fix(ocr): Arabic+Latin font fix and added tests

src/tests/compare/ocr-page.test.ts

@@ -32,14 +32,26 @@ describe('compare OCR page recognition', () => {
 
     mockWorker.recognize.mockResolvedValue({
       data: {
-        words: [
+        blocks: [
           {
-            text: 'Hello',
-            bbox: { x0: 10, y0: 20, x1: 60, y1: 40 },
-          },
-          {
-            text: 'world',
-            bbox: { x0: 70, y0: 20, x1: 120, y1: 40 },
+            paragraphs: [
+              {
+                lines: [
+                  {
+                    words: [
+                      {
+                        text: 'Hello',
+                        bbox: { x0: 10, y0: 20, x1: 60, y1: 40 },
+                      },
+                      {
+                        text: 'world',
+                        bbox: { x0: 70, y0: 20, x1: 120, y1: 40 },
+                      },
+                    ],
+                  },
+                ],
+              },
+            ],
           },
         ],
       },
@@ -52,7 +64,11 @@ describe('compare OCR page recognition', () => {
       1,
       expect.any(Function)
     );
-    expect(mockWorker.recognize).toHaveBeenCalledWith(canvas);
+    expect(mockWorker.recognize).toHaveBeenCalledWith(
+      canvas,
+      {},
+      { text: true, blocks: true }
+    );
     expect(mockWorker.terminate).toHaveBeenCalledTimes(1);
     expect(model.source).toBe('ocr');
     expect(model.hasText).toBe(true);

src/tests/font-loader.test.ts

@@ -12,11 +12,11 @@ describe('font-loader', () => {
 
   it('builds a self-hosted font URL when an OCR font base URL is configured', () => {
     expect(
-      resolveFontUrl('Noto Sans Arabic', {
+      resolveFontUrl('Noto Naskh Arabic', {
         VITE_OCR_FONT_BASE_URL: 'https://internal.example.com/wasm/ocr/fonts/',
       })
     ).toBe(
-      'https://internal.example.com/wasm/ocr/fonts/NotoSansArabic-Regular.ttf'
+      'https://internal.example.com/wasm/ocr/fonts/NotoNaskhArabic-Regular.ttf'
     );
   });
 

src/tests/hocr-transform.test.ts

@@ -8,6 +8,7 @@ import {
   normalizeText,
   calculateWordTransform,
   calculateSpaceTransform,
+  calculateLineFontSize,
 } from '../js/utils/hocr-transform';
 import type { OcrWord } from '@/types';
 
@@ -298,6 +299,32 @@ describe('hocr-transform', () => {
     });
   });
 
+  describe('calculateLineFontSize', () => {
+    it('returns line height + baseline intercept', () => {
+      const line = {
+        bbox: { x0: 0, y0: 100, x1: 500, y1: 130 },
+        baseline: { slope: 0, intercept: 2 },
+        textangle: 0,
+        words: [] as OcrWord[],
+        direction: 'ltr' as const,
+        injectWordBreaks: true,
+      };
+      expect(calculateLineFontSize(line)).toBe(32);
+    });
+
+    it('clamps to a minimum of 1 for degenerate lines', () => {
+      const line = {
+        bbox: { x0: 0, y0: 0, x1: 0, y1: 0 },
+        baseline: { slope: 0, intercept: 0 },
+        textangle: 0,
+        words: [] as OcrWord[],
+        direction: 'ltr' as const,
+        injectWordBreaks: true,
+      };
+      expect(calculateLineFontSize(line)).toBe(1);
+    });
+  });
+
   describe('calculateSpaceTransform', () => {
     const baseLine = {
       bbox: { x0: 0, y0: 100, x1: 500, y1: 130 },

src/tests/ocr.test.ts

@@ -1,4 +1,4 @@
-import { beforeEach, describe, expect, it, vi } from 'vitest';
+import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 
 const {
   createConfiguredTesseractWorker,
@@ -23,6 +23,7 @@ const mockWorker = {
 const mockPdfPage = {
   getViewport: vi.fn(() => ({ width: 200, height: 100 })),
   render: vi.fn(() => ({ promise: Promise.resolve() })),
+  cleanup: vi.fn(),
 };
 
 const mockPdfOutputPage = {
@@ -177,27 +178,31 @@ describe('performOcr', () => {
       expect.any(Function)
     );
     expect(mockWorker.setParameters).toHaveBeenCalledWith({
-      tessjs_create_hocr: '1',
       tessedit_pageseg_mode: '3',
     });
     expect(mockWorker.recognize).toHaveBeenCalledTimes(1);
     expect(mockWorker.terminate).toHaveBeenCalledTimes(1);
     expect(result.fullText).toContain('Recognized text');
     expect(result.pdfBytes).toBeInstanceOf(Uint8Array);
+    expect(result.warnings).toEqual([]);
   });
 
-  it('terminates the Tesseract worker when OCR fails', async () => {
+  it('records a warning and still terminates the worker when OCR fails on a page', async () => {
     mockWorker.recognize.mockRejectedValueOnce(new Error('ocr failed'));
 
-    await expect(
-      performOcr(new Uint8Array([1, 2, 3]), {
-        language: 'eng',
-        resolution: 2,
-        binarize: false,
-        whitelist: '',
-      })
-    ).rejects.toThrow('ocr failed');
+    const result = await performOcr(new Uint8Array([1, 2, 3]), {
+      language: 'eng',
+      resolution: 2,
+      binarize: false,
+      whitelist: '',
+    });
 
     expect(mockWorker.terminate).toHaveBeenCalledTimes(1);
+    expect(result.warnings).toHaveLength(1);
+    expect(result.warnings[0]).toMatchObject({
+      page: 1,
+      kind: 'recognize-error',
+    });
+    expect(result.warnings[0].message).toContain('ocr failed');
   });
 });