Desktop: Add support for OCR (#8975)

packages/lib/services/ocr/utils/filterOcrText.test.ts

@@ -0,0 +1,28 @@
+import filterOcrText from './filterOcrText';
+
+const testData: string[][] = [
+	['— !',
+		'',
+	],
+
+	[
+		`- = = — ‘ =
+—`,
+		'',
+	],
+
+	['', ''],
+
+	['   testing   ', 'testing'],
+
+];
+
+describe('filterOcrText', () => {
+
+	it('should filter text', () => {
+		for (const [input, expected] of testData) {
+			expect(filterOcrText(input)).toBe(expected);
+		}
+	});
+
+});

packages/lib/services/ocr/utils/filterOcrText.ts

@@ -0,0 +1,8 @@
+export default (text: string) => {
+	// Remove all non-letter characters from the string
+	const filtered = text.replace(/\P{Letter}/ug, '');
+	// If there's nothing left, this is most likely an invalid detection, so we
+	// clear the string.
+	if (!filtered.trim()) return '';
+	return text.trim();
+};

packages/lib/services/ocr/utils/types.ts

@@ -0,0 +1,23 @@
+export const emptyRecognizeResult = (): RecognizeResult => {
+	return {
+		text: '',
+		lines: [],
+	};
+};
+
+export type RecognizeResultBoundingBox = [number, number, number, number]; // x0, y0, x1, y1
+
+export interface RecognizeResultWord {
+	t: string;
+	bb: RecognizeResultBoundingBox; // Bounding box;
+	bl?: RecognizeResultBoundingBox; // Baseline
+}
+
+export interface RecognizeResultLine {
+	words: RecognizeResultWord[];
+}
+
+export interface RecognizeResult {
+	text: string;
+	lines?: RecognizeResultLine[]; // We do not store detailed data for PDFs
+}