@embedpdf/engines 1.0.4 → 1.0.6

package/dist/pdfium.d.ts

@@ -1,4 +1,4 @@
-import { SearchTarget, PdfImage, PdfEngine, Logger, Task, PdfErrorReason, PdfFileUrl, PdfUrlOptions, PdfDocumentObject, PdfFile, PdfFileLoader, PdfSignatureObject, PdfBookmarkObject, PdfPageObject, Rotation, PdfRenderOptions, PdfTask, Rect, PdfAnnotationObject, PdfAnnotationTransformation, PdfTextRectObject, PdfAttachmentObject, PdfWidgetAnnoObject, FormFieldValue, PdfPageFlattenFlag, PdfPageFlattenResult, PdfInkListObject, PdfStampAnnoObjectContents, Position, PdfPageGeometry, PdfGlyphObject, MatchFlag, SearchAllPagesResult, PdfEngineMethodName, PdfEngineMethodArgs, TaskReturn, PdfEngineMethodReturnType } from '@embedpdf/models';
+import { SearchTarget, PdfImage, ImageConversionTypes, PdfEngine, Logger, Task, PdfErrorReason, PdfFileUrl, PdfUrlOptions, PdfDocumentObject, PdfFile, PdfFileLoader, PdfSignatureObject, PdfBookmarkObject, PdfPageObject, Rotation, PdfRenderOptions, PdfTask, Rect, PdfAnnotationObject, PdfAnnotationTransformation, PdfTextRectObject, PdfAttachmentObject, PdfWidgetAnnoObject, FormFieldValue, PdfPageFlattenFlag, PdfPageFlattenResult, PageTextSlice, PdfInkListObject, PdfStampAnnoObjectContents, Position, PdfPageGeometry, PdfGlyphObject, MatchFlag, SearchAllPagesResult, PdfEngineMethodName, PdfEngineMethodArgs, TaskReturn, PdfEngineMethodReturnType } from '@embedpdf/models';
 import { WrappedPdfiumModule, PdfiumRuntimeMethods, PdfiumModule } from '@embedpdf/pdfium';
 
 /**
@@ -61,7 +61,7 @@ declare enum PdfiumErrorCode {
  * In browser: uses OffscreenCanvas
  * In Node.js: can use Sharp or other image processing libraries
  */
-type ImageDataConverter<T = Blob> = (imageData: PdfImage) => Promise<T>;
+type ImageDataConverter<T = Blob> = (imageData: PdfImage, imageType?: ImageConversionTypes) => Promise<T>;
 declare const browserImageDataToBlobConverter: ImageDataConverter<Blob>;
 /**
  * Pdf engine that based on pdfium wasm
@@ -183,13 +183,13 @@ declare class PdfiumEngine<T = Blob> implements PdfEngine<T> {
      *
      * @public
      */
-    renderPage(doc: PdfDocumentObject, page: PdfPageObject, scaleFactor?: number, rotation?: Rotation, dpr?: number, options?: PdfRenderOptions): PdfTask<T>;
+    renderPage(doc: PdfDocumentObject, page: PdfPageObject, scaleFactor?: number, rotation?: Rotation, dpr?: number, options?: PdfRenderOptions, imageType?: ImageConversionTypes): PdfTask<T>;
     /**
      * {@inheritDoc @embedpdf/models!PdfEngine.renderPageRect}
      *
      * @public
      */
-    renderPageRect(doc: PdfDocumentObject, page: PdfPageObject, scaleFactor: number, rotation: Rotation, dpr: number, rect: Rect, options: PdfRenderOptions): PdfTask<T>;
+    renderPageRect(doc: PdfDocumentObject, page: PdfPageObject, scaleFactor: number, rotation: Rotation, dpr: number, rect: Rect, options: PdfRenderOptions, imageType?: ImageConversionTypes): PdfTask<T>;
     /**
      * {@inheritDoc @embedpdf/models!PdfEngine.getAllAnnotations}
      *
@@ -269,6 +269,12 @@ declare class PdfiumEngine<T = Blob> implements PdfEngine<T> {
      * @public
      */
     extractText(doc: PdfDocumentObject, pageIndexes: number[]): Task<any, PdfErrorReason> | Task<string, PdfErrorReason>;
+    /**
+     * {@inheritDoc @embedpdf/models!PdfEngine.getTextSlices}
+     *
+     * @public
+     */
+    getTextSlices(doc: PdfDocumentObject, slices: PageTextSlice[]): PdfTask<string[]>;
     /**
      * {@inheritDoc @embedpdf/models!PdfEngine.merge}
      *

package/dist/pdfium.js

@@ -1,4 +1,4 @@
-import { NoopLogger, PdfTaskHelper, PdfErrorCode, Task, Rotation, PdfAnnotationSubtype, PdfPageObjectType, PdfAnnotationObjectStatus, quadToRect, PDF_FORM_FIELD_TYPE, toIntRect, transformRect, toIntSize, transformSize, PdfActionType, PdfZoomMode, AppearanceMode, MatchFlag } from '@embedpdf/models';
+import { NoopLogger, PdfTaskHelper, PdfErrorCode, Task, Rotation, PdfAnnotationSubtype, stripPdfUnwantedMarkers, PdfPageObjectType, PdfAnnotationObjectStatus, quadToRect, PDF_FORM_FIELD_TYPE, toIntRect, transformRect, toIntSize, transformSize, PdfActionType, PdfZoomMode, AppearanceMode, MatchFlag } from '@embedpdf/models';
 import { init } from '@embedpdf/pdfium';
 
 /**
@@ -265,7 +265,7 @@ var PdfiumErrorCode;
     PdfiumErrorCode[PdfiumErrorCode["XFALoad"] = 7] = "XFALoad";
     PdfiumErrorCode[PdfiumErrorCode["XFALayout"] = 8] = "XFALayout";
 })(PdfiumErrorCode || (PdfiumErrorCode = {}));
-const browserImageDataToBlobConverter = (pdfImageData) => {
+const browserImageDataToBlobConverter = (pdfImageData, imageType = 'image/webp') => {
     // Check if we're in a browser environment
     if (typeof OffscreenCanvas === 'undefined') {
         throw new Error('OffscreenCanvas is not available in this environment. ' +
@@ -275,7 +275,7 @@ const browserImageDataToBlobConverter = (pdfImageData) => {
     const imageData = new ImageData(pdfImageData.data, pdfImageData.width, pdfImageData.height);
     const off = new OffscreenCanvas(imageData.width, imageData.height);
     off.getContext('2d').putImageData(imageData, 0, 0);
-    return off.convertToBlob({ type: 'image/webp' });
+    return off.convertToBlob({ type: imageType });
 };
 /**
  * Pdf engine that based on pdfium wasm
@@ -781,7 +781,7 @@ class PdfiumEngine {
      *
      * @public
      */
-    renderPage(doc, page, scaleFactor = 1, rotation = Rotation.Degree0, dpr = 1, options = { withAnnotations: false }) {
+    renderPage(doc, page, scaleFactor = 1, rotation = Rotation.Degree0, dpr = 1, options = { withAnnotations: false }, imageType = 'image/webp') {
         const task = new Task();
         this.logger.debug(LOG_SOURCE$1, LOG_CATEGORY$1, 'renderPage', doc, page, scaleFactor, rotation, dpr, options);
         this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, `RenderPage`, 'Begin', `${doc.id}-${page.index}`);
@@ -798,7 +798,7 @@ class PdfiumEngine {
             size: page.size,
         }, scaleFactor, rotation, dpr, options);
         this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, `RenderPage`, 'End', `${doc.id}-${page.index}`);
-        this.imageDataConverter(imageData).then((blob) => task.resolve(blob));
+        this.imageDataConverter(imageData, imageType).then((blob) => task.resolve(blob));
         return task;
     }
     /**
@@ -806,7 +806,7 @@ class PdfiumEngine {
      *
      * @public
      */
-    renderPageRect(doc, page, scaleFactor, rotation, dpr, rect, options) {
+    renderPageRect(doc, page, scaleFactor, rotation, dpr, rect, options, imageType = 'image/webp') {
         const task = new Task();
         this.logger.debug(LOG_SOURCE$1, LOG_CATEGORY$1, 'renderPageRect', doc, page, scaleFactor, rotation, dpr, rect, options);
         this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, `RenderPageRect`, 'Begin', `${doc.id}-${page.index}`);
@@ -820,7 +820,7 @@ class PdfiumEngine {
         }
         const imageData = this.renderPageRectToImageData(ctx, page, rect, scaleFactor, rotation, dpr, options);
         this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, `RenderPageRect`, 'End', `${doc.id}-${page.index}`);
-        this.imageDataConverter(imageData).then((blob) => task.resolve(blob));
+        this.imageDataConverter(imageData, imageType).then((blob) => task.resolve(blob));
         return task;
     }
     /**
@@ -1362,6 +1362,62 @@ class PdfiumEngine {
         this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, `ExtractText`, 'End', doc.id);
         return PdfTaskHelper.resolve(text);
     }
+    /**
+     * {@inheritDoc @embedpdf/models!PdfEngine.getTextSlices}
+     *
+     * @public
+     */
+    getTextSlices(doc, slices) {
+        this.logger.debug(LOG_SOURCE$1, LOG_CATEGORY$1, 'getTextSlices', doc, slices);
+        this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, 'GetTextSlices', 'Begin', doc.id);
+        /* ⚠︎ 1 — trivial case */
+        if (slices.length === 0) {
+            this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, 'GetTextSlices', 'End', doc.id);
+            return PdfTaskHelper.resolve([]);
+        }
+        /* ⚠︎ 2 — document must be open */
+        const ctx = this.cache.getContext(doc.id);
+        if (!ctx) {
+            this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, 'GetTextSlices', 'End', doc.id);
+            return PdfTaskHelper.reject({
+                code: PdfErrorCode.DocNotOpen,
+                message: 'document does not open',
+            });
+        }
+        try {
+            /* keep caller order */
+            const out = new Array(slices.length);
+            /* group → open each page once */
+            const byPage = new Map();
+            slices.forEach((s, i) => {
+                (byPage.get(s.pageIndex) ?? byPage.set(s.pageIndex, []).get(s.pageIndex)).push({
+                    slice: s,
+                    pos: i,
+                });
+            });
+            for (const [pageIdx, list] of byPage) {
+                const pageCtx = ctx.acquirePage(pageIdx);
+                const textPagePtr = pageCtx.getTextPage();
+                for (const { slice, pos } of list) {
+                    const bufPtr = this.malloc(2 * (slice.charCount + 1)); // UTF-16 + NIL
+                    this.pdfiumModule.FPDFText_GetText(textPagePtr, slice.charIndex, slice.charCount, bufPtr);
+                    out[pos] = stripPdfUnwantedMarkers(this.pdfiumModule.pdfium.UTF16ToString(bufPtr));
+                    this.free(bufPtr);
+                }
+                pageCtx.release();
+            }
+            this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, 'GetTextSlices', 'End', doc.id);
+            return PdfTaskHelper.resolve(out);
+        }
+        catch (e) {
+            this.logger.error(LOG_SOURCE$1, LOG_CATEGORY$1, 'getTextSlices error', e);
+            this.logger.perf(LOG_SOURCE$1, LOG_CATEGORY$1, 'GetTextSlices', 'End', doc.id);
+            return PdfTaskHelper.reject({
+                code: PdfErrorCode.Unknown,
+                message: String(e),
+            });
+        }
+    }
     /**
      * {@inheritDoc @embedpdf/models!PdfEngine.merge}
      *
@@ -1891,14 +1947,12 @@ class PdfiumEngine {
         const runs = [];
         let current = null;
         let curObjPtr = null;
+        let bounds = null;
         /** ── main loop ──────────────────────────────────────────── */
         for (let i = 0; i < glyphs.length; i++) {
             const g = glyphs[i];
             /* 1 — find the CPDF_TextObject this glyph belongs to */
             const objPtr = this.pdfiumModule.FPDFText_GetTextObject(textPagePtr, i);
-            if (g.isEmpty) {
-                continue;
-            }
             /* 2 — start a new run when the text object changes */
             if (objPtr !== curObjPtr) {
                 curObjPtr = objPtr;
@@ -1912,6 +1966,12 @@ class PdfiumEngine {
                     charStart: i,
                     glyphs: [],
                 };
+                bounds = {
+                    minX: g.origin.x,
+                    minY: g.origin.y,
+                    maxX: g.origin.x + g.size.width,
+                    maxY: g.origin.y + g.size.height,
+                };
                 runs.push(current);
             }
             /* 3 — append the slim glyph record */
@@ -1920,16 +1980,24 @@ class PdfiumEngine {
                 y: g.origin.y,
                 width: g.size.width,
                 height: g.size.height,
-                flags: g.isSpace ? 1 : 0,
+                flags: g.isEmpty ? 2 : g.isSpace ? 1 : 0,
             });
             /* 4 — expand the run's bounding rect */
+            if (g.isEmpty) {
+                continue;
+            }
             const right = g.origin.x + g.size.width;
             const bottom = g.origin.y + g.size.height;
-            current.rect.width =
-                Math.max(current.rect.x + current.rect.width, right) - current.rect.x;
-            current.rect.y = Math.min(current.rect.y, g.origin.y);
-            current.rect.height =
-                Math.max(current.rect.y + current.rect.height, bottom) - current.rect.y;
+            // Update bounds
+            bounds.minX = Math.min(bounds.minX, g.origin.x);
+            bounds.minY = Math.min(bounds.minY, g.origin.y);
+            bounds.maxX = Math.max(bounds.maxX, right);
+            bounds.maxY = Math.max(bounds.maxY, bottom);
+            // Calculate final rect from bounds
+            current.rect.x = bounds.minX;
+            current.rect.y = bounds.minY;
+            current.rect.width = bounds.maxX - bounds.minX;
+            current.rect.height = bounds.maxY - bounds.minY;
         }
         return runs;
     }
@@ -4175,6 +4243,9 @@ class EngineRunner {
                 case 'extractText':
                     task = this.engine[name](...args);
                     break;
+                case 'getTextSlices':
+                    task = this.engine[name](...args);
+                    break;
                 case 'getPageGlyphs':
                     task = this.engine[name](...args);
                     break;