Errorf("histogram not trimmed to its occupied bins: %v", r.hist.counts)
	}

	// Album loudness is computed from these histograms (#4996), so the
	// histogram has to reproduce ffmpeg's own figure.
	got, ok := r.hist.gatedLoudness()
	if !ok || math.Abs(got-(-10.7)) > 0.05 {
		t.Errorf("loudness from the histogram = %.3f (ok=%v), want ffmpeg's -10.7 within 0.05", got, ok)
	}
}

func TestEbur128Parser_SilenceIsAVerdictNotAMeasurement(t *testing.T) {
	log := `[Parsed_ebur128_0 @ 0x1] t: 0.1  TARGET:-23 LUFS    M:-120.7 S:-120.7     I: -70.0 LUFS       LRA:   0.0 LU  FTPK:  -inf  -inf dBFS  TPK:  -inf  -inf dBFS
[Parsed_ebur128_0 @ 0x1] t: 0.2  TARGET:-23 LUFS    M:-120.7 S:-120.7     I: -70.0 LUFS       LRA:   0.0 LU  FTPK:  -inf  -inf dBFS  TPK:  -inf  -inf dBFS
[Parsed_ebur128_0 @ 0x1] Summary:

  Integrated loudness:
    I:         -70.0 LUFS
    Threshold:   0.0 LUFS

  Loudness range:
    LRA:         0.0 LU
    Threshold:   0.0 LUFS
    LRA low:     0.0 LUFS
    LRA high:    0.0 LUFS

  True peak:
    Peak:       -inf dBFS
`
	r := parseEbur128(log).result()
	if r.err != nil {
		t.Fatalf("err = %v, want a clean result", r.err)
	}
	if r.integratedLUFS != nil || r.truePeakDBTP != nil || r.rangeLU != nil {
		t.Errorf("silence measured as integrated=%v peak=%v range=%v, want all nil",
			r.integratedLUFS, r.truePeakDBTP, r.rangeLU)
	}
	if !r.hist.empty() {
		t.Errorf("silence left blocks in the histogram: %+v", r.hist)
	}
	if r.inconclusive() {
		t.Errorf("silence reported as inconclusive; it is settled until the file changes")
	}
}

func TestEbur128Parser_NoSummaryIsUnreadable(t *testing.T) {
	r := parseEbur128("[in#0 @ 0x1] Error opening input: Invalid data found when processing input\n").result()
	if r.err == nil {
		t.Fatal("a log with no summary produced a measurement")
	}
	if r.inconclusive() {
		t.Errorf("err %v reads as inconclusive; ffmpeg ran and found nothing to measure", r.err)
	}
	if !strings.Contains(r.err.Error(), "Invalid data found") {
		t.Errorf("err %q does not carry ffmpeg's reason", r.err)
	}
}

func TestEbur128Parser_LoudBlocksLandInTheTopBin(t *testing.T) {
	p := parseEbur128(`[Parsed_ebur128_0 @ 0x1] t: 0.4  TARGET:-23 LUFS    M:  12.4 S:  12.4     I:  12.4 LUFS
[Parsed_ebur128_0 @ 0x1] t: 0.5  TARGET:-23 LUFS    M:  -5.0 S:  -5.0     I:  -5.0 LUFS
`)
	h := p.histogram()
	if int(h.start)+len(h.counts) != loudnessHistBins || h.counts[len(h.counts)-1] != 1 {
		t.Errorf("a +12.4 LUFS block was not counted in the top bin: %+v", h)
	}
	if h.start != 650 || h.counts[0] != 1 {
		t.Errorf("a -5.0 LUFS block is not in bin 650: start %d, first %d", h.start, h.counts[0])
	}
}

func TestBlockHistogram_GatedLoudness(t *testing.T) {
	bin := func(lufs float64) int { return int(math.Round((lufs - loudnessHistFloor) * loudnessHistPerLU)) }
	hist := func(blocks map[float64]int32) blockHistogram {
		lo, hi := loudnessHistBins, 0
		for l := range blocks {
			lo, hi = min(lo, bin(l)), max(hi, bin(l))
		}
		h := blockHistogram{start: int16(lo), counts: make([]int32, hi-lo+1)}
		for l, n := range blocks {
			h.counts[bin(l)-lo] = n
		}
		return h
	}

	cases := []struct {
		name   string
		blocks map[float64]int32
		want   float64
	}{
		// One level throughout is that level.
		{"steady", map[float64]int32{-14: 50}, -14},
		// The quiet half is 30 LU below the loud half, past the relative gate
		// (10 LU below the ungated loudness, here about -13), so it is dropped
		// and the result is the loud half alone.
		{"quiet passage gated out", map[float64]int32{-10: 100, -40: 100}, -10},
		// 6 LU apart is inside the gate, so both count, energy-weighted: the
		// result sits nearer the louder level than the midpoint (-15) does.
		{"both inside the gate", map[float64]int32{-12: 100, -18: 100}, -14.037},
	}
	for _, c := range cases {
		got, ok := hist(c.blocks).gatedLoudness()
		if !ok || math.Abs(got-c.want) > 0.01 {
			t.Errorf("%s: gatedLoudness = %.3f (ok=%v), want %.3f", c.name, got, ok, c.want)
		}
	}
	if _, ok := (blockHistogram{}).gatedLoudness(); ok {
		t.Errorf("an empty histogram reported a loudness")
	}
}

func TestEbur128Args(t *testing.T) {
	args := ebur128Args("/music/a.flac")
	joined := strings.Join(args, " ")
	for _, want := range []string{"peak=true", "dualmono=true", "framelog=info", "-map 0:a:0", "-nostdin", "-f null -"} {
		if !strings.Contains(joined, want) {
			t.Errorf("args %q lack %q", joined, want)
		}
	}
	// The path is its own argument, never spliced into the filter string.
	if i := slices.Index(args, "-i"); i < 0 || args[i+1] != "/music/a.flac" {
		t.Errorf("args %q do not pass the path after -i", args)
	}
}

func TestLoudnessTimeout_ScalesWithLength(t *testing.T) {
	if got := loudnessTimeout(0); got != loudnessBaseTimeout {
		t.Errorf("unknown length: %s, want the base %s", got, loudnessBaseTimeout)
	}
	if got, want := loudnessTimeout(int32((2 * time.Hour).Milliseconds())), loudnessBaseTimeout+30*time.Minute; got != want {
		t.Errorf("two-hour mix: %s, want %s", got, want)
	}
	if got := loudnessTimeout(-5); got != loudnessBaseTimeout {
		t.Errorf("negative length: %s, want the base %s", got, loudnessBaseTimeout)
	}
}

func TestLoudnessResult_Inconclusive(t *testing.T) {
	for _, err := range []error{
		fmt.Errorf("ffmpeg: %w", errLoudnessTimeout),
		fmt.Errorf("ffmpeg: %w", context.Canceled),
		fmt.Errorf("ffmpeg: %w", exec.ErrNotFound),
	} {
		if !(loudnessResult{err: err}).inconclusive() {
			t.Errorf("%v: not inconclusive, so it would be stored as a verdict", err)
		}
	}
	if (loudnessResult{err: errors.New("ffmpeg exited 1: moov atom not found")}).inconclusive() {
		t.Errorf("a decode failure read as inconclusive; it would be retried every pass")
	}
	if (loudnessResult{}).inconclusive() {
		t.Errorf("a clean result read as inconclusive")
	}
}

func TestBackfillLoudnessResult_Add(t *testing.T) {
	var r BackfillLoudnessResult
	for _, o := range []loudnessOutcome{
		loudnessMeasured, loudnessMeasured, loudnessSilent, loudnessUnreadable,
		loudnessInconclusive, loudnessStoreFailed,
	} {
		r.add(o)
	}
	// A failed write stored nothing, so it is retried like an inconclusive one.
	want := BackfillLoudnessResult{Processed: 6, Measured: 2, Silent: 1, Unreadable: 1, Inconclusive: 2}
	if r != want {
		t.Errorf("tally = %+v, want %+v", r, want)
	}
}

func TestValidateLoudnessSettings(t *testing.T) {
	with := func(concurrency, cacheMB int32) LoudnessSettings {
		s := DefaultLoudnessSettings
		s.BackfillConcurrency, s.LeveledCacheMB = concurrency, cacheMB
		return s
	}
	for _, s := range []LoudnessSettings{
		with(minBackfillConcurrency, minLeveledCacheMB),
		with(maxBackfillConcurrency, maxLeveledCacheMB),
	} {
		if err := validateLoudnessSettings(s); err != nil {
			t.Errorf("%+v rejected: %v", s, err)
		}
	}
	for _, s := range []LoudnessSettings{
		with(0, minLeveledCacheMB),
		with(maxBackfillConcurrency+1, minLeveledCacheMB),
		with(minBackfillConcurrency, minLeveledCacheMB-1),
		with(minBackfillConcurrency, maxLeveledCacheMB+1),
	} {
		if err := validateLoudnessSettings(s); !errors.Is(err, ErrLoudnessSettingOutOfRange) {
			t.Errorf("%+v: err = %v, want ErrLoudnessSettingOutOfRange", s, err)
		}
	}
	var nilSvc *LoudnessSettingsService
	if got := nilSvc.Get(); got != DefaultLoudnessSettings {
		t.Errorf("nil service Get = %+v, want the defaults", got)
	}
}