fix(hints): cap OCR processes per scan

commit: ce28c3608856048f341730bb6e56b14f5be52534
author: brice · 3d ago · 2026-10-19 15:28 UTC
parent: 71feaa7
signature: GPG VerifiedCD88F9E564E0102646FB6077A1A07C7C7C3E349D · arch-gpg Verified for the committer email

2 files · +58 -1

M internal/adapters/hints/ocr.go +13 -1


@@ -154,12 +154,24 @@ func bandsOf(height int) []band {
154	154		func (w worker) recognize(ctx context.Context, jobs []bandJob) {
155	155		ctx, cancel := context.WithTimeout(ctx, 15*time.Second)
156	156		defer cancel()
	157	+	// One scan-wide pool: at most ocrWorkers tesseract processes run at once,
	158	+	// whatever the number of windows.
	159	+	sem := make(chan struct{}, max(ocrWorkers, 1))
157	160		var wg sync.WaitGroup
158	161		for _, j := range jobs {
159	162		if j.hit {
160	163		continue
161	164		}
162		-	wg.Go(func() { j.words, j.err = readBand(ctx, &w.ocr, j) })
	165	+	select {
	166	+	case sem <- struct{}{}:
	167	+	case <-ctx.Done():
	168	+	j.err = ctx.Err()
	169	+	continue
	170	+	}
	171	+	wg.Go(func() {
	172	+	defer func() { <-sem }()
	173	+	j.words, j.err = readBand(ctx, &w.ocr, j)
	174	+	})
163	175		}
164	176		wg.Wait()
165	177		}

 func (w *worker) recognize(ctx context.Context, jobs []*bandJob) {
 	ctx, cancel := context.WithTimeout(ctx, 15*time.Second)
 	defer cancel()
+	// One scan-wide pool: at most ocrWorkers tesseract processes run at once,
+	// whatever the number of windows.
+	sem := make(chan struct{}, max(ocrWorkers, 1))
 	var wg sync.WaitGroup
 	for _, j := range jobs {
 		if j.hit {
 			continue
 		}
-		wg.Go(func() { j.words, j.err = readBand(ctx, &w.ocr, j) })
+		select {
+		case sem <- struct{}{}:
+		case <-ctx.Done():
+			j.err = ctx.Err()
+			continue
+		}
+		wg.Go(func() {
+			defer func() { <-sem }()
+			j.words, j.err = readBand(ctx, &w.ocr, j)
+		})
 	}
 	wg.Wait()
 }

M internal/adapters/hints/ocr_test.go +45 -0


@@ -8,6 +8,7 @@ import (
8	8		"io"
9	9		"os"
10	10		"os/exec"
	11	+	"path/filepath"
11	12		"reflect"
12	13		"strings"
13	14		"testing"

@@ -191,3 +192,47 @@ func TestPGMMatchesPNG(t *testing.T) {
191	192		t.Fatalf("PNG %d words, PGM %d words", len(res[0]), len(res[1]))
192	193		}
193	194		}
	195	+
	196	+	// TestRecognizeCapsProcesses runs the bands of several windows through a
	197	+	// stand-in tesseract that logs when each process starts and ends.
	198	+	func TestRecognizeCapsProcesses(t *testing.T) {
	199	+	dir := t.TempDir()
	200	+	logPath := filepath.Join(dir, "log")
	201	+	script := filepath.Join(dir, "tesseract")
	202	+	body := "#!/bin/sh\ncat >/dev/null\necho s >> " + logPath + "\nsleep 0.1\necho e >> " + logPath + "\nprintf 'level\\tpage_num\\tblock_num\\tpar_num\\tline_num\\tword_num\\tleft\\ttop\\twidth\\theight\\tconf\\ttext\\n'\n"
	203	+	if err := os.WriteFile(script, []byte(body), 0o700); err != nil {
	204	+	t.Fatal(err)
	205	+	}
	206	+	old := ocrWorkers
	207	+	ocrWorkers = 2
	208	+	t.Cleanup(func() { ocrWorkers = old })
	209	+	w := &worker{ocr: ocrState{done: true, binary: script, lang: "eng"}}
	210	+	var jobs []*bandJob
	211	+	for range 8 { // several windows' bands
	212	+	g := image.NewGray(image.Rect(0, 0, 8, 8))
	213	+	jobs = append(jobs, &bandJob{ext: g.Rect, own: [2]int{0, 8}, gray: g})
	214	+	}
	215	+	w.recognize(context.Background(), jobs)
	216	+	data, err := os.ReadFile(logPath)
	217	+	if err != nil {
	218	+	t.Fatal(err)
	219	+	}
	220	+	cur, peak, starts := 0, 0, 0
	221	+	for _, l := range strings.Fields(string(data)) {
	222	+	if l == "s" {
	223	+	cur++
	224	+	starts++
	225	+	peak = max(peak, cur)
	226	+	} else {
	227	+	cur--
	228	+	}
	229	+	}
	230	+	if starts != 8 \|\| peak > 2 {
	231	+	t.Fatalf("starts=%d peak=%d", starts, peak)
	232	+	}
	233	+	for _, j := range jobs {
	234	+	if j.err != nil {
	235	+	t.Fatal(j.err)
	236	+	}
	237	+	}
	238	+	}


@@ -8,6 +8,7 @@ import (
8	"io"	8		"io"
9	"os"	9		"os"
10	"os/exec"	10		"os/exec"
		11	+	"path/filepath"
11	"reflect"	12		"reflect"
12	"strings"	13		"strings"
13	"testing"	14		"testing"

@@ -191,3 +192,47 @@ func TestPGMMatchesPNG(t *testing.T) {
191	t.Fatalf("PNG %d words, PGM %d words", len(res[0]), len(res[1]))	192		t.Fatalf("PNG %d words, PGM %d words", len(res[0]), len(res[1]))
192	}	193		}
193	}	194		}
		195	+
		196	+	// TestRecognizeCapsProcesses runs the bands of several windows through a
		197	+	// stand-in tesseract that logs when each process starts and ends.
		198	+	func TestRecognizeCapsProcesses(t *testing.T) {
		199	+	dir := t.TempDir()
		200	+	logPath := filepath.Join(dir, "log")
		201	+	script := filepath.Join(dir, "tesseract")
		202	+	body := "#!/bin/sh\ncat >/dev/null\necho s >> " + logPath + "\nsleep 0.1\necho e >> " + logPath + "\nprintf 'level\\tpage_num\\tblock_num\\tpar_num\\tline_num\\tword_num\\tleft\\ttop\\twidth\\theight\\tconf\\ttext\\n'\n"
		203	+	if err := os.WriteFile(script, []byte(body), 0o700); err != nil {
		204	+	t.Fatal(err)
		205	+	}
		206	+	old := ocrWorkers
		207	+	ocrWorkers = 2
		208	+	t.Cleanup(func() { ocrWorkers = old })
		209	+	w := &worker{ocr: ocrState{done: true, binary: script, lang: "eng"}}
		210	+	var jobs []*bandJob
		211	+	for range 8 { // several windows' bands
		212	+	g := image.NewGray(image.Rect(0, 0, 8, 8))
		213	+	jobs = append(jobs, &bandJob{ext: g.Rect, own: [2]int{0, 8}, gray: g})
		214	+	}
		215	+	w.recognize(context.Background(), jobs)
		216	+	data, err := os.ReadFile(logPath)
		217	+	if err != nil {
		218	+	t.Fatal(err)
		219	+	}
		220	+	cur, peak, starts := 0, 0, 0
		221	+	for _, l := range strings.Fields(string(data)) {
		222	+	if l == "s" {
		223	+	cur++
		224	+	starts++
		225	+	peak = max(peak, cur)
		226	+	} else {
		227	+	cur--
		228	+	}
		229	+	}
		230	+	if starts != 8 \|\| peak > 2 {
		231	+	t.Fatalf("starts=%d peak=%d", starts, peak)
		232	+	}
		233	+	for _, j := range jobs {
		234	+	if j.err != nil {
		235	+	t.Fatal(j.err)
		236	+	}
		237	+	}
		238	+	}