Improved import and export support; misc changes (#49)

Author: Balearica

cli/main.js

@@ -103,7 +103,7 @@ async function main(func, params) {
     const debugDir = `${outputDir}/${outputStem}_debug`;
     fs.mkdirSync(debugDir, { recursive: true });
     const outputPathCsv = `${debugDir}/_debug.csv`;
-    scribe.utils.writeDebugCsv(scribe.data.ocr.active, outputPathCsv);
+    scribe.utils.writeDebugCsv({ pages: scribe.data.ocr.active, fileName: outputPathCsv });
 
     scribe.utils.dumpDebugImages(debugDir);
     scribe.utils.dumpHOCR(debugDir);

js/clear.js

@@ -5,7 +5,7 @@ import {
   layoutRegions,
   ocrAll,
   ocrAllRaw,
-  pageMetricsArr,
+  pageMetricsAll,
 } from './containers/dataContainer.js';
 import { FontCont } from './containers/fontContainer.js';
 import { ImageCache } from './containers/imageContainer.js';
@@ -19,7 +19,7 @@ export function clearData() {
   ocrAllRaw.active = [];
   layoutRegions.pages.length = 0;
   layoutDataTables.pages.length = 0;
-  pageMetricsArr.length = 0;
+  pageMetricsAll.length = 0;
   convertPageWarn.length = 0;
   ImageCache.clear();
   FontCont.clear();

js/containers/dataContainer.js

@@ -63,7 +63,7 @@ export const ocrAll = { active: [] };
 export const ocrAllRaw = { active: [] };
 
 /** @type {Array<PageMetrics>} */
-export const pageMetricsArr = [];
+export const pageMetricsAll = [];
 
 /**
  * Class that stores various debug data.

js/containers/imageContainer.js

@@ -5,15 +5,15 @@ import {
 import { initMuPDFWorker } from '../../mupdf/mupdf-async.js';
 
 import { updateFontContWorkerMain } from '../fontContainerMain.js';
-import { pageMetricsArr } from './dataContainer.js';
+import { pageMetricsAll } from './dataContainer.js';
 import {
   FontCont,
   FontContainerFont,
   loadOpentype,
 } from './fontContainer.js';
 
 import { gs } from '../generalWorkerMain.js';
-import { imageUtils } from '../objects/imageObjects.js';
+import { imageUtils, ImageWrapper } from '../objects/imageObjects.js';
 import { range } from '../utils/miscUtils.js';
 import { opt } from './app.js';
 
@@ -42,32 +42,6 @@ export class MuPDFScheduler {
   }
 }
 
-export class ImageWrapper {
-  /**
-   * @param {number} n - Page number
-   * @param {string} imageStr - Base-64 encoded image string. Should start with "data:image/png" or "data:image/jpeg".
-   * @param {string} colorMode - Color mode ("color", "gray", or "binary").
-   * @param {boolean} rotated - Whether image has been rotated.
-   * @param {boolean} upscaled - Whether image has been upscaled.
-   *
-   * All properties of this object must be serializable, as ImageWrapper objects are sent between threads.
-   * This means that no promises can be used.
-   */
-  constructor(n, imageStr, colorMode, rotated = false, upscaled = false) {
-    this.n = n;
-    this.src = imageStr;
-    const format0 = imageStr.match(/^data:image\/(png|jpeg)/)?.[1];
-    if (!format0 || !['png', 'jpeg'].includes(format0)) throw new Error(`Invalid image format: ${format0}`);
-    this.format = format0;
-    this._dims = null;
-    this.rotated = rotated;
-    this.upscaled = upscaled;
-    this.colorMode = colorMode;
-    /** @type {?ImageBitmap} */
-    this.imageBitmap = null;
-  }
-}
-
 /**
  * @typedef {Object} ImageProperties
  * @property {boolean} [rotated]
@@ -126,7 +100,7 @@ export class ImageCache {
       colorMode = color ? 'color' : 'gray';
     }
 
-    let pageAngle = pageMetricsArr[n].angle || 0;
+    let pageAngle = pageMetricsAll[n].angle || 0;
     if (Math.abs(pageAngle) < 0.05) pageAngle = 0;
 
     // If no preference is specified for rotation, default to true.
@@ -213,7 +187,7 @@ export class ImageCache {
     if (ImageCache.inputModes.image) {
       return ImageCache.nativeSrc[n];
     } if (ImageCache.inputModes.pdf) {
-      const pageMetrics = pageMetricsArr[n];
+      const pageMetrics = pageMetricsAll[n];
       const targetWidth = pageMetrics.dims.width;
       const dpi = 300 * (targetWidth / ImageCache.pdfDims300[n].width);
       const muPDFScheduler = await ImageCache.getMuPDFScheduler();
@@ -232,7 +206,7 @@ export class ImageCache {
    * @param {boolean} [saveNativeImage=true] - Whether the native image should be saved.
    */
   static transformImage = async (inputImage, n, props, saveNativeImage = true) => {
-    let pageAngle = pageMetricsArr[n].angle || 0;
+    let pageAngle = pageMetricsAll[n].angle || 0;
     if (Math.abs(pageAngle) < 0.05) pageAngle = 0;
 
     // If no preference is specified for rotation, default to true.
@@ -245,8 +219,8 @@ export class ImageCache {
     await gs.getGeneralScheduler();
 
     const resPromise = (async () => {
-    // Wait for non-rotated version before replacing with promise
-      if (typeof process === 'undefined') await gs.initTesseract({ anyOk: true });
+      // Wait for non-rotated version before replacing with promise
+      await gs.initTesseract({ anyOk: true });
       return gs.recognize({
         image: inputImage.src,
         options: { rotateRadians: angleArg, upscale: upscaleArg },
@@ -280,7 +254,7 @@ export class ImageCache {
       return { native: undefined, binary: undefined };
     }
 
-    const significantRotation = Math.abs(pageMetricsArr[n].angle || 0) > 0.05;
+    const significantRotation = Math.abs(pageMetricsAll[n].angle || 0) > 0.05;
 
     const newNative = !ImageCache.native[n] || !imageUtils.compatible(ImageCache.nativeProps[n], props, significantRotation);
     const newBinary = !nativeOnly && (!ImageCache.binary[n] || !imageUtils.compatible(ImageCache.binaryProps[n], props, significantRotation));
@@ -426,7 +400,7 @@ export class ImageCache {
 
     ImageCache.pdfDims300.forEach((x, i) => {
       const pageDims = { width: Math.round(x.width * pageDPI[i] / 300), height: Math.round(x.height * pageDPI[i] / 300) };
-      pageMetricsArr[i] = new PageMetrics(pageDims);
+      pageMetricsAll[i] = new PageMetrics(pageDims);
     });
 
     // WIP: Extract fonts embedded in PDFs.

js/coordinates.js

@@ -3,7 +3,7 @@
 // Image Coordinate Space: coordinate space of a particular image
 // Canvas Coordinate Space: coordinate space of canvas, used for user interactions
 
-import { pageMetricsArr } from './containers/dataContainer.js';
+import { pageMetricsAll } from './containers/dataContainer.js';
 import { ImageCache } from './containers/imageContainer.js';
 
 /**
@@ -27,7 +27,7 @@ function rotateBoundingBox(boundingBox, rotateAngle, n) {
   let angleAdjXRect = 0;
   let angleAdjYRect = 0;
 
-  const pageDims = pageMetricsArr[n].dims;
+  const pageDims = pageMetricsAll[n].dims;
 
   const sinAngle = Math.sin(rotateAngle * (Math.PI / 180));
   const cosAngle = Math.cos(rotateAngle * (Math.PI / 180));
@@ -103,7 +103,7 @@ async function ocrToImage(ocrCoords, n, binary = false) {
 
   if (imageN.rotated) {
   // Otherwise, we must also account for rotation applied by the canvas
-    const rotateAngle = (pageMetricsArr[n].angle || 0) * -1;
+    const rotateAngle = (pageMetricsAll[n].angle || 0) * -1;
 
     rotateBoundingBox(ocrCoords, rotateAngle, n);
   }

js/debug.js

@@ -1,5 +1,5 @@
 import { opt } from './containers/app.js';
-import { pageMetricsArr } from './containers/dataContainer.js';
+import { pageMetricsAll } from './containers/dataContainer.js';
 import { ImageCache } from './containers/imageContainer.js';
 import { gs } from './generalWorkerMain.js';
 import { loadImageElem } from './utils/imageUtils.js';
@@ -125,7 +125,7 @@ export async function renderPageStatic(page) {
   const res = gs.renderPageStaticImp({
     page,
     image,
-    angle: pageMetricsArr[page.n].angle,
+    angle: pageMetricsAll[page.n].angle,
   });
 
   return res;

js/export/export.js

@@ -1,11 +1,11 @@
 import { inputData, opt } from '../containers/app.js';
 import {
-  layoutDataTables, layoutRegions, ocrAll, pageMetricsArr,
+  layoutDataTables, layoutRegions, ocrAll, pageMetricsAll,
 } from '../containers/dataContainer.js';
 import { ImageCache } from '../containers/imageContainer.js';
 import { reorderOcrPage } from '../modifyOCR.js';
 import { saveAs } from '../utils/miscUtils.js';
-import { writePdf } from './writePdf.js';
+import { writePdf } from './pdf/writePdf.js';
 import { writeHocr } from './writeHocr.js';
 import { writeText } from './writeText.js';
 import { writeHtml } from './writeHtml.js';
@@ -45,8 +45,8 @@ export async function exportData(format = 'txt', minPage = 0, maxPage = -1) {
     const dimsLimit = { width: -1, height: -1 };
     if (opt.standardizePageSize) {
       for (let i = minPage; i <= maxPage; i++) {
-        dimsLimit.height = Math.max(dimsLimit.height, pageMetricsArr[i].dims.height);
-        dimsLimit.width = Math.max(dimsLimit.width, pageMetricsArr[i].dims.width);
+        dimsLimit.height = Math.max(dimsLimit.height, pageMetricsAll[i].dims.height);
+        dimsLimit.width = Math.max(dimsLimit.width, pageMetricsAll[i].dims.width);
       }
     }
 
@@ -58,10 +58,30 @@ export async function exportData(format = 'txt', minPage = 0, maxPage = -1) {
 
       const rotateText = !rotateBackground;
 
+      const includeImages = false;
+      /** @type {ImageWrapper[]} */
+      let images = [];
+      if (includeImages) {
+        images = await Promise.all(ImageCache.nativeSrc);
+      }
+
       // Page sizes should not be standardized at this step, as the overlayText/overlayTextImage functions will perform this,
       // and assume that the overlay PDF is the same size as the input images.
-      const pdfStr = await writePdf(ocrDownload, minPage, maxPage, opt.displayMode, rotateText, rotateBackground,
-        { width: -1, height: -1 }, opt.confThreshHigh, opt.confThreshMed, opt.overlayOpacity / 100);
+      const pdfStr = await writePdf({
+        ocrArr: ocrDownload,
+        pageMetricsArr: pageMetricsAll,
+        minpage: minPage,
+        maxpage: maxPage,
+        textMode: opt.displayMode,
+        rotateText,
+        rotateBackground,
+        dimsLimit: { width: -1, height: -1 },
+        confThreshHigh: opt.confThreshHigh,
+        confThreshMed: opt.confThreshMed,
+        proofOpacity: opt.overlayOpacity / 100,
+        images,
+        includeImages,
+      });
 
       const enc = new TextEncoder();
       const pdfEnc = enc.encode(pdfStr);
@@ -121,7 +141,7 @@ export async function exportData(format = 'txt', minPage = 0, maxPage = -1) {
 
         await w.convertImageStart({ humanReadable: opt.humanReadablePDF });
         for (let i = minPage; i < maxPage + 1; i++) {
-          /** @type {import('../containers/imageContainer.js').ImageWrapper} */
+          /** @type {ImageWrapper} */
           let image;
           if (binary) {
             image = await ImageCache.getBinary(i, props);
@@ -134,7 +154,7 @@ export async function exportData(format = 'txt', minPage = 0, maxPage = -1) {
           // Angle the PDF viewer is instructed to rotated the image by.
           // This method is currently only used when rotation is needed but the user's (unrotated) source images are being used.
           // If the images are being rendered, then rotation is expected to be applied within the rendering process.
-          const angleImagePdf = rotateBackground && !renderImage ? (pageMetricsArr[i].angle || 0) * -1 : 0;
+          const angleImagePdf = rotateBackground && !renderImage ? (pageMetricsAll[i].angle || 0) * -1 : 0;
 
           await w.convertImageAddPage({
             image: image.src, i, pagewidth: dimsLimit.width, pageheight: dimsLimit.height, angle: angleImagePdf,
@@ -157,8 +177,19 @@ export async function exportData(format = 'txt', minPage = 0, maxPage = -1) {
 
       w.freeDocument(pdfOverlay);
     } else {
-      const pdfStr = await writePdf(ocrDownload, minPage, maxPage, opt.displayMode, false, true, dimsLimit, opt.confThreshHigh, opt.confThreshMed,
-        opt.overlayOpacity / 100);
+      const pdfStr = await writePdf({
+        ocrArr: ocrDownload,
+        pageMetricsArr: pageMetricsAll,
+        minpage: minPage,
+        maxpage: maxPage,
+        textMode: opt.displayMode,
+        rotateText: false,
+        rotateBackground: true,
+        dimsLimit,
+        confThreshHigh: opt.confThreshHigh,
+        confThreshMed: opt.confThreshMed,
+        proofOpacity: opt.overlayOpacity / 100,
+      });
 
       // The PDF is still run through muPDF, even thought in eBook mode no background layer is added.
       // This is because muPDF cleans up the PDF we made in the previous step, including:
@@ -186,7 +217,7 @@ export async function exportData(format = 'txt', minPage = 0, maxPage = -1) {
       w.freeDocument(pdf);
     }
   } else if (format === 'hocr') {
-    content = writeHocr(ocrDownload, minPage, maxPage);
+    content = writeHocr({ ocrData: ocrDownload, minValue: minPage, maxValue: maxPage });
   } else if (format === 'html') {
     const images = /** @type {Array<ImageWrapper>} */ ([]);
     if (opt.includeImages) {
@@ -218,18 +249,29 @@ export async function exportData(format = 'txt', minPage = 0, maxPage = -1) {
       ocrPages: ocrDownload, images, minpage: minPage, maxpage: maxPage, reflowText: opt.reflow, removeMargins: opt.removeMargins,
     });
   } else if (format === 'txt') {
-    content = writeText(ocrDownload, minPage, maxPage, opt.reflow, false);
+    content = writeText({
+      ocrCurrent: ocrDownload,
+      minpage: minPage,
+      maxpage: maxPage,
+      reflowText: opt.reflow,
+      docxMode: false,
+    });
   // Defining `DISABLE_DOCX_XLSX` disables docx/xlsx exports when using build tools.
   // @ts-ignore
   } else if (typeof DISABLE_DOCX_XLSX === 'undefined' && format === 'docx') {
     // Less common export formats are loaded dynamically to reduce initial load time.
     const writeDocx = (await import('./writeDocx.js')).writeDocx;
-    content = await writeDocx(ocrDownload, minPage, maxPage);
+    content = await writeDocx({ hocrCurrent: ocrDownload, minpage: minPage, maxpage: maxPage });
   // @ts-ignore
   } else if (typeof DISABLE_DOCX_XLSX === 'undefined' && format === 'xlsx') {
     // Less common export formats are loaded dynamically to reduce initial load time.
     const writeXlsx = (await import('./writeTabular.js')).writeXlsx;
-    content = await writeXlsx(ocrDownload, layoutDataTables.pages, minPage, maxPage);
+    content = await writeXlsx({
+      ocrPageArr: ocrDownload,
+      layoutPageArr: layoutDataTables.pages,
+      minpage: minPage,
+      maxpage: maxPage,
+    });
   } else if (format === 'scribe') {
     const data = {
       ocr: removeCircularRefsOcr(ocrDownload),

js/export/exportDebugCsv.js

@@ -39,11 +39,12 @@ export const convertToCsv = (data) => {
 
 /**
  *
- * @param {Array<OcrPage>} pages
- * @param {string} fileName
+ * @param {Object} params
+ * @param {Array<OcrPage>} params.pages
+ * @param {string} params.fileName
  * @returns
  */
-export const writeDebugCsv = (pages, fileName) => {
+export const writeDebugCsv = ({ pages, fileName }) => {
   let csvStr = '';
 
   for (let i = 0; i < pages.length; i++) {