Add first pass of Textract support (#45)

Author: Balearica

build-deno-compile.sh

@@ -17,12 +17,15 @@ mkdir -p build
 # Build for different platforms
 echo "Building for Linux x64..."
 deno compile --allow-sys --allow-read  --allow-write --target x86_64-unknown-linux-gnu --output build/scribe-linux-x64 cli/scribe.js
+# deno compile --allow-sys --allow-read  --allow-write --target x86_64-unknown-linux-gnu --output build/scribe-linux-x64 --include mupdf --include fonts --include js/worker cli/scribe.js
 
 echo "Building for macOS x64..."
 deno compile --allow-sys --allow-read  --allow-write --target x86_64-apple-darwin --output build/scribe-macos-x64 cli/scribe.js
+# deno compile --allow-sys --allow-read  --allow-write --target x86_64-apple-darwin --output build/scribe-macos-x64 --include mupdf --include fonts --include js/worker cli/scribe.js
 
 echo "Building for Windows x64..."
 deno compile --allow-sys --allow-read  --allow-write --target x86_64-pc-windows-msvc --output build/scribe-windows-x64.exe cli/scribe.js
+# deno compile --allow-sys --allow-read  --allow-write --target x86_64-pc-windows-msvc --output build/scribe-windows-x64.exe --include mupdf --include fonts --include js/worker cli/scribe.js
 
 # Create checksums
 cd build

fonts/Dingbats.woff

@@ -137,7 +137,8 @@ export function writeHtml({
       top += pageMetricsArr[g].dims.height + 10;
     }
 
-    if (reflowText) {
+    // Do not overwrite paragraphs from Abbyy or Textract.
+    if (reflowText && (!pageObj.textSource || !['textract', 'abbyy'].includes(pageObj.textSource))) {
       const angle = pageMetricsArr[g].angle || 0;
       assignParagraphs(pageObj, angle);
     }

js/export/writeHtml.js

@@ -26,7 +26,8 @@ export function writeText(ocrCurrent, minpage = 0, maxpage = -1, reflowText = fa
 
     const pageObj = ocrCurrent[g];
 
-    if (reflowText) {
+    // Do not overwrite paragraphs from Abbyy or Textract.
+    if (reflowText && (!pageObj.textSource || !['textract', 'abbyy'].includes(pageObj.textSource))) {
       const angle = pageMetricsArr[g].angle || 0;
       assignParagraphs(pageObj, angle);
     }

js/export/writeText.js

@@ -223,6 +223,35 @@ export async function loadChiSimFont() {
   return chiReady;
 }
 
+let dingbatsReadyRes;
+let dingbatsReady;
+
+/**
+ * Loads dingbats font. Returns early if already loaded.
+ */
+export async function loadDingbatsFont() {
+  console.log('Loading Dingbats font');
+  if (dingbatsReady) return dingbatsReady;
+
+  dingbatsReady = new Promise((resolve, reject) => {
+    dingbatsReadyRes = resolve;
+  });
+
+  let /** @type {Promise<ArrayBuffer>} */ dingbatsSrc;
+  if (typeof process === 'undefined') {
+    dingbatsSrc = fetch(new URL('../fonts/Dingbats.woff', import.meta.url)).then((res) => res.arrayBuffer());
+  } else {
+    const { readFile } = await import('node:fs/promises');
+    dingbatsSrc = readFile(new URL('../fonts/Dingbats.woff', import.meta.url)).then((res) => res.buffer);
+  }
+
+  FontCont.supp.dingbats = await loadFont('Dingbats', 'normal', 'sans', await dingbatsSrc, false);
+
+  dingbatsReadyRes();
+
+  return dingbatsReady;
+}
+
 /**
  * Enable or disable font optimization settings.
  * This function is used rather than exposing the settings using the `opt` object, as these settings exist on the font container in both the main thread and the worker threads.

js/fontContainerMain.js

@@ -95,6 +95,7 @@ export async function initGeneralWorker() {
     obj.convertPageHocr = wrap('convertPageHocr');
     obj.convertPageAbbyy = wrap('convertPageAbbyy');
     obj.convertPageStext = wrap('convertPageStext');
+    obj.convertDocTextract = wrap('convertDocTextract');
 
     obj.optimizeFont = wrap('optimizeFont');
 
@@ -174,6 +175,12 @@ export class gs {
    */
   static convertPageAbbyy = async (args) => (await gs.schedulerInner.addJob('convertPageAbbyy', args));
 
+  /**
+   * @param {Parameters<typeof import('./import/convertDocTextract.js').convertDocTextract>[0]} args
+   * @returns {ReturnType<typeof import('./import/convertDocTextract.js').convertDocTextract>}
+   */
+  static convertDocTextract = async (args) => (await gs.schedulerInner.addJob('convertDocTextract', args));
+
   /**
    * @param {Parameters<typeof import('./import/convertPageStext.js').convertPageStext>[0]} args
    * @returns {ReturnType<typeof import('./import/convertPageStext.js').convertPageStext>}

js/generalWorkerMain.js

@@ -10,6 +10,10 @@ declare global {
         sup: boolean;
         dropcap: boolean;
     };
+
+    // Strings representing supported sources of text.
+    // `stext` indicates the text was extracted directly from a PDF using mupdf.
+    type TextSource = null | 'tesseract' | 'textract' | 'abbyy' | 'stext' | 'hocr';
 
     type FontState = {
         enableOpt: boolean;
@@ -176,6 +180,53 @@ declare global {
     type LayoutDataColumn = import("./objects/layoutObjects.js").LayoutDataColumn;
     type LayoutRegion = import("./objects/layoutObjects.js").LayoutRegion;
 
+    interface Point {
+        x: number;
+        y: number;
+    }
+
+    interface Polygon {
+        br: Point;
+        bl: Point;
+        tr: Point;
+        tl: Point;
+    }
+
+    interface TextractBoundingBox {
+        Width: number;
+        Height: number;
+        Left: number;
+        Top: number;
+    }
+
+    interface TextractPoint {
+        X: number;
+        Y: number;
+    }
+
+    interface TextractGeometry {
+        BoundingBox: TextractBoundingBox;
+        Polygon: TextractPoint[];
+        RotationAngle: number;
+    }
+
+    interface Relationship {
+        Type: string;
+        Ids: string[];
+    }
+
+    interface TextractBlock {
+        BlockType: "WORD" | "LINE" | "PAGE" | "KEY_VALUE_SET" | "CELL" | "MERGED_CELL" | "SELECTION_ELEMENT" | "TABLE"; 
+        Confidence: number;
+        Text: string;
+        TextType: "PRINTED" | "HANDWRITING";
+        Geometry: TextractGeometry;
+        Id: string;
+        Page?: number;
+        Relationships?: Relationship[];
+    }
+
+
 }
 
 export { };