reamkit 1.29.0 → 1.30.0

package/dist/esm/core/document-model/types.d.ts

@@ -1127,6 +1127,8 @@ export interface ShapeLine {
     readonly customDash?: ReadonlyArray<number>;
     readonly cap?: 'flat' | 'round' | 'square';
     readonly fill?: 'solid' | 'none';
+    /** §20.1.2.3.1 `a:alpha` on the line's colour — how opaque the stroke is, `0..1`. */
+    readonly alpha?: number;
     /** §20.1.8.24 `a:headEnd` — the decoration at the line's first point. */
     readonly headEnd?: LineEnd;
     /** §20.1.8.42 `a:tailEnd` — the decoration at the line's last point. */

package/dist/esm/core/drawingml/shape-render.js

@@ -192,13 +192,25 @@ function buildStroke(line) {
 	const widthPt = line.width ?? DEFAULT_LINE_WIDTH_EMU / 12700;
 	const dash = line.customDash?.map((n) => Math.max(.01, n * widthPt)) ?? (line.dash && line.dash !== "solid" ? dashPattern(line.dash, widthPt) : void 0);
 	const cap = line.cap === "flat" ? "butt" : line.cap;
+	const colorHex = line.colorHex ?? "000000";
 	return {
-		colorHex: line.colorHex ?? "000000",
+		colorHex: line.alpha !== void 0 && line.alpha < 1 ? overWhite(colorHex, line.alpha) : colorHex,
 		widthPt,
 		...dash ? { dash } : {},
 		...cap ? { cap } : {}
 	};
 }
+/** A 6-hex colour drawn at `alpha` over white, as the opaque colour it comes to. */
+function overWhite(hex, alpha) {
+	const n = parseInt(hex, 16);
+	if (Number.isNaN(n)) return hex;
+	const a = Math.max(0, Math.min(1, alpha));
+	return [
+		16,
+		8,
+		0
+	].map((shift) => Math.round(255 - (255 - (n >> shift & 255)) * a).toString(16).padStart(2, "0")).join("").toUpperCase();
+}
 function dashPattern(dash, w) {
 	const u = Math.max(w, .1);
 	switch (dash) {

package/dist/esm/core/fonts/index.d.ts

@@ -1,3 +1,3 @@
-export { fetchFontSet, fetchScriptFont, isScriptKey, resolveFamilyKey, resolveFamilyStyle, clearFontCache, } from './remote-fonts.js';
+export { fetchFontSet, fetchScriptFont, isScriptKey, knowsFamily, resolveFamilyKey, resolveFamilyStyle, clearFontCache, } from './remote-fonts.js';
 export { scriptForCodepoint, scriptsInFlow } from './scripts.js';
 export type { FamilyKey, FamilyStyle, FetchFontSetOptions, FetchLike, ScriptKey, SubstituteKey, } from './remote-fonts.js';

package/dist/esm/core/fonts/remote-fonts.d.ts

@@ -52,6 +52,14 @@ export interface FamilyStyle {
  * @returns The chosen family and what the name said about the face.
  */
 export declare function resolveFamilyStyle(name: string | undefined): FamilyStyle;
+/**
+ * Whether the tables KNOW a family name — so the substitute it maps to is the
+ * name's own answer, and not the sans that every name nobody knows falls to.
+ *
+ * @param name The referenced family or face name.
+ * @returns `true` where the name matched a twin, a class or a stem.
+ */
+export declare function knowsFamily(name: string | undefined): boolean;
 /**
  * The curated substitute for a family name — {@link resolveFamilyStyle} without
  * the face it also carries.

package/dist/esm/core/fonts/remote-fonts.js

@@ -98,7 +98,24 @@ var SERIF = new Set([
 	"minion pro",
 	"serif"
 ]);
+var SANS = new Set([
+	"roboto",
+	"noto sans",
+	"notosans",
+	"dejavu sans",
+	"dejavusans",
+	"lucida sans",
+	"lucidasans",
+	"ubuntu",
+	"lato",
+	"inter",
+	"din",
+	"sans",
+	"sans serif"
+]);
 var MONO = new Set([
+	"mono",
+	"monospaced",
 	"consolas",
 	"monaco",
 	"menlo",
@@ -138,14 +155,59 @@ var STEMS = [
 	["charter", "tinos"],
 	["baskerville", "tinos"],
 	["caslon", "tinos"],
-	["minionpro", "tinos"],
+	["minion", "tinos"],
+	["stix", "tinos"],
 	["stoneserif", "tinos"],
 	["stonesans", "arimo"],
-	["myriadpro", "arimo"]
+	["myriad", "arimo"],
+	["arial", "arimo"],
+	["helvetica", "arimo"],
+	["frutiger", "arimo"],
+	["univers", "arimo"],
+	["futura", "arimo"],
+	["gillsans", "arimo"],
+	["optima", "arimo"],
+	["avenir", "arimo"],
+	["verdana", "arimo"],
+	["tahoma", "arimo"],
+	["trebuchet", "arimo"],
+	["segoeui", "arimo"],
+	["franklingothic", "arimo"],
+	["newsgothic", "arimo"],
+	["tradegothic", "arimo"],
+	["centurygothic", "arimo"],
+	["akzidenz", "arimo"],
+	["eurostile", "arimo"],
+	["avantgarde", "arimo"],
+	["gotham", "arimo"],
+	["proximanova", "arimo"],
+	["opensans", "arimo"],
+	["sourcesans", "arimo"],
+	["ptsans", "arimo"],
+	["firasans", "arimo"],
+	["montserrat", "arimo"],
+	["candara", "arimo"],
+	["corbel", "arimo"],
+	["lucidagrande", "arimo"],
+	["mssansserif", "arimo"],
+	["microsoftsansserif", "arimo"],
+	["dinpro", "arimo"],
+	["dinnext", "arimo"],
+	["monospace", "cousine"],
+	["lucidatypewriter", "cousine"],
+	["lucidasanstypewriter", "cousine"]
 ];
 /** The family a name STARTS with, for the families named by stem and size. */
 function familyFromStem(name) {
-	for (const [stem, key] of STEMS) if (name.startsWith(stem)) return key;
+	for (const [stem, key] of STEMS) {
+		if (!name.startsWith(stem)) continue;
+		if (key === "arimo") {
+			const cut = name.slice(stem.length);
+			if (/mono|typewriter/u.test(cut)) return "cousine";
+			if (/serif|slab/u.test(cut)) return "tinos";
+		}
+		return key;
+	}
 }
 var WEIGHT_WORDS = new Map([
 	["black", "bold"],
@@ -185,6 +247,26 @@ var NARROW_SCALE = .82;
 */
 function resolveFamilyStyle(name) {
 	if (!name) return { key: "arimo" };
+	const { words, bold, italic, narrow } = readName(name);
+	return {
+		key: familyOfWords(words) ?? "arimo",
+		...bold ? { bold } : {},
+		...italic ? { italic } : {},
+		...narrow ? { widthScale: NARROW_SCALE } : {}
+	};
+}
+/**
+* Whether the tables KNOW a family name — so the substitute it maps to is the
+* name's own answer, and not the sans that every name nobody knows falls to.
+*
+* @param name The referenced family or face name.
+* @returns `true` where the name matched a twin, a class or a stem.
+*/
+function knowsFamily(name) {
+	return name !== void 0 && name !== "" && familyOfWords(readName(name).words) !== void 0;
+}
+/** A name's family words, with the face words that ended it taken off and read. */
+function readName(name) {
 	const words = name.trim().toLowerCase().split(/[\s\-_,]+/u).filter((w) => w !== "");
 	let bold = false;
 	let italic = false;
@@ -197,25 +279,26 @@ function resolveFamilyStyle(name) {
 		if (word === "narrow") narrow = true;
 		words.pop();
 	}
+	return {
+		words,
+		bold,
+		italic,
+		narrow
+	};
+}
+/** The substitute a family's words name, or `undefined` where no table knows them. */
+function familyOfWords(words) {
+	if (words.length === 0) return void 0;
+	const whole = words.join(" ");
 	const tries = [
-		words.join(" "),
+		whole,
 		words[words.length - 1],
 		words[0]
 	];
-	let key = "arimo";
 	for (const n of tries) {
-		const found = EXACT[n] ?? (MONO.has(n) ? "cousine" : SERIF.has(n) ? "tinos" : void 0) ?? familyFromStem(n.replace(/[^a-z]/gu, ""));
-		if (found) {
-			key = found;
-			break;
-		}
+		const found = EXACT[n] ?? (MONO.has(n) ? "cousine" : SERIF.has(n) ? "tinos" : void 0) ?? (n === whole && SANS.has(n) ? "arimo" : void 0) ?? familyFromStem(n.replace(/[^a-z]/gu, ""));
+		if (found) return found;
 	}
-	return {
-		key,
-		...bold ? { bold } : {},
-		...italic ? { italic } : {},
-		...narrow ? { widthScale: NARROW_SCALE } : {}
-	};
 }
 /**
 * The curated substitute for a family name — {@link resolveFamilyStyle} without
@@ -279,4 +362,4 @@ async function fetchFontSet(options = {}) {
 	};
 }
 //#endregion
-export { fetchFontSet, fetchScriptFont, resolveFamilyKey, resolveFamilyStyle };
+export { fetchFontSet, fetchScriptFont, knowsFamily, resolveFamilyKey, resolveFamilyStyle };

package/dist/esm/core/ir/flow.d.ts

@@ -1,6 +1,15 @@
 import { BodyElement, Chart, Comment, DocumentInfo, Numbering, Section, SectionProperties, ShapeFill, StyleSheet } from '../document-model/index.js';
 import { FontRegistry } from '../font/index.js';
 import { ResourceStore } from './resources.js';
+/**
+ * A face's family as a word processor names it (ECMA-376 §17.8.3.9 `w:font`).
+ */
+export interface FaceFamily {
+    /** The family's own name: `Inter` for `Inter-SemiBold`, `Arial` for `ArialMT`. */
+    readonly family: string;
+    /** §17.8.3.10 `w:family` — the kind of face, for a reader that has to substitute. */
+    readonly generic: 'roman' | 'swiss' | 'modern';
+}
 /**
  * The semantic IR tree (ir-design §5): everything a reader extracts from the
  * document bytes, format-neutrally — the flow `body` plus its document-scoped
@@ -42,6 +51,14 @@ export interface FlowDoc {
     readonly resources: ResourceStore;
     /** Fonts embedded in the source document itself (docx fontTable), by name. */
     readonly embeddedFonts?: ReadonlyMap<string, FontRegistry>;
+    /**
+     * The family each face a run names belongs to, where the two are not the
+     * same name. A PDF names a FACE — `Inter-SemiBold` — where a word processor
+     * names a family and states the weight beside it (`Inter`, bold). The layout
+     * finds a document's own program by the face; a writer that hands the text
+     * to another program names the family, which is the name that program knows.
+     */
+    readonly faceFamilies?: ReadonlyMap<string, FaceFamily>;
     /** Document metadata from docProps/core.xml. */
     readonly info?: DocumentInfo;
     /** Document natural language hint (BCP-47), e.g. for tagged-PDF /Lang. */

package/dist/esm/index.d.ts

@@ -34,7 +34,7 @@ export { getHyphenator, createLanguageHyphenator, createHyphenator, splitPattern
 export type { Hyphenator, HyphenatorOptions, SupportedLanguage } from './core/hyphenation/index.js';
 export type { Pt, ResourceId, Feature, KnownFeature, Loss, LossReport, LossSeverity, NativeBag, } from './core/ir/index.js';
 export { FEATURES, ResourceStore, ConversionLossError, formatLoss, pt, twipsToPt, halfPtToPt, eighthPtToPt, emuToPt, pxToPt, inchToPt, mmToPt, } from './core/ir/index.js';
-export type { FlowDoc } from './core/ir/flow.js';
+export type { FaceFamily, FlowDoc } from './core/ir/flow.js';
 export type { DocumentReader, DocumentWriter, ReadOptions, ReadResult, WriteOptions, WriteResult, } from './core/ir/adapters.js';
 export { docxReader, readDocx } from './word/docx-reader.js';
 export { xlsxReader, readXlsx } from './excel/xlsx-reader.js';

package/dist/esm/pdf-reader/annot-draw.js

@@ -1,4 +1,6 @@
-import { PDF_NULL, PdfName, PdfStream } from "../pdf/objects.js";
+import { PDF_NULL, PdfHexString, PdfName, PdfStream } from "../pdf/objects.js";
+import { reorderVisual } from "../core/bidi/algorithm.js";
+import { analyzeString, hasBidiCharacters } from "../core/bidi/index.js";
 //#region src/pdf-reader/annot-draw.ts
 /** The circle constant: how far a Bézier handle reaches to round a quarter. */
 var KAPPA = .5523;
@@ -20,6 +22,7 @@ function drawnAppearance(file, annot) {
 	const subtype = file.get(annot, "Subtype");
 	if (!(subtype instanceof PdfName)) return void 0;
 	if (subtype.value === "Widget") return typedValue(file, annot) ?? checkedBox(file, annot);
+	if (subtype.value === "FreeText") return freeText(file, annot);
 	const markup = markupShape(file, annot, subtype.value);
 	if (markup) return markup;
 	const pen = borderWidth(file, annot);
@@ -162,6 +165,10 @@ var SQUIGGLE_AMP_EM = .07;
 */
 function drawnResources(file, annot) {
 	const subtype = file.get(annot, "Subtype");
+	if (subtype instanceof PdfName && subtype.value === "FreeText") {
+		const text = freeTextContents(file, annot);
+		return text === void 0 ? void 0 : new Map([["Font", new Map([[FREE_TEXT_FONT, spelledFace(text)]])]]);
+	}
 	if (!(subtype instanceof PdfName) || subtype.value !== "Widget") return void 0;
 	const acro = file.get(file.catalog, "AcroForm");
 	if (!(acro instanceof Map)) return void 0;
@@ -211,6 +218,91 @@ function typedValue(file, annot) {
 	ops.push("ET", "Q");
 	return new PdfStream(/* @__PURE__ */ new Map(), new TextEncoder().encode(ops.join("\n")));
 }
+/**
+* §12.5.6.6 — a free-text annotation with no appearance: its contents, a line
+* to each line break, from the top of its `/Rect` in the size and colour its
+* `/DA` asks. Left undrawn the words were simply gone: bug1865341.pdf's one
+* note, "Załącznik", and nothing else on the page.
+*/
+function freeText(file, annot) {
+	const text = freeTextContents(file, annot);
+	const rect = rectangle(file.get(annot, "Rect"));
+	if (text === void 0 || !rect) return void 0;
+	const da = defaultAppearance(file, annot);
+	const box = inset(rect, 2);
+	const size = da.size > 0 ? da.size : 12;
+	const codes = codesOf(text);
+	const ops = [
+		"q",
+		"BT",
+		`/${FREE_TEXT_FONT} ${num(size)} Tf`,
+		`${da.color} rg`
+	];
+	text.split(/\r\n|[\r\n]/u).forEach((line, i) => {
+		const y = box[3] - size * (.8 + i * 1.16);
+		const hex = visualOrder(line).map((ch) => (codes.get(ch) ?? 0).toString(16).padStart(2, "0")).join("");
+		ops.push(`1 0 0 1 ${num(box[0])} ${num(y)} Tm`, `<${hex}> Tj`);
+	});
+	ops.push("ET", "Q");
+	return new PdfStream(/* @__PURE__ */ new Map(), new TextEncoder().encode(ops.join("\n")));
+}
+/** UAX #9 — a line's characters in the order they stand on the page, left to right. */
+function visualOrder(line) {
+	const chars = [...line];
+	if (!hasBidiCharacters(line)) return chars;
+	const { levels } = analyzeString(line);
+	return reorderVisual(levels).map((i) => chars[i] ?? "");
+}
+/** The resource name the face of a drawn free-text annotation goes by. */
+var FREE_TEXT_FONT = "FreeText";
+/** §7.9.2.2 — an annotation's `/Contents`, decoded as the text string it is. */
+function freeTextContents(file, annot) {
+	const found = file.get(annot, "Contents");
+	const raw = found instanceof PdfHexString ? String.fromCharCode(...found.bytes) : typeof found === "string" ? found : void 0;
+	if (raw === void 0) return void 0;
+	const text = textString(raw).replace(/\s+$/u, "");
+	return text.length > 0 ? text : void 0;
+}
+/**
+* §7.9.2.2 — a text string's characters: UTF-16BE behind its byte-order mark,
+* UTF-8 behind its own, and otherwise one byte to a character.
+*/
+function textString(raw) {
+	if (raw.startsWith("þÿ")) {
+		let out = "";
+		for (let i = 2; i + 1 < raw.length; i += 2) out += String.fromCharCode(raw.charCodeAt(i) << 8 | raw.charCodeAt(i + 1));
+		return out;
+	}
+	if (raw.startsWith("")) return new TextDecoder("utf-8").decode(Uint8Array.from([...raw.slice(3)].map((c) => c.charCodeAt(0))));
+	return raw;
+}
+/**
+* A code for each character a text shows, in the order it first shows them —
+* so a simple face can set a text no standard encoding covers.
+*/
+function codesOf(text) {
+	const codes = /* @__PURE__ */ new Map();
+	for (const ch of text.replace(/[\r\n]/gu, "")) if (!codes.has(ch) && codes.size < 255) codes.set(ch, codes.size + 1);
+	return codes;
+}
+/**
+* The Helvetica a drawn free-text annotation is set in, its encoding spelling
+* each code's character out as a glyph name (`uni0142` for ł) — which is how
+* the text reads back as what it is.
+*/
+function spelledFace(text) {
+	const differences = [1];
+	for (const ch of codesOf(text).keys()) {
+		const cp = ch.codePointAt(0) ?? 32;
+		differences.push(new PdfName(`uni${cp.toString(16).toUpperCase().padStart(4, "0")}`));
+	}
+	return new Map([
+		["Type", new PdfName("Font")],
+		["Subtype", new PdfName("Type1")],
+		["BaseFont", new PdfName("Helvetica")],
+		["Encoding", new Map([["Differences", differences]])]
+	]);
+}
 /** §12.7.4.3 — the field is a multi-line text box (bit 13 of `/Ff`). */
 var MULTILINE = 4096;
 /**

package/dist/esm/pdf-reader/annots.d.ts

@@ -8,5 +8,23 @@ export interface Appearance {
     readonly ctm: Matrix;
     /** The appearance's `/Resources`, when it states its own. */
     readonly resources: PdfDict | undefined;
+    /** §8.10.1 — the form's `/BBox`, in its own space, which clips what it paints. */
+    readonly bbox?: readonly [number, number, number, number];
 }
+/**
+ * §8.10.1 — an appearance's content as it paints: clipped to its `/BBox`.
+ *
+ * The box is the form's clip, not a hint. A form-filling tool that wrote its
+ * field backgrounds in PAGE coordinates inside a box forty points wide paints
+ * them nowhere any viewer shows — and painted anyway, with the fill colour it
+ * never set, bug1669099.pdf's form came back with three black slabs over its
+ * letterhead and its terms of payment, each twice as far across the sheet as
+ * the field it belonged to. The clip is installed the way the content would
+ * install one, so every pass that reads the appearance sees the same page.
+ *
+ * @param file       The owning file.
+ * @param appearance The appearance.
+ * @returns Its content, preceded by the clip to its box where it has one.
+ */
+export declare function appearanceContent(file: PdfFile, appearance: Appearance): Uint8Array;
 export declare function collectPageAppearances(file: PdfFile, page: PdfPage): Array<Appearance>;

package/dist/esm/pdf-reader/annots.js

@@ -2,6 +2,32 @@ import { PDF_NULL, PdfName, PdfStream } from "../pdf/objects.js";
 import { IDENTITY, multiply } from "./content.js";
 import { drawnAppearance, drawnResources, textMarkupOf } from "./annot-draw.js";
 //#region src/pdf-reader/annots.ts
+/**
+* §8.10.1 — an appearance's content as it paints: clipped to its `/BBox`.
+*
+* The box is the form's clip, not a hint. A form-filling tool that wrote its
+* field backgrounds in PAGE coordinates inside a box forty points wide paints
+* them nowhere any viewer shows — and painted anyway, with the fill colour it
+* never set, bug1669099.pdf's form came back with three black slabs over its
+* letterhead and its terms of payment, each twice as far across the sheet as
+* the field it belonged to. The clip is installed the way the content would
+* install one, so every pass that reads the appearance sees the same page.
+*
+* @param file       The owning file.
+* @param appearance The appearance.
+* @returns Its content, preceded by the clip to its box where it has one.
+*/
+function appearanceContent(file, appearance) {
+	const data = file.streamData(appearance.stream);
+	const box = appearance.bbox;
+	if (!box) return data;
+	const [x0, y0, x1, y1] = box;
+	const clip = new TextEncoder().encode(`${String(x0)} ${String(y0)} ${String(x1 - x0)} ${String(y1 - y0)} re W n\n`);
+	const out = new Uint8Array(clip.length + data.length);
+	out.set(clip, 0);
+	out.set(data, clip.length);
+	return out;
+}
 /** §12.5.3 `/F` — the annotation is not painted at all. */
 var FLAG_HIDDEN = 2;
 var FLAG_NOVIEW = 32;
@@ -35,6 +61,7 @@ function collectPageAppearances(file, page) {
 	const annots = file.get(page.dict, "Annots");
 	if (!Array.isArray(annots)) return [];
 	const out = [];
+	const regenerate = needsAppearances(file);
 	for (const entry of annots) {
 		const annot = file.resolve(entry);
 		if (!(annot instanceof Map)) continue;
@@ -44,6 +71,18 @@ function collectPageAppearances(file, page) {
 		const flags = file.get(annot, "F");
 		if (typeof flags === "number" && (flags & FLAG_HIDDEN || flags & FLAG_NOVIEW)) continue;
 		const stream = normalAppearance(file, annot);
+		const typed = regenerate && isWidget(file, annot) ? drawnAppearance(file, annot) : void 0;
+		if (typed && typedField(file, annot)) {
+			const kept = stream ? withoutVariableText(file, stream) : void 0;
+			const rect = rectangle(file.get(annot, "Rect"));
+			if (kept && rect) out.push(fitted(file, kept, rect));
+			out.push({
+				stream: typed,
+				ctm: IDENTITY,
+				resources: drawnResources(file, annot)
+			});
+			continue;
+		}
 		if (!stream) {
 			const drawn = drawnAppearance(file, annot);
 			if (drawn) out.push({
@@ -55,17 +94,55 @@ function collectPageAppearances(file, page) {
 		}
 		const rect = rectangle(file.get(annot, "Rect"));
 		if (!rect) continue;
-		const matrix = matrixOf(file, stream.dict);
-		const bbox = rectangle(file.get(stream.dict, "BBox"));
-		const resources = file.get(stream.dict, "Resources");
-		out.push({
-			stream,
-			ctm: multiply(matrix, fitToRect(bbox, matrix, rect)),
-			resources: resources instanceof Map ? resources : void 0
-		});
+		out.push(fitted(file, stream, rect));
 	}
 	return out;
 }
+/** An appearance stream placed on its annotation's `/Rect` (§12.5.5). */
+function fitted(file, stream, rect) {
+	const matrix = matrixOf(file, stream.dict);
+	const bbox = rectangle(file.get(stream.dict, "BBox"));
+	const resources = file.get(stream.dict, "Resources");
+	return {
+		stream,
+		ctm: multiply(matrix, fitToRect(bbox, matrix, rect)),
+		resources: resources instanceof Map ? resources : void 0,
+		...bbox ? { bbox } : {}
+	};
+}
+/** §12.7.2 — whether the form asks the viewer to build its fields' appearances. */
+function needsAppearances(file) {
+	const acro = file.get(file.catalog, "AcroForm");
+	return acro instanceof Map && file.get(acro, "NeedAppearances") === true;
+}
+function isWidget(file, annot) {
+	const subtype = file.get(annot, "Subtype");
+	return subtype instanceof PdfName && subtype.value === "Widget";
+}
+/** Whether a widget belongs to a text field — `/FT /Tx`, on it or up its field tree. */
+function typedField(file, annot) {
+	let node = annot;
+	for (let depth = 0; node && depth < 32; depth++) {
+		const type = file.get(node, "FT");
+		if (type instanceof PdfName) return type.value === "Tx";
+		const parent = file.get(node, "Parent");
+		node = parent instanceof Map ? parent : void 0;
+	}
+	return false;
+}
+/**
+* §12.7.3.3 — an appearance with its variable text taken out: the part it
+* marks `/Tx BMC … EMC`, which is what a viewer rebuilds from the value.
+* bug1844583.pdf's stale appearance still reads "Dlrow Olleh" there, over a
+* field whose value is "Hello World".
+*/
+function withoutVariableText(file, stream) {
+	const stripped = new TextDecoder("latin1").decode(file.streamData(stream)).replace(/\/Tx\s+BMC[\s\S]*?EMC/gu, "");
+	const dict = new Map(stream.dict);
+	dict.delete("Filter");
+	dict.delete("DecodeParms");
+	return new PdfStream(dict, Uint8Array.from([...stripped].map((c) => c.charCodeAt(0))));
+}
 /** §12.5.5 `/AP` `/N` — the normal appearance, through `/AS` when it is a set. */
 function normalAppearance(file, annot) {
 	const ap = file.get(annot, "AP");
@@ -149,4 +226,4 @@ function rectangle(v) {
 	];
 }
 //#endregion
-export { collectPageAppearances };
+export { appearanceContent, collectPageAppearances };

package/dist/esm/pdf-reader/cmap.js

@@ -45,8 +45,11 @@ function parseToUnicodeCMap(bytes) {
 			const hiN = bytesToInt(hi.bytes);
 			const dst = lexer.nextToken();
 			if (dst.kind === "hexstr") {
-				const base = stated(utf16be(dst.bytes));
-				if (base !== void 0) for (let i = 0; loN + i <= hiN && i < MAX_RANGE; i++) map.set(loN + i, incString(base, i));
+				const base = utf16be(dst.bytes);
+				for (let i = 0; loN + i <= hiN && i < MAX_RANGE; i++) {
+					const text = stated(incString(base, i));
+					if (text !== void 0) map.set(loN + i, text);
+				}
 			} else if (dst.kind === "arrayOpen") {
 				let i = 0;
 				for (;;) {

package/dist/esm/pdf-reader/content.d.ts

@@ -1,6 +1,5 @@
-import { ColorSpaceInfo, GsPaint } from './shading.js';
+import { ColorSpaceInfo, GsPaint, PageGradient } from './shading.js';
 import { TextMarkup } from './annot-draw.js';
-import { ShapeGradient } from '../core/vector.js';
 import { PdfDict, PdfStream } from '../pdf/objects.js';
 /**
  * A page font as the interpreter needs it (built from the font dictionaries in
@@ -136,6 +135,12 @@ export interface TextRun {
      * leaves it to the picture.
      */
     readonly invisible?: boolean;
+    /**
+     * §12.5.5 — drawn by an annotation's appearance rather than by the page: a
+     * field's value, a button's caption, a tick. It stands in the annotation's
+     * box, which is placed where the page has it, not in the page's reading.
+     */
+    readonly annotation?: boolean;
     /**
      * §9.3.6 — the colour the glyphs are STROKED in, when the rendering mode
      * asks for a stroke, and how wide the pen is.
@@ -245,7 +250,7 @@ export interface VectorPlacement {
     /** Fill colour (6-hex), present iff the path is filled (`f` / `F` / `f*` / `B` / `b`). */
     readonly fillHex?: string;
     /** Shading pattern, present iff filled with one (EP16c). */
-    readonly gradient?: ShapeGradient;
+    readonly gradient?: PageGradient;
     /** §11.6.4.4 `/ca` — how opaque the fill is, when the page asked for less. */
     readonly alpha?: number;
     /**
@@ -271,10 +276,18 @@ export interface VectorPlacement {
      * known by walking into it; the `fillHex` beside this is not the fill.
      */
     readonly patternName?: string;
+    /** §8.7.3.3 — the colour `scn` gave an uncoloured pattern, where it gave one. */
+    readonly patternPaint?: string;
     /** Stroke colour (6-hex), present iff the path is stroked (`S` / `s` / `B` / `b`) — EP11. */
     readonly strokeHex?: string;
     /** Stroke width in page-space points — EP11. */
     readonly lineWidth?: number;
+    /** §8.4.3.6 — the stroke's dash lengths in page-space points, where it is dashed. */
+    readonly dash?: ReadonlyArray<number>;
+    /** §8.4.3.3 — the stroke's cap, where it is not the butt cap every line starts with. */
+    readonly cap?: 'round' | 'square';
+    /** §11.6.4.4 `/CA` — how opaque the stroke is, when the page asked for less than all. */
+    readonly strokeAlpha?: number;
     readonly mcid?: number;
 }
 /** Everything {@link interpretContent} extracts from one page's content stream. */
@@ -357,7 +370,7 @@ export type Matrix = readonly [number, number, number, number, number, number];
 export declare const IDENTITY: Matrix;
 /** Compose two {@link Matrix matrices}: `a` applied first, then `b`. */
 export declare function multiply(a: Matrix, b: Matrix): Matrix;
-export declare function interpretContent(bytes: Uint8Array, fonts: ReadonlyMap<string, ContentFont>, initialCtm?: Matrix, shadings?: ReadonlyMap<string, ShapeGradient>, alphas?: ReadonlyMap<string, GsPaint>, spaces?: ReadonlyMap<string, ColorSpaceInfo>, hiddenOc?: ReadonlySet<string>): InterpretResult;
+export declare function interpretContent(bytes: Uint8Array, fonts: ReadonlyMap<string, ContentFont>, initialCtm?: Matrix, shadings?: ReadonlyMap<string, PageGradient>, alphas?: ReadonlyMap<string, GsPaint>, spaces?: ReadonlyMap<string, ColorSpaceInfo>, hiddenOc?: ReadonlySet<string>): InterpretResult;
 /**
  * Whether a string is wholly right-to-left: at least one letter of an RTL
  * script and nothing of any other, spaces and joiners aside.