@jasy/pdf 1.0.0-alpha.1 → 1.0.0-alpha.11

package/dist/utils/gsub.js

@@ -0,0 +1,178 @@
+import { u16, u32, i16, latin1FromBytes } from "./bytes.js";
+const EXTENSION = 7;
+/** A coverage table may not expand past this: a glyph id is a uint16, so no honest one can. */
+const MAX_COVERAGE_ENTRIES = 65536;
+export class GsubTable {
+    constructor(data, base) {
+        this.data = data;
+        this.base = base;
+        this.lookupCache = new Map();
+        this.coverageCache = new Map();
+        this.scriptList = base + u16(data, base + 4);
+        this.featureList = base + u16(data, base + 6);
+        this.lookupList = base + u16(data, base + 8);
+    }
+    /** Whether the font carries this script at all - `arab` is the question the shaper asks. */
+    hasScript(script) {
+        return this.scriptOffset(script) !== undefined;
+    }
+    /** A feature's lookup indices, under the script's DEFAULT language system - Arabic shaping features
+     *  are script-wide, so a language-specific override would only add typographic niceties. */
+    lookups(script, feature) {
+        const scriptOff = this.scriptOffset(script);
+        if (scriptOff === undefined)
+            return [];
+        // The default LangSys is optional; fall back to the first named one - a font declaring only
+        // `ARA ` still shapes Arabic.
+        const defaultOff = u16(this.data, scriptOff);
+        const langSys = defaultOff !== 0
+            ? scriptOff + defaultOff
+            : u16(this.data, scriptOff + 2) > 0
+                ? // The first LangSysRecord is at +4 and is tag(4) + offset(2), so its offset is at +8.
+                    scriptOff + u16(this.data, scriptOff + 8)
+                : undefined;
+        if (langSys === undefined)
+            return [];
+        const out = [];
+        const count = u16(this.data, langSys + 4);
+        for (let i = 0; i < count; i++) {
+            const featureIndex = u16(this.data, langSys + 6 + i * 2);
+            const record = this.featureList + 2 + featureIndex * 6;
+            if (latin1FromBytes(this.data.subarray(record, record + 4)) !== feature)
+                continue;
+            const featureOff = this.featureList + u16(this.data, record + 4);
+            const lookupCount = u16(this.data, featureOff + 2);
+            for (let j = 0; j < lookupCount; j++)
+                out.push(u16(this.data, featureOff + 4 + j * 2));
+        }
+        return out;
+    }
+    /** Type 1: substitute one glyph. `null` when no subtable covers it - the feature does not apply. */
+    substituteSingle(lookupIndex, glyph) {
+        const lookup = this.lookup(lookupIndex);
+        if (!lookup || lookup.type !== 1)
+            return null;
+        for (const sub of lookup.subtables) {
+            const coverage = this.coverage(sub + u16(this.data, sub + 2));
+            const index = coverage.get(glyph);
+            if (index === undefined)
+                continue;
+            // Format 1 shifts every covered glyph by one delta; format 2 lists a replacement per glyph.
+            return u16(this.data, sub) === 1
+                ? (glyph + i16(this.data, sub + 4)) & 0xffff
+                : u16(this.data, sub + 6 + index * 2);
+        }
+        return null;
+    }
+    /**
+     * Type 4: collapse a run of glyphs into one, starting at `at`. Longest match wins, as the spec
+     * requires - a three-glyph ligature beats the two-glyph one that starts the same way.
+     */
+    substituteLigature(lookupIndex, glyphs, at) {
+        const lookup = this.lookup(lookupIndex);
+        if (!lookup || lookup.type !== 4)
+            return null;
+        let best = null;
+        for (const sub of lookup.subtables) {
+            const coverage = this.coverage(sub + u16(this.data, sub + 2));
+            const setIndex = coverage.get(glyphs[at]);
+            if (setIndex === undefined)
+                continue;
+            const set = sub + u16(this.data, sub + 6 + setIndex * 2);
+            const ligatureCount = u16(this.data, set);
+            for (let i = 0; i < ligatureCount; i++) {
+                const lig = set + u16(this.data, set + 2 + i * 2);
+                const components = u16(this.data, lig + 2); // includes the glyph coverage already matched
+                if (at + components > glyphs.length)
+                    continue;
+                let matches = true;
+                for (let c = 1; c < components && matches; c++) {
+                    matches = glyphs[at + c] === u16(this.data, lig + 2 + c * 2);
+                }
+                if (matches && (best === null || components > best.consumed)) {
+                    best = { glyph: u16(this.data, lig), consumed: components };
+                }
+            }
+        }
+        return best;
+    }
+    /** The type of a lookup, for reporting what a font needs that we do not do. */
+    lookupType(lookupIndex) {
+        return this.lookup(lookupIndex)?.type;
+    }
+    scriptOffset(script) {
+        const count = u16(this.data, this.scriptList);
+        for (let i = 0; i < count; i++) {
+            const record = this.scriptList + 2 + i * 6;
+            if (latin1FromBytes(this.data.subarray(record, record + 4)) === script) {
+                return this.scriptList + u16(this.data, record + 4);
+            }
+        }
+        return undefined;
+    }
+    /**
+     * Resolve a lookup, unwrapping type 7 (extension) - which exists only because the 16-bit offsets
+     * elsewhere cannot reach far enough in a big font. Callers never learn it was wrapped.
+     */
+    lookup(index) {
+        const cached = this.lookupCache.get(index);
+        if (cached !== undefined)
+            return cached;
+        let result = null;
+        if (index < u16(this.data, this.lookupList)) {
+            const off = this.lookupList + u16(this.data, this.lookupList + 2 + index * 2);
+            const type = u16(this.data, off);
+            const count = u16(this.data, off + 4);
+            const subtables = [];
+            let realType = type;
+            for (let i = 0; i < count; i++) {
+                const sub = off + u16(this.data, off + 6 + i * 2);
+                if (type === EXTENSION) {
+                    realType = u16(this.data, sub + 2);
+                    subtables.push(sub + u32(this.data, sub + 4));
+                }
+                else {
+                    subtables.push(sub);
+                }
+            }
+            result = { type: realType, subtables };
+        }
+        this.lookupCache.set(index, result);
+        return result;
+    }
+    /** Coverage table -> glyph id to its INDEX in the table, which is how subtables address glyphs. */
+    coverage(off) {
+        const cached = this.coverageCache.get(off);
+        if (cached)
+            return cached;
+        const map = new Map();
+        const count = u16(this.data, off + 2);
+        if (u16(this.data, off) === 1) {
+            for (let i = 0; i < count; i++)
+                map.set(u16(this.data, off + 4 + i * 2), i);
+        }
+        else {
+            // Format 2 is RANGES, and a malformed file can claim far more glyphs than any font holds. Such
+            // a table is refused whole rather than expanded - the feature then simply does not apply.
+            let total = 0;
+            for (let i = 0; i < count; i++) {
+                const record = off + 4 + i * 6;
+                const start = u16(this.data, record);
+                const end = u16(this.data, record + 2);
+                if (end < start)
+                    continue; // an inverted range covers nothing
+                total += end - start + 1;
+                if (total > MAX_COVERAGE_ENTRIES)
+                    return this.cacheCoverage(off, new Map());
+                const startIndex = u16(this.data, record + 4);
+                for (let g = start; g <= end; g++)
+                    map.set(g, startIndex + (g - start));
+            }
+        }
+        return this.cacheCoverage(off, map);
+    }
+    cacheCoverage(off, map) {
+        this.coverageCache.set(off, map);
+        return map;
+    }
+}

package/dist/utils/image-helper.d.ts

@@ -7,7 +7,7 @@ interface ImageDimensions {
     width: number;
     height: number;
 }
-export declare function getImageDimensions(buffer: Buffer): Promise<ImageDimensions>;
+export declare function getImageDimensions(buffer: Uint8Array): Promise<ImageDimensions>;
 /**
  * Converts the given image to grayscale and returns its binary data.
  * @param imagePath Path to the input image file.
@@ -16,15 +16,16 @@ export declare function getImageDimensions(buffer: Buffer): Promise<ImageDimensi
 /**
  * Decodes a PNG into raw DeviceRGB samples, Flate-compressed, ready to embed as a PDF
  * image XObject. A PNG file is NOT a valid `/FlateDecode` stream (it's a signature plus
- * chunks of filtered, zlib-compressed scanlines), so it must be decoded first. PDF has
- * no alpha channel for DeviceRGB, so transparent pixels are composited over white.
+ * chunks of filtered, zlib-compressed scanlines), so it must be decoded first. DeviceRGB has
+ * no alpha itself, so any transparency rides alongside as a DeviceGray `/SMask` (the return's `smask`).
  */
-export declare function decodePngToRgbFlate(pngBuffer: Buffer): Promise<{
+export declare function decodePngToRgbFlate(pngBuffer: Uint8Array): Promise<{
     data: string;
+    smask?: string;
     width: number;
     height: number;
 }>;
-export declare function convertImageToGrayscaleBuffer(imagePath: string): Promise<Buffer>;
+export declare function convertImageToGrayscaleBuffer(imagePath: string): Promise<Uint8Array>;
 export interface FitResult {
     width: number;
     height: number;

package/dist/utils/image-helper.js

@@ -1,23 +1,6 @@
-"use strict";
-var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
-    function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }
-    return new (P || (P = Promise))(function (resolve, reject) {
-        function fulfilled(value) { try { step(generator.next(value)); } catch (e) { reject(e); } }
-        function rejected(value) { try { step(generator["throw"](value)); } catch (e) { reject(e); } }
-        function step(result) { result.done ? resolve(result.value) : adopt(result.value).then(fulfilled, rejected); }
-        step((generator = generator.apply(thisArg, _arguments || [])).next());
-    });
-};
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.getImageDimensions = getImageDimensions;
-exports.decodePngToRgbFlate = decodePngToRgbFlate;
-exports.convertImageToGrayscaleBuffer = convertImageToGrayscaleBuffer;
-exports.applyContainFit = applyContainFit;
-exports.applyCoverFit = applyCoverFit;
-exports.applyFillFit = applyFillFit;
-exports.applyFitNone = applyFitNone;
-const jimp_1 = require("jimp");
-const zlib_1 = require("zlib");
+import { zlibSync } from "fflate";
+import { latin1FromBytes } from "./bytes.js";
+import { pngToRgba } from "../platform/node-image.js";
 // Implement the method to handle 24-bit integers
 DataView.prototype.getUint24 = function (byteOffset, littleEndian = false) {
     const b1 = this.getUint8(byteOffset);
@@ -25,61 +8,59 @@ DataView.prototype.getUint24 = function (byteOffset, littleEndian = false) {
     const b3 = this.getUint8(byteOffset + 2);
     return littleEndian ? (b3 << 16) | (b2 << 8) | b1 : (b1 << 16) | (b2 << 8) | b3;
 };
-function getImageDimensions(buffer) {
-    return __awaiter(this, void 0, void 0, function* () {
-        const dataView = new DataView(buffer.buffer);
-        // Check for JPEG (0xFFD8 is the start of JPEG file)
-        if (dataView.getUint16(0) === 0xffd8) {
-            let offset = 2;
-            while (offset < buffer.byteLength) {
-                const marker = dataView.getUint16(offset, false);
-                offset += 2;
-                if (marker === 0xffc0 || marker === 0xffc2) {
-                    // SOF0 or SOF2
-                    return {
-                        height: dataView.getUint16(offset + 3, false),
-                        width: dataView.getUint16(offset + 5, false),
-                    };
-                }
-                else {
-                    offset += dataView.getUint16(offset, false);
-                }
+export async function getImageDimensions(buffer) {
+    const dataView = new DataView(buffer.buffer);
+    // Check for JPEG (0xFFD8 is the start of JPEG file)
+    if (dataView.getUint16(0) === 0xffd8) {
+        let offset = 2;
+        while (offset < buffer.byteLength) {
+            const marker = dataView.getUint16(offset, false);
+            offset += 2;
+            if (marker === 0xffc0 || marker === 0xffc2) {
+                // SOF0 or SOF2
+                return {
+                    height: dataView.getUint16(offset + 3, false),
+                    width: dataView.getUint16(offset + 5, false),
+                };
+            }
+            else {
+                offset += dataView.getUint16(offset, false);
             }
         }
-        // Check for PNG (0x89504E47 is the PNG signature)
-        if (dataView.getUint32(0) === 0x89504e47) {
+    }
+    // Check for PNG (0x89504E47 is the PNG signature)
+    if (dataView.getUint32(0) === 0x89504e47) {
+        return {
+            width: dataView.getUint32(16, false),
+            height: dataView.getUint32(20, false),
+        };
+    }
+    // Check for BMP (0x424D is the BMP signature)
+    if (dataView.getUint16(0) === 0x424d) {
+        return {
+            width: dataView.getUint32(18, true),
+            height: dataView.getUint32(22, true),
+        };
+    }
+    // Check for WebP (0x52494646 is the WebP signature 'RIFF')
+    if (dataView.getUint32(0) === 0x52494646 && dataView.getUint32(8) === 0x57454250) {
+        // 'WEBP'
+        if (dataView.getUint32(12) === 0x56503820) {
+            // 'VP8 '
             return {
-                width: dataView.getUint32(16, false),
-                height: dataView.getUint32(20, false),
+                width: dataView.getUint16(26, true),
+                height: dataView.getUint16(28, true),
             };
         }
-        // Check for BMP (0x424D is the BMP signature)
-        if (dataView.getUint16(0) === 0x424d) {
+        else if (dataView.getUint32(12) === 0x56503858) {
+            // 'VP8X'
             return {
-                width: dataView.getUint32(18, true),
-                height: dataView.getUint32(22, true),
+                width: dataView.getUint24(24, true) + 1,
+                height: dataView.getUint24(27, true) + 1,
             };
         }
-        // Check for WebP (0x52494646 is the WebP signature 'RIFF')
-        if (dataView.getUint32(0) === 0x52494646 && dataView.getUint32(8) === 0x57454250) {
-            // 'WEBP'
-            if (dataView.getUint32(12) === 0x56503820) {
-                // 'VP8 '
-                return {
-                    width: dataView.getUint16(26, true),
-                    height: dataView.getUint16(28, true),
-                };
-            }
-            else if (dataView.getUint32(12) === 0x56503858) {
-                // 'VP8X'
-                return {
-                    width: dataView.getUint24(24, true) + 1,
-                    height: dataView.getUint24(27, true) + 1,
-                };
-            }
-        }
-        throw new Error("Unsupported image format");
-    });
+    }
+    throw new Error("Unsupported image format");
 }
 /**
  * Converts the given image to grayscale and returns its binary data.
@@ -89,50 +70,59 @@ function getImageDimensions(buffer) {
 /**
  * Decodes a PNG into raw DeviceRGB samples, Flate-compressed, ready to embed as a PDF
  * image XObject. A PNG file is NOT a valid `/FlateDecode` stream (it's a signature plus
- * chunks of filtered, zlib-compressed scanlines), so it must be decoded first. PDF has
- * no alpha channel for DeviceRGB, so transparent pixels are composited over white.
+ * chunks of filtered, zlib-compressed scanlines), so it must be decoded first. DeviceRGB has
+ * no alpha itself, so any transparency rides alongside as a DeviceGray `/SMask` (the return's `smask`).
  */
-function decodePngToRgbFlate(pngBuffer) {
-    return __awaiter(this, void 0, void 0, function* () {
-        const image = yield jimp_1.Jimp.fromBuffer(pngBuffer);
-        const { width, height, data: rgba } = image.bitmap;
-        const rgb = Buffer.allocUnsafe(width * height * 3);
-        for (let i = 0, j = 0; i < rgba.length; i += 4, j += 3) {
-            const alpha = rgba[i + 3] / 255;
-            rgb[j] = Math.round(rgba[i] * alpha + 255 * (1 - alpha));
-            rgb[j + 1] = Math.round(rgba[i + 1] * alpha + 255 * (1 - alpha));
-            rgb[j + 2] = Math.round(rgba[i + 2] * alpha + 255 * (1 - alpha));
-        }
-        // Compress so the existing `/Filter /FlateDecode` XObject path embeds it correctly.
-        return { data: (0, zlib_1.deflateSync)(rgb).toString("binary"), width, height };
-    });
+export async function decodePngToRgbFlate(pngBuffer) {
+    // The decode (PNG bytes → RGBA) is platform-specific: jimp on Node, an OffscreenCanvas in the browser
+    // (`platform/{node,browser}-image.ts`, swapped by the package `browser` field). The split below is shared.
+    const { width, height, rgba } = await pngToRgba(pngBuffer);
+    // Raw DeviceRGB samples + (only when the PNG actually has transparency) a DeviceGray alpha channel that
+    // rides as the XObject's /SMask. An opaque PNG yields the same raw RGB as before and no SMask, so it stays
+    // byte-identical; a transparent one now composites correctly over whatever sits behind it.
+    const rgb = new Uint8Array(width * height * 3);
+    const alpha = new Uint8Array(width * height);
+    let hasAlpha = false;
+    for (let i = 0, j = 0, k = 0; i < rgba.length; i += 4, j += 3, k++) {
+        rgb[j] = rgba[i];
+        rgb[j + 1] = rgba[i + 1];
+        rgb[j + 2] = rgba[i + 2];
+        alpha[k] = rgba[i + 3];
+        if (rgba[i + 3] !== 255)
+            hasAlpha = true;
+    }
+    return {
+        data: latin1FromBytes(zlibSync(rgb)),
+        smask: hasAlpha ? latin1FromBytes(zlibSync(alpha)) : undefined,
+        width,
+        height,
+    };
 }
-function convertImageToGrayscaleBuffer(imagePath) {
-    return __awaiter(this, void 0, void 0, function* () {
-        // We get the image from the buffer
-        const image = yield jimp_1.Jimp.read(imagePath);
-        // Get MIME type. If emtpy throw error
-        const mime = image.mime;
-        if (!mime)
-            throw new Error("Cannot read MIME type");
-        // We need to check the MIME type (the exact union `getBuffer` accepts).
-        let mimeType;
-        switch (mime) {
-            case jimp_1.JimpMime.png:
-            case jimp_1.JimpMime.jpeg:
-            case jimp_1.JimpMime.bmp:
-                mimeType = mime;
-                break;
-            default:
-                throw new Error("Unsupported MIME type");
-        }
-        image.greyscale();
-        // Convert the image back to buffer with current MIME type
-        const grayscaleBuffer = yield image.getBuffer(mimeType);
-        return grayscaleBuffer;
-    });
+export async function convertImageToGrayscaleBuffer(imagePath) {
+    // We get the image from the buffer (jimp lazily loaded - see decodePngToRgbFlate).
+    const { Jimp, JimpMime } = await import("jimp");
+    const image = await Jimp.read(imagePath);
+    // Get MIME type. If emtpy throw error
+    const mime = image.mime;
+    if (!mime)
+        throw new Error("Cannot read MIME type");
+    // We need to check the MIME type (the exact union `getBuffer` accepts).
+    let mimeType;
+    switch (mime) {
+        case JimpMime.png:
+        case JimpMime.jpeg:
+        case JimpMime.bmp:
+            mimeType = mime;
+            break;
+        default:
+            throw new Error("Unsupported MIME type");
+    }
+    image.greyscale();
+    // Convert the image back to buffer with current MIME type
+    const grayscaleBuffer = await image.getBuffer(mimeType);
+    return grayscaleBuffer;
 }
-function applyContainFit(imageWidth, imageHeight, containerWidth, containerHeight) {
+export function applyContainFit(imageWidth, imageHeight, containerWidth, containerHeight) {
     const imageAspectRatio = imageWidth / imageHeight;
     const containerAspectRatio = containerWidth / containerHeight;
     let width, height, offsetX, offsetY;
@@ -159,7 +149,7 @@ function applyContainFit(imageWidth, imageHeight, containerWidth, containerHeigh
         offsetY,
     };
 }
-function applyCoverFit(imageWidth, imageHeight, containerWidth, containerHeight) {
+export function applyCoverFit(imageWidth, imageHeight, containerWidth, containerHeight) {
     const imageAspectRatio = imageWidth / imageHeight;
     const containerAspectRatio = containerWidth / containerHeight;
     let width, height, offsetX, offsetY;
@@ -186,7 +176,7 @@ function applyCoverFit(imageWidth, imageHeight, containerWidth, containerHeight)
         offsetY,
     };
 }
-function applyFillFit(containerWidth, containerHeight) {
+export function applyFillFit(containerWidth, containerHeight) {
     return {
         width: containerWidth,
         height: containerHeight,
@@ -194,7 +184,7 @@ function applyFillFit(containerWidth, containerHeight) {
         offsetY: 0,
     };
 }
-function applyFitNone(imageWidth, imageHeight, containerWidth, containerHeight) {
+export function applyFitNone(imageWidth, imageHeight, containerWidth, containerHeight) {
     const offsetX = (containerWidth - imageWidth) / 2; // Center horizontally
     const offsetY = (containerHeight - imageHeight) / 2; // Center vertically
     return {

package/dist/utils/inflate.d.ts

@@ -0,0 +1,16 @@
+/**
+ * Inflating with a ceiling, so a hostile or mistaken input cannot exhaust memory.
+ *
+ * Used by BOTH sides: the reader inflating a PDF stream, and the font path unpacking a WOFF table.
+ * It lives in `utils/` for exactly that reason - the generate path must not import from `edit/`.
+ *
+ * fflate's `unzlibSync(data, { out })` is NOT usable for this: it truncates silently instead of failing.
+ * Hence the streamed form - input in slices, so the total can be checked while it grows.
+ */
+/** Thrown when a stream expands past the ceiling; distinct from a stream that simply will not decode. */
+export declare class PdfStreamTooLargeError extends Error {
+    readonly limit: number;
+    constructor(limit: number, message: string);
+}
+export declare const DEFAULT_MAX_STREAM_SIZE: number;
+export declare function inflateBounded(data: Uint8Array, limit: number): Uint8Array;

package/dist/utils/inflate.js

@@ -0,0 +1,52 @@
+import { Unzlib, unzlibSync } from "fflate";
+/**
+ * Inflating with a ceiling, so a hostile or mistaken input cannot exhaust memory.
+ *
+ * Used by BOTH sides: the reader inflating a PDF stream, and the font path unpacking a WOFF table.
+ * It lives in `utils/` for exactly that reason - the generate path must not import from `edit/`.
+ *
+ * fflate's `unzlibSync(data, { out })` is NOT usable for this: it truncates silently instead of failing.
+ * Hence the streamed form - input in slices, so the total can be checked while it grows.
+ */
+/** Thrown when a stream expands past the ceiling; distinct from a stream that simply will not decode. */
+export class PdfStreamTooLargeError extends Error {
+    constructor(limit, message) {
+        super(`@jasy/pdf: ${message}`);
+        this.limit = limit;
+    }
+}
+export const DEFAULT_MAX_STREAM_SIZE = 64 * 1024 * 1024;
+/** Only the granularity of the check: fflate buffers internally, so the abort lands one block past. */
+const INPUT_SLICE = 16 * 1024;
+/** DEFLATE's maximum expansion (a 258-byte match in its shortest encoding). A format property. */
+const MAX_DEFLATE_RATIO = 1032;
+export function inflateBounded(data, limit) {
+    // Too small to reach the ceiling even at maximum compression, so checking as it grows is pointless.
+    // Nearly every stream in a real PDF lands here, which is why the guard costs nothing.
+    if (data.length <= limit / MAX_DEFLATE_RATIO)
+        return unzlibSync(data);
+    const parts = [];
+    let total = 0;
+    const inflater = new Unzlib();
+    inflater.ondata = (chunk) => {
+        total += chunk.length;
+        if (total > limit) {
+            throw new PdfStreamTooLargeError(limit, `a stream in this PDF expands past the ${limit} byte limit for a single stream; if the file is ` +
+                `genuinely that large, raise it with { maxStreamSize }`);
+        }
+        parts.push(chunk);
+    };
+    for (let at = 0; at < data.length; at += INPUT_SLICE) {
+        const end = Math.min(at + INPUT_SLICE, data.length);
+        inflater.push(data.subarray(at, end), end >= data.length);
+    }
+    if (parts.length === 1)
+        return parts[0];
+    const out = new Uint8Array(total);
+    let at = 0;
+    for (const part of parts) {
+        out.set(part, at);
+        at += part.length;
+    }
+    return out;
+}

package/dist/utils/md5.d.ts

@@ -0,0 +1,4 @@
+/** The raw 16-byte MD5 digest of `input`. */
+export declare function md5(input: Uint8Array): Uint8Array;
+/** Lowercase hex of the MD5 digest (matches `digest("hex")`). */
+export declare function md5Hex(input: Uint8Array): string;

package/dist/utils/md5.js

@@ -0,0 +1,79 @@
+// MD5 (RFC 1321), vendored + isomorphic - so the font subset tag and the documentId /ID hash work in the
+// browser too: Node `crypto` is absent there, and Web Crypto has no MD5 and is async. Verified byte-for-byte
+// identical to Node's `createHash("md5")`. Operates on bytes; callers UTF-8-encode strings (TextEncoder),
+// matching `hash.update(string)`'s default encoding.
+// Per-round left-rotate amounts.
+const S = [
+    7, 12, 17, 22, 7, 12, 17, 22, 7, 12, 17, 22, 7, 12, 17, 22, 5, 9, 14, 20, 5, 9, 14, 20, 5, 9, 14,
+    20, 5, 9, 14, 20, 4, 11, 16, 23, 4, 11, 16, 23, 4, 11, 16, 23, 4, 11, 16, 23, 6, 10, 15, 21, 6,
+    10, 15, 21, 6, 10, 15, 21, 6, 10, 15, 21,
+];
+// Per-round constants K[i] = floor(abs(sin(i+1)) * 2^32).
+const K = new Int32Array(64);
+for (let i = 0; i < 64; i++)
+    K[i] = Math.floor(Math.abs(Math.sin(i + 1)) * 2 ** 32) | 0;
+/** The raw 16-byte MD5 digest of `input`. */
+export function md5(input) {
+    const len = input.length;
+    // Pad to a multiple of 64: message + 0x80 + zeros + 64-bit little-endian bit length.
+    const total = (((len + 8) >> 6) + 1) << 6;
+    const b = new Uint8Array(total);
+    b.set(input);
+    b[len] = 0x80;
+    const bits = len * 8;
+    b[total - 8] = bits & 0xff;
+    b[total - 7] = (bits >>> 8) & 0xff;
+    b[total - 6] = (bits >>> 16) & 0xff;
+    b[total - 5] = (bits >>> 24) & 0xff;
+    // High 32 bits of the length stay 0 (inputs are well under 512 MB).
+    let a0 = 0x67452301, b0 = 0xefcdab89, c0 = 0x98badcfe, d0 = 0x10325476;
+    const M = new Int32Array(16);
+    for (let off = 0; off < total; off += 64) {
+        for (let i = 0; i < 16; i++) {
+            const j = off + i * 4;
+            M[i] = b[j] | (b[j + 1] << 8) | (b[j + 2] << 16) | (b[j + 3] << 24);
+        }
+        let A = a0, B = b0, C = c0, D = d0;
+        for (let i = 0; i < 64; i++) {
+            let f, g;
+            if (i < 16) {
+                f = (B & C) | (~B & D);
+                g = i;
+            }
+            else if (i < 32) {
+                f = (D & B) | (~D & C);
+                g = (5 * i + 1) % 16;
+            }
+            else if (i < 48) {
+                f = B ^ C ^ D;
+                g = (3 * i + 5) % 16;
+            }
+            else {
+                f = C ^ (B | ~D);
+                g = (7 * i) % 16;
+            }
+            f = (f + A + K[i] + M[g]) | 0;
+            A = D;
+            D = C;
+            C = B;
+            B = (B + ((f << S[i]) | (f >>> (32 - S[i])) | 0)) | 0;
+        }
+        a0 = (a0 + A) | 0;
+        b0 = (b0 + B) | 0;
+        c0 = (c0 + C) | 0;
+        d0 = (d0 + D) | 0;
+    }
+    const out = new Uint8Array(16);
+    const w = [a0, b0, c0, d0];
+    for (let i = 0; i < 4; i++) {
+        out[i * 4] = w[i] & 0xff;
+        out[i * 4 + 1] = (w[i] >>> 8) & 0xff;
+        out[i * 4 + 2] = (w[i] >>> 16) & 0xff;
+        out[i * 4 + 3] = (w[i] >>> 24) & 0xff;
+    }
+    return out;
+}
+/** Lowercase hex of the MD5 digest (matches `digest("hex")`). */
+export function md5Hex(input) {
+    return [...md5(input)].map((x) => x.toString(16).padStart(2, "0")).join("");
+}

package/dist/utils/outline.d.ts

@@ -0,0 +1,21 @@
+import type { PDFObjectManager } from "./pdf-object-manager.js";
+interface Entry {
+    title: string;
+    level: number;
+    pageObjNum: number;
+    top: number;
+}
+/**
+ * Builds the PDF document outline (`/Outlines`) - the viewer's bookmark panel. Entries are collected
+ * flat in reading order with a 1-based `level`; `finalize` nests each entry under the nearest preceding
+ * entry of a smaller level (so level 2s hang under the last level 1), emits the outline objects and
+ * returns the catalog addition `/Outlines N 0 R`. A no-op returning "" when nothing was added, so a
+ * document without bookmarks stays byte-identical.
+ */
+export declare class OutlineBuilder {
+    private entries;
+    add(entry: Entry): void;
+    get isEmpty(): boolean;
+    finalize(om: PDFObjectManager): string;
+}
+export {};