@poe-platform/safe-bash 0.1.788 → 0.1.790

package/dist/safe-bash-command-pdftotext/index.js

@@ -4,6 +4,8 @@ import {
 
 // packages/safe-bash-command-pdftotext/dist/index.js
 init_buffer();
+import { yieldTurn, drainCooperativeSteps as drainSteps } from "../safe-bash-contracts/yield.js";
+import { InputByteBudget } from "../safe-bash-contracts/io.js";
 import { writeFileOutput } from "../safe-bash-contracts/filesystem-output-budget.js";
 import {
   commandRuntimeIdentity,
@@ -27100,6 +27102,11 @@ function tokenizeCos(bytes, start = 0, end = bytes.length) {
   }
   return tokens;
 }
+function drainWork(work) {
+  let step = work.next();
+  while (!step.done) step = work.next();
+  return step.value;
+}
 function bytesToString3(bytes) {
   if (typeof bytes !== "object" || bytes?.length === void 0) {
     throw new Error("Invalid argument for bytesToString");
@@ -27695,7 +27702,9 @@ function concatByteArrays(chunks) {
   }
   return out;
 }
-function serializeCosDocument(options) {
+function* serializeCosDocumentSteps(options) {
+  yield;
+  let work = 0;
   const maxOutputBytes = options.maxOutputBytes ?? Infinity;
   const maxObjects = options.maxObjects ?? Infinity;
   const sorted = [...options.objects].sort((a, b) => a.objectNumber - b.objectNumber);
@@ -27718,6 +27727,7 @@ function serializeCosDocument(options) {
     const packable = [];
     const direct = [];
     for (const obj of sorted) {
+      if (++work % 16 === 0) yield;
       const isLinearized = obj.value.kind === "dict" && dictGet(obj.value, "Linearized") !== void 0;
       if (obj.value.kind !== "stream" && obj.generationNumber === 0 && !isLinearized) {
         packable.push(obj);
@@ -27732,6 +27742,7 @@ function serializeCosDocument(options) {
       const headerPairs = [];
       let relOffset = 0;
       for (const pObj of packable) {
+        if (++work % 16 === 0) yield;
         headerPairs.push(`${pObj.objectNumber} ${relOffset}`);
         const bBytes = serializeCosNodeBytes(pObj.value);
         const nl = textEncoder2.encode("\n");
@@ -27757,6 +27768,7 @@ function serializeCosDocument(options) {
       ];
       const directOffsets = /* @__PURE__ */ new Map();
       for (const dObj of allDirect) {
+        if (++work % 16 === 0) yield;
         directOffsets.set(dObj.objectNumber, { offset, generation: dObj.generationNumber });
         const prefix = textEncoder2.encode(`${dObj.objectNumber} ${dObj.generationNumber} obj
 `);
@@ -27767,6 +27779,7 @@ function serializeCosDocument(options) {
       }
       const compressedMap = /* @__PURE__ */ new Map();
       for (let idx = 0; idx < packable.length; idx++) {
+        if (++work % 16 === 0) yield;
         compressedMap.set(packable[idx].objectNumber, idx);
       }
       const xrefOffset2 = offset;
@@ -27784,6 +27797,7 @@ function serializeCosDocument(options) {
       };
       writeEntry7(0, 0, 0, 65535);
       for (let i = 1; i < size2; i++) {
+        if (++work % 16 === 0) yield;
         if (i === xrefStmNum) {
           writeEntry7(i, 1, xrefOffset2, 0);
         } else if (compressedMap.has(i)) {
@@ -27864,6 +27878,7 @@ ${xrefOffset2}
     let pageCount = 0;
     let firstPageObjNum = options.rootRef.objectNumber;
     for (const o of sorted) {
+      if (++work % 16 === 0) yield;
       if (o.value.kind === "dict") {
         const t = dictGet(o.value, "Type");
         if (t?.kind === "name" && t.decoded === "Page") {
@@ -27911,6 +27926,7 @@ ${xrefOffset2}
   }
   const objectOffsets = /* @__PURE__ */ new Map();
   for (const obj of sorted) {
+    if (++work % 16 === 0) yield;
     objectOffsets.set(obj.objectNumber, { offset, generation: obj.generationNumber });
     let value = obj.value;
     if (options.normalizeContent && value.kind === "stream" && value.decodedBytes) {
@@ -27942,6 +27958,7 @@ ${xrefOffset2}
 0000000000 65535 f 
 `;
   for (let i = 1; i < size; i++) {
+    if (++work % 16 === 0) yield;
     const entry = objectOffsets.get(i);
     if (entry) {
       xrefText += `${String(entry.offset).padStart(10, "0")} ${String(entry.generation).padStart(5, "0")} n 
@@ -27988,6 +28005,7 @@ ${xrefOffset}
     let endFirstPage = xrefOffset;
     const sortedOffsets = [...objectOffsets.entries()].sort((a, b) => a[1].offset - b[1].offset);
     for (let i = 0; i < sortedOffsets.length; i++) {
+      if (++work % 16 === 0) yield;
       const [objNum, info3] = sortedOffsets[i];
       const nextOff = sortedOffsets[i + 1]?.[1].offset ?? xrefOffset;
       if (objNum === hintObjAfter.objectNumber) {
@@ -28021,6 +28039,9 @@ ${xrefOffset}
   }
   return fullBytes;
 }
+function serializeCosDocument(options) {
+  return drainWork(serializeCosDocumentSteps(options));
+}
 function appendIncrementalRevision(originalBytes, doc, updatedObjects) {
   const chunks = [originalBytes, textEncoder2.encode("\n")];
   let offset = originalBytes.length + 1;
@@ -28294,48 +28315,30 @@ function encryptPdfBuffer(state, objectNumber, generationNumber, plaintext) {
   const objKey = md5Bytes([state.fileKey, suffix]).subarray(0, Math.min(16, state.fileKey.length + 5));
   return rc4Transform(objKey, plaintext);
 }
-function transformNodeStringsAndStreams(node, transform, transformStream) {
+function* transformNodeStringsAndStreamsSteps(node, transform, transformStream) {
+  yield;
   switch (node.kind) {
     case "string":
-      return {
-        kind: "string",
-        format: "hex",
-        bytes: transform(node.bytes)
-      };
-    case "array":
-      return {
-        kind: "array",
-        items: node.items.map((item) => transformNodeStringsAndStreams(item, transform, transformStream))
-      };
+      return { kind: "string", format: "hex", bytes: transform(node.bytes) };
+    case "array": {
+      const items = [];
+      for (const item of node.items) items.push(yield* transformNodeStringsAndStreamsSteps(item, transform, transformStream));
+      return { kind: "array", items };
+    }
     case "dict": {
-      const type = dictGet(node, "Type");
-      const fieldType = dictGet(node, "FT");
-      const byteRange = dictGet(node, "ByteRange");
+      const type = dictGet(node, "Type"), fieldType = dictGet(node, "FT"), byteRange = dictGet(node, "ByteRange");
       const signature = type?.kind === "name" && type.decoded === "Sig" || fieldType?.kind === "name" && fieldType.decoded === "Sig" || byteRange?.kind === "array" && dictGet(node, "Filter")?.kind === "name";
-      return {
-        kind: "dict",
-        entries: node.entries.map((e) => ({
-          key: e.key,
-          value: signature && e.key.decoded === "Contents" ? e.value : transformNodeStringsAndStreams(e.value, transform, transformStream)
-        }))
-      };
+      const entries = [];
+      for (const e of node.entries) entries.push({ key: e.key, value: signature && e.key.decoded === "Contents" ? e.value : yield* transformNodeStringsAndStreamsSteps(e.value, transform, transformStream) });
+      return { kind: "dict", entries };
     }
     case "stream": {
       const type = dictGet(node.dict, "Type");
       if (type?.kind === "name" && type.decoded === "XRef") return node;
       const stream = transformStream ? transformStream(node) : { ...node, rawBytes: transform(node.rawBytes) };
-      const updatedDict = {
-        kind: "dict",
-        entries: stream.dict.entries.map((e) => ({
-          key: e.key,
-          value: e.key.decoded === "Length" ? cosNumber(stream.rawBytes.length) : transformNodeStringsAndStreams(e.value, transform, transformStream)
-        }))
-      };
-      return {
-        kind: "stream",
-        dict: updatedDict,
-        rawBytes: stream.rawBytes
-      };
+      const entries = [];
+      for (const e of stream.dict.entries) entries.push({ key: e.key, value: e.key.decoded === "Length" ? cosNumber(stream.rawBytes.length) : yield* transformNodeStringsAndStreamsSteps(e.value, transform, transformStream) });
+      return { kind: "stream", dict: { kind: "dict", entries }, rawBytes: stream.rawBytes };
     }
     default:
       return node;
@@ -28404,7 +28407,8 @@ function createR6Encryption(userPassword, ownerPassword, pMask) {
     StrF: cosName("StdCF")
   }) };
 }
-function encryptCosDocument(doc, options = {}) {
+function* encryptCosDocumentSteps(doc, options = {}) {
+  yield;
   const userPassword = options.userPassword ?? "";
   const ownerPassword = options.ownerPassword ?? userPassword;
   const revision = options.revision ?? 6;
@@ -28426,14 +28430,15 @@ function encryptCosDocument(doc, options = {}) {
   const encryptObjNum = maxObjNum + 1;
   const encryptedObjects = [];
   for (const obj of doc.objects.values()) {
-    const transformed = transformNodeStringsAndStreams(
+    yield;
+    const transformed = yield* transformNodeStringsAndStreamsSteps(
       obj.value,
       (plain) => encryptPdfBuffer(state, obj.objectNumber, obj.generationNumber, plain)
     );
     encryptedObjects.push({ objectNumber: obj.objectNumber, generationNumber: obj.generationNumber, value: transformed });
   }
   encryptedObjects.push({ objectNumber: encryptObjNum, generationNumber: 0, value: encryptDict });
-  return serializeCosDocument({
+  return yield* serializeCosDocumentSteps({
     objects: encryptedObjects,
     rootRef: doc.rootRef,
     infoRef: doc.infoRef,
@@ -28483,17 +28488,19 @@ function decryptStream(doc, stream, transform) {
   }
   return { kind: "stream", dict, rawBytes: bytes };
 }
-function decryptCosDocument(doc, state) {
+function* decryptCosDocumentSteps(doc, state) {
+  yield;
   const encryptObjNum = doc.encryptRef?.objectNumber;
   const handler = securityHandlers.get(state);
   for (const [num, obj] of doc.objects.entries()) {
+    yield;
     if (num === encryptObjNum || handler?.plaintextObjects.has(num)) continue;
     if (obj.value.kind === "stream") {
       const type = doc.resolve(dictGet(obj.value.dict, "Type"));
       if (type?.kind === "name" && (type.decoded === "XRef" || type.decoded === "Metadata" && !state.encryptMetadata)) continue;
     }
     const transform = handler?.factory.createCipherTransform(obj.objectNumber, obj.generationNumber);
-    const decrypted = transformNodeStringsAndStreams(
+    const decrypted = yield* transformNodeStringsAndStreamsSteps(
       obj.value,
       (cipher) => decryptPdfBuffer(state, obj.objectNumber, obj.generationNumber, cipher),
       transform ? (stream) => decryptStream(doc, stream, transform) : void 0
@@ -28607,9 +28614,12 @@ var ParsedCosDocument = class {
     return void 0;
   }
 };
-function parseNodeFromLexer(lexer, bytes, maxDepth, repair = false) {
+function* parseNodeFromLexerSteps(lexer, bytes, maxDepth, repair = false) {
+  yield;
+  let work = 0;
   const stack = [];
   while (true) {
+    if (++work % 16 === 0) yield;
     const tok = lexer.nextToken();
     const parent = stack.at(-1);
     if (!tok) {
@@ -28805,7 +28815,8 @@ function parseHeaderVersion(bytes) {
   }
   return head.slice(idx + 5, idx + 8);
 }
-function parseObjectAtOffset(bytes, offset, maxRecursionDepth, repair = false) {
+function* parseObjectAtOffsetSteps(bytes, offset, maxRecursionDepth, repair = false) {
+  yield;
   const lexer = new CosByteLexer(bytes, offset);
   const objNumTok = lexer.nextToken();
   const genNumTok = lexer.nextToken();
@@ -28813,7 +28824,7 @@ function parseObjectAtOffset(bytes, offset, maxRecursionDepth, repair = false) {
   if (objNumTok?.kind !== "number" || genNumTok?.kind !== "number" || objKwTok?.kind !== "keyword" || objKwTok.value !== "obj") {
     throw new PdfError("E_PARSE", `Malformed indirect object header at byte offset ${offset}`);
   }
-  const value = parseNodeFromLexer(lexer, bytes, maxRecursionDepth, repair);
+  const value = yield* parseNodeFromLexerSteps(lexer, bytes, maxRecursionDepth, repair);
   if (!value) {
     throw new PdfError("E_PARSE", `Empty indirect object ${objNumTok.value} at offset ${offset}`);
   }
@@ -28824,7 +28835,9 @@ function parseObjectAtOffset(bytes, offset, maxRecursionDepth, repair = false) {
     span: { start: objNumTok.span.start, end: lexer.offset }
   };
 }
-function parseXrefRevisionAt(bytes, xrefOffset, maxDecompressedBytes, maxRecursionDepth, maxObjects) {
+function* parseXrefRevisionAtSteps(bytes, xrefOffset, maxDecompressedBytes, maxRecursionDepth, maxObjects) {
+  yield;
+  let work = 0;
   if (xrefOffset < 0 || xrefOffset >= bytes.length) {
     throw new PdfError("E_PARSE", `Invalid xref offset: ${xrefOffset}`);
   }
@@ -28836,6 +28849,7 @@ function parseXrefRevisionAt(bytes, xrefOffset, maxDecompressedBytes, maxRecursi
   if (firstTok.kind === "keyword" && firstTok.value === "xref") {
     const entries2 = /* @__PURE__ */ new Map();
     while (true) {
+      if (++work % 16 === 0) yield;
       const saved = lexer.offset;
       const tok = lexer.nextToken();
       if (!tok) {
@@ -28859,6 +28873,7 @@ function parseXrefRevisionAt(bytes, xrefOffset, maxDecompressedBytes, maxRecursi
       }
       if (count > maxObjects + 1) throw new PdfError("E_LIMIT", "PDF object count limit exceeded while reading xref entries");
       for (let i = 0; i < count; i++) {
+        if (++work % 16 === 0) yield;
         const offTok = lexer.nextToken();
         const genTok = lexer.nextToken();
         const flagTok = lexer.nextToken();
@@ -28884,7 +28899,7 @@ function parseXrefRevisionAt(bytes, xrefOffset, maxDecompressedBytes, maxRecursi
         if (entries2.size > maxObjects + 1) throw new PdfError("E_LIMIT", "PDF object count limit exceeded while reading xref entries");
       }
     }
-    const trailerNode = parseNodeFromLexer(lexer, bytes, maxRecursionDepth);
+    const trailerNode = yield* parseNodeFromLexerSteps(lexer, bytes, maxRecursionDepth);
     if (!trailerNode || trailerNode.kind !== "dict") {
       throw new PdfError("E_PARSE", "Missing trailer dictionary after xref table");
     }
@@ -28896,7 +28911,7 @@ function parseXrefRevisionAt(bytes, xrefOffset, maxDecompressedBytes, maxRecursi
       previousXrefOffset: prevNode2?.kind === "number" ? prevNode2.value : void 0
     };
   }
-  const xrefObj = parseObjectAtOffset(bytes, xrefOffset, maxRecursionDepth);
+  const xrefObj = yield* parseObjectAtOffsetSteps(bytes, xrefOffset, maxRecursionDepth);
   if (xrefObj.value.kind !== "stream") {
     throw new PdfError("E_PARSE", `Expected xref table or XRef stream at offset ${xrefOffset}`);
   }
@@ -28917,6 +28932,7 @@ function parseXrefRevisionAt(bytes, xrefOffset, maxDecompressedBytes, maxRecursi
   if (indexArray) {
     if (indexArray.kind !== "array" || indexArray.items.length % 2 !== 0) throw new PdfError("E_PARSE", "Invalid XRef range fields");
     for (let i = 0; i < indexArray.items.length; i += 2) {
+      if (++work % 16 === 0) yield;
       const first = indexArray.items[i];
       const count = indexArray.items[i + 1];
       if (first?.kind !== "number" || count?.kind !== "number" || !Number.isSafeInteger(first.value) || first.value < 0 || !Number.isSafeInteger(count.value) || count.value < 0 || !Number.isSafeInteger(first.value + count.value)) {
@@ -28939,9 +28955,11 @@ function parseXrefRevisionAt(bytes, xrefOffset, maxDecompressedBytes, maxRecursi
     return v;
   };
   for (const [startObj, count] of subsections) {
+    if (++work % 16 === 0) yield;
     if (count > (decoded.length - pos) / stride) throw new PdfError("E_PARSE", "Truncated XRef stream");
     if (count > maxObjects + 1) throw new PdfError("E_LIMIT", "PDF object count limit exceeded while reading xref entries");
     for (let i = 0; i < count; i++) {
+      if (++work % 16 === 0) yield;
       const objNum = startObj + i;
       const field0 = w0 > 0 ? readInt(w0) : 1;
       const field1 = readInt(w1);
@@ -28987,12 +29005,16 @@ function isAsciiWhitespace(ch3) {
 function isAsciiDigit(ch3) {
   return ch3 >= 48 && ch3 <= 57;
 }
-function repairScanCosDocument(bytes, maxObjects, maxRecursionDepth) {
+function* repairScanCosDocumentSteps(bytes, maxObjects, maxRecursionDepth) {
+  yield;
+  let work = 0;
   const objects = /* @__PURE__ */ new Map();
   const trailers = [];
   let pos = 0;
   while (pos < bytes.length) {
+    if (++work % 16 === 0) yield;
     while (pos < bytes.length && !isAsciiDigit(bytes[pos])) {
+      if (++work % 16 === 0) yield;
       if (bytes[pos] === 37) {
         while (pos < bytes.length && bytes[pos] !== 10 && bytes[pos] !== 13) pos++;
         continue;
@@ -29002,7 +29024,7 @@ function repairScanCosDocument(bytes, maxObjects, maxRecursionDepth) {
         try {
           const token = lexer.nextToken();
           if (token?.kind === "keyword" && token.value === "trailer") {
-            const trailer = parseNodeFromLexer(lexer, bytes, maxRecursionDepth, true);
+            const trailer = yield* parseNodeFromLexerSteps(lexer, bytes, maxRecursionDepth, true);
             if (trailer?.kind === "dict") trailers.push(trailer);
             pos = lexer.offset;
             continue;
@@ -29035,7 +29057,7 @@ function repairScanCosDocument(bytes, maxObjects, maxRecursionDepth) {
     bytes[pos + 3] === 40)) {
       pos += 3;
       try {
-        const parsed = parseObjectAtOffset(bytes, headerStart, maxRecursionDepth, true);
+        const parsed = yield* parseObjectAtOffsetSteps(bytes, headerStart, maxRecursionDepth, true);
         objects.set(parsed.objectNumber, parsed);
         if (parsed.value.kind === "stream") {
           const type = dictGet(parsed.value.dict, "Type");
@@ -29052,13 +29074,17 @@ function repairScanCosDocument(bytes, maxObjects, maxRecursionDepth) {
   }
   return { objects, trailers };
 }
-function unpackRecoveredObjects(objects, maxObjects, maxDecompressedBytes, maxRecursionDepth) {
+function* unpackRecoveredObjectsSteps(objects, maxObjects, maxDecompressedBytes, maxRecursionDepth) {
+  yield;
+  let work = 0;
   for (const obj of [...objects.values()]) {
+    if (++work % 16 === 0) yield;
     if (obj.value.kind === "stream") {
       const t = dictGet(obj.value.dict, "Type");
       if (t?.kind === "name" && t.decoded === "ObjStm") {
         try {
-          for (const [unpackedNum, unpackedVal] of unpackObjectStream(obj.value, maxDecompressedBytes, maxRecursionDepth, true).entries()) {
+          for (const [unpackedNum, unpackedVal] of (yield* unpackObjectStreamSteps(obj.value, maxDecompressedBytes, maxRecursionDepth, true)).entries()) {
+            if (++work % 16 === 0) yield;
             if (!objects.has(unpackedNum)) {
               objects.set(unpackedNum, {
                 objectNumber: unpackedNum,
@@ -29094,7 +29120,9 @@ function recoveredReferences(objects) {
   }
   return { ...rootRef ? { rootRef } : {}, ...infoRef ? { infoRef } : {} };
 }
-function unpackObjectStream(streamObj, maxDecompressedBytes, maxRecursionDepth, repair = false) {
+function* unpackObjectStreamSteps(streamObj, maxDecompressedBytes, maxRecursionDepth, repair = false) {
+  yield;
+  let work = 0;
   const nNode = dictGet(streamObj.dict, "N");
   const firstNode = dictGet(streamObj.dict, "First");
   if (nNode?.kind !== "number" || firstNode?.kind !== "number") {
@@ -29104,6 +29132,7 @@ function unpackObjectStream(streamObj, maxDecompressedBytes, maxRecursionDepth,
   const headerLexer = new CosByteLexer(decoded, 0);
   const pairs = [];
   for (let i = 0; i < nNode.value; i++) {
+    if (++work % 16 === 0) yield;
     const numTok = headerLexer.nextToken();
     const offTok = headerLexer.nextToken();
     if (numTok?.kind === "number" && offTok?.kind === "number") {
@@ -29112,9 +29141,10 @@ function unpackObjectStream(streamObj, maxDecompressedBytes, maxRecursionDepth,
   }
   const result = /* @__PURE__ */ new Map();
   for (const pair of pairs) {
+    if (++work % 16 === 0) yield;
     try {
       const valLexer = new CosByteLexer(decoded, firstNode.value + pair.relativeOffset);
-      const node = parseNodeFromLexer(valLexer, decoded, maxRecursionDepth, repair);
+      const node = yield* parseNodeFromLexerSteps(valLexer, decoded, maxRecursionDepth, repair);
       if (node) {
         result.set(pair.objectNumber, node);
       }
@@ -29124,7 +29154,9 @@ function unpackObjectStream(streamObj, maxDecompressedBytes, maxRecursionDepth,
   }
   return result;
 }
-function parseCosDocument(bytes, options = {}) {
+function* parseCosDocumentSteps(bytes, options = {}) {
+  yield;
+  let work = 0;
   const version = parseHeaderVersion(bytes);
   const recovery = options.recovery ?? "strict";
   const maxObjects = options.maxObjects ?? Infinity;
@@ -29159,10 +29191,12 @@ function parseCosDocument(bytes, options = {}) {
         let currentXrefOffset = offsetTok.value;
         const visitedOffsets = /* @__PURE__ */ new Set();
         while (currentXrefOffset !== void 0 && !visitedOffsets.has(currentXrefOffset)) {
+          if (++work % 16 === 0) yield;
           visitedOffsets.add(currentXrefOffset);
-          const rev = parseXrefRevisionAt(bytes, currentXrefOffset, maxDecompressedBytes, maxRecursionDepth, maxObjects);
+          const rev = yield* parseXrefRevisionAtSteps(bytes, currentXrefOffset, maxDecompressedBytes, maxRecursionDepth, maxObjects);
           revisions.push(rev);
           for (const [num, entry] of rev.entries.entries()) {
+            if (++work % 16 === 0) yield;
             if (!mergedXref.has(num)) {
               mergedXref.set(num, entry);
               if (mergedXref.size > maxObjects + 1) throw new PdfError("E_LIMIT", "PDF object count limit exceeded while reading xref entries");
@@ -29195,8 +29229,9 @@ function parseCosDocument(bytes, options = {}) {
   if (!useRepair && rootRef) {
     try {
       for (const [objNum, entry] of mergedXref.entries()) {
+        if (++work % 16 === 0) yield;
         if (entry.type !== "uncompressed") continue;
-        const parsed = parseObjectAtOffset(bytes, entry.offset ?? 0, maxRecursionDepth, recovery === "repair");
+        const parsed = yield* parseObjectAtOffsetSteps(bytes, entry.offset ?? 0, maxRecursionDepth, recovery === "repair");
         if (parsed.objectNumber !== objNum || parsed.generationNumber !== (entry.generationNumber ?? 0)) {
           throw new PdfError("E_PARSE", `Bad uncompressed XRef entry for object ${objNum}`);
         }
@@ -29214,7 +29249,7 @@ function parseCosDocument(bytes, options = {}) {
       throw new PdfError("E_PARSE", "PDF trailer missing /Root reference");
     }
     useRepair = true;
-    const scanned = repairScanCosDocument(bytes, maxObjects, maxRecursionDepth);
+    const scanned = yield* repairScanCosDocumentSteps(bytes, maxObjects, maxRecursionDepth);
     objects = scanned.objects;
     revisions.length = 0;
     mergedXref.clear();
@@ -29249,7 +29284,7 @@ function parseCosDocument(bytes, options = {}) {
     encryptNode = encrypt?.kind === "ref" || encrypt?.kind === "dict" ? encrypt : void 0;
     encryptRef = encrypt?.kind === "ref" ? encrypt : void 0;
     idArray = id?.kind === "array" ? id : void 0;
-    if (!rootRef && !encryptNode) unpackRecoveredObjects(objects, maxObjects, maxDecompressedBytes, maxRecursionDepth);
+    if (!rootRef && !encryptNode) yield* unpackRecoveredObjectsSteps(objects, maxObjects, maxDecompressedBytes, maxRecursionDepth);
     const fallback = recoveredReferences(objects);
     rootRef ??= fallback.rootRef;
     infoRef ??= fallback.infoRef;
@@ -29284,20 +29319,21 @@ function parseCosDocument(bytes, options = {}) {
       maxRecursionDepth
     });
     doc.encryption = encState;
-    decryptCosDocument(doc, encState);
+    yield* decryptCosDocumentSteps(doc, encState);
   }
   if (useRepair) {
-    unpackRecoveredObjects(objects, maxObjects, maxDecompressedBytes, maxRecursionDepth);
+    yield* unpackRecoveredObjectsSteps(objects, maxObjects, maxDecompressedBytes, maxRecursionDepth);
     doc.infoRef ??= recoveredReferences(objects).infoRef;
   }
   const objStmCache = /* @__PURE__ */ new Map();
   for (const [objNum, entry] of mergedXref.entries()) {
+    if (++work % 16 === 0) yield;
     if (entry.type === "compressed" && entry.objectStreamNumber !== void 0) {
       let unpacked = objStmCache.get(entry.objectStreamNumber);
       if (!unpacked) {
         const stmObj = objects.get(entry.objectStreamNumber)?.value;
         if (stmObj?.kind === "stream") {
-          unpacked = unpackObjectStream(stmObj, maxDecompressedBytes, maxRecursionDepth, recovery === "repair");
+          unpacked = yield* unpackObjectStreamSteps(stmObj, maxDecompressedBytes, maxRecursionDepth, recovery === "repair");
           objStmCache.set(entry.objectStreamNumber, unpacked);
         }
       }
@@ -29314,6 +29350,9 @@ function parseCosDocument(bytes, options = {}) {
   }
   return doc;
 }
+function parseCosDocument(bytes, options = {}) {
+  return drainWork(parseCosDocumentSteps(bytes, options));
+}
 var WIN_ANSI_HIGH_BYTES = {
   128: 8364,
   130: 8218,
@@ -40689,7 +40728,8 @@ function resolveInheritedFieldEntry(doc, dict, key) {
   }
   return void 0;
 }
-function collectFieldDicts(doc, node, prefix = "", out = [], parentFt = "", parentFf = 0, parentQ = 0, parentMaxLen) {
+function* collectFieldDictsSteps(doc, node, prefix = "", out = [], parentFt = "", parentFf = 0, parentQ = 0, parentMaxLen, yieldState = { work: 0 }) {
+  if (++yieldState.work % 256 === 0) yield;
   const dict = doc.resolveDict(node);
   if (!dict) return out;
   const tNode = doc.resolve(dictGet(dict, "T"));
@@ -40722,7 +40762,8 @@ function collectFieldDicts(doc, node, prefix = "", out = [], parentFt = "", pare
       return out;
     }
     for (const kid of kidsArr.items) {
-      collectFieldDicts(doc, kid, fullName, out, currentFt, currentFf, currentQ, currentMaxLen);
+      if (++yieldState.work % 16 === 0) yield;
+      yield* collectFieldDictsSteps(doc, kid, fullName, out, currentFt, currentFf, currentQ, currentMaxLen, yieldState);
     }
   } else if (fullName) {
     out.push({
@@ -41163,6 +41204,9 @@ function escapePdfLiteralString(str) {
   }
   return out;
 }
+function collectFieldDicts(doc, node, prefix = "", out = [], parentFt = "", parentFf = 0, parentQ = 0, parentMaxLen) {
+  drainWork(collectFieldDictsSteps(doc, node, prefix, out, parentFt, parentFf, parentQ, parentMaxLen));
+}
 var PdfPage = class {
   cosDoc;
   pageRef;
@@ -43843,14 +43887,18 @@ var PdfDocument = class _PdfDocument {
     });
     return new _PdfDocument(parseCosDocument(rawBytes));
   }
-  static load(bytes, options) {
+  static *loadSteps(bytes, options) {
+    yield;
     return new _PdfDocument(
-      parseCosDocument(bytes, {
+      yield* parseCosDocumentSteps(bytes, {
         ...options,
         recovery: options?.recovery ?? "repair"
       })
     );
   }
+  static load(bytes, options) {
+    return drainWork(_PdfDocument.loadSteps(bytes, options));
+  }
   rebuildPagesList() {
     const catalog = this.cos.resolveDict(this.cos.rootRef);
     const pagesNode = catalog ? dictGet(catalog, "Pages") : void 0;
@@ -43932,20 +43980,30 @@ var PdfDocument = class _PdfDocument {
     this.pages.splice(index, 1);
     this.syncPageTree();
   }
-  copyPagesFrom(sourceDoc, indices) {
+  *copyPagesFromSteps(sourceDoc, indices) {
+    function* cloneItems(items) {
+      const result = [];
+      for (const item of items) result.push(yield* cloneNode(item));
+      return result;
+    }
+    yield;
+    let work = 0;
     const memo = /* @__PURE__ */ new Map();
     const sourcePageObjNums = /* @__PURE__ */ new Set();
     for (let i = 0; i < sourceDoc.getPageCount(); i++) {
+      if (++work % 16 === 0) yield;
       sourcePageObjNums.add(sourceDoc.getPage(i).ref.objectNumber);
     }
     const targetPageRefs = [];
     for (const idx of indices) {
+      if (++work % 16 === 0) yield;
       const srcPage = sourceDoc.getPage(idx);
       const pageRef = this.cos.allocateObject({ kind: "null" });
       targetPageRefs.push(pageRef);
       memo.set(srcPage.ref.objectNumber, pageRef);
     }
-    const cloneNode = (node) => {
+    const cloneNode = function* (node) {
+      if (++work % 16 === 0) yield;
       if (node.kind === "ref") {
         const existing = memo.get(node.objectNumber);
         if (existing) return existing;
@@ -43956,12 +44014,12 @@ var PdfDocument = class _PdfDocument {
         if (!target) return cosRef(0);
         const placeholderRef = this.cos.allocateObject({ kind: "null" });
         memo.set(node.objectNumber, placeholderRef);
-        const clonedTarget = cloneNode(target);
+        const clonedTarget = yield* cloneNode(target);
         this.cos.setObject(placeholderRef.objectNumber, clonedTarget, 0);
         return placeholderRef;
       }
       if (node.kind === "array") {
-        return cosArray(node.items.map(cloneNode));
+        return cosArray(yield* cloneItems(node.items));
       }
       if (node.kind === "dict") {
         const typeEntry = dictGet(node, "Type");
@@ -43971,13 +44029,13 @@ var PdfDocument = class _PdfDocument {
           if (isPageTreeNode && entry.key.decoded === "Parent") continue;
           newEntries.push({
             key: { ...entry.key },
-            value: cloneNode(entry.value)
+            value: yield* cloneNode(entry.value)
           });
         }
         return { kind: "dict", entries: newEntries };
       }
       if (node.kind === "stream") {
-        const clonedDict = cloneNode(node.dict);
+        const clonedDict = yield* cloneNode(node.dict);
         return {
           kind: "stream",
           dict: clonedDict,
@@ -43986,11 +44044,12 @@ var PdfDocument = class _PdfDocument {
         };
       }
       return node;
-    };
+    }.bind(this);
     const copiedPages = [];
     for (const [position, idx] of indices.entries()) {
+      if (++work % 16 === 0) yield;
       const srcPage = sourceDoc.getPage(idx);
-      const clonedPageDict = cloneNode(srcPage.pageDict);
+      const clonedPageDict = yield* cloneNode(srcPage.pageDict);
       const size = srcPage.getSize();
       if (!dictGet(clonedPageDict, "MediaBox")) {
         dictSet(
@@ -44000,13 +44059,14 @@ var PdfDocument = class _PdfDocument {
         );
       }
       if (!dictGet(clonedPageDict, "Resources")) {
-        dictSet(clonedPageDict, "Resources", cloneNode(srcPage.getResourcesDict()));
+        dictSet(clonedPageDict, "Resources", yield* cloneNode(srcPage.getResourcesDict()));
       }
       if (!dictGet(clonedPageDict, "Rotate") && srcPage.getRotation() !== 0) {
         dictSet(clonedPageDict, "Rotate", cosNumber(srcPage.getRotation()));
       }
       const inheritedBoxKeys = ["CropBox", "BleedBox", "TrimBox", "ArtBox"];
       for (const boxKey of inheritedBoxKeys) {
+        if (++work % 16 === 0) yield;
         if (!dictGet(clonedPageDict, boxKey)) {
           const boxVal = boxKey === "CropBox" ? srcPage.getCropBox() : boxKey === "BleedBox" ? srcPage.getBleedBox() : boxKey === "TrimBox" ? srcPage.getTrimBox() : srcPage.getArtBox();
           if (boxVal[0] !== 0 || boxVal[1] !== 0 || boxVal[2] !== size.width || boxVal[3] !== size.height) {
@@ -44048,8 +44108,9 @@ var PdfDocument = class _PdfDocument {
       };
       const clonedRootFieldRefs = [];
       for (const item of srcFieldsArr.items) {
+        if (++work % 16 === 0) yield;
         if (!hasClonedDescendant(item)) continue;
-        const cloned = cloneNode(item);
+        const cloned = yield* cloneNode(item);
         if (cloned.kind === "ref") {
           const clonedDict = this.cos.resolveDict(cloned);
           const kidsArr = clonedDict ? this.cos.resolveArray(dictGet(clonedDict, "Kids")) : void 0;
@@ -44083,6 +44144,7 @@ var PdfDocument = class _PdfDocument {
             dstFields.items.filter((x) => x.kind === "ref").map((x) => x.objectNumber)
           );
           for (const fRef of clonedRootFieldRefs) {
+            if (++work % 16 === 0) yield;
             if (!existingNums.has(fRef.objectNumber)) {
               dstFields.items.push(fRef);
               existingNums.add(fRef.objectNumber);
@@ -44090,11 +44152,11 @@ var PdfDocument = class _PdfDocument {
           }
           const srcDr = dictGet(srcAcroForm, "DR");
           if (srcDr && !dictGet(dstAcroForm, "DR")) {
-            dictSet(dstAcroForm, "DR", cloneNode(srcDr));
+            dictSet(dstAcroForm, "DR", yield* cloneNode(srcDr));
           }
           const srcDa = dictGet(srcAcroForm, "DA");
           if (srcDa && !dictGet(dstAcroForm, "DA")) {
-            dictSet(dstAcroForm, "DA", cloneNode(srcDa));
+            dictSet(dstAcroForm, "DA", yield* cloneNode(srcDa));
           }
         }
       }
@@ -44103,12 +44165,15 @@ var PdfDocument = class _PdfDocument {
     if (srcOcProps) {
       const dstCatalog = this.cos.resolveDict(this.cos.rootRef);
       if (dstCatalog && !dictGet(dstCatalog, "OCProperties")) {
-        dictSet(dstCatalog, "OCProperties", cloneNode(srcOcProps));
+        dictSet(dstCatalog, "OCProperties", yield* cloneNode(srcOcProps));
       }
     }
     this.syncPageTree();
     return copiedPages;
   }
+  copyPagesFrom(sourceDoc, indices) {
+    return drainWork(this.copyPagesFromSteps(sourceDoc, indices));
+  }
   embedStandardFont(fontName) {
     return createStandardFontHandle(fontName);
   }
@@ -44308,14 +44373,15 @@ var PdfDocument = class _PdfDocument {
     });
     return buildSemanticAstFromPages(extractedPages, tablesByPage, displayLists);
   }
-  save(options = {}) {
+  *saveSteps(options = {}) {
+    yield;
     this.syncPageTree();
     if (this.cos.requiresFullRewrite) {
       pruneUnusedXObjects(this.cos, this.pages);
       discardUnreachableObjects(this.cos);
     }
     if (options.encrypt) {
-      return encryptCosDocument(this.cos, options.encrypt);
+      return yield* encryptCosDocumentSteps(this.cos, options.encrypt);
     }
     const allObjects = [...this.cos.objects.values()].sort(
       (a, b) => a.objectNumber - b.objectNumber
@@ -44323,7 +44389,7 @@ var PdfDocument = class _PdfDocument {
     if (options.incremental && !this.cos.requiresFullRewrite && this.cos.bytes.length > 0) {
       return appendIncrementalRevision(this.cos.bytes, this.cos, allObjects);
     }
-    return serializeCosDocument({
+    return yield* serializeCosDocumentSteps({
       objects: allObjects,
       rootRef: this.cos.rootRef,
       infoRef: this.cos.infoRef,
@@ -44334,6 +44400,9 @@ var PdfDocument = class _PdfDocument {
       linearize: options.linearize
     });
   }
+  save(options = {}) {
+    return drainWork(this.saveSteps(options));
+  }
 };
 function resolveDestinationPageIndex(doc, destOrActionNode) {
   const cos = doc.cos;
@@ -44895,7 +44964,7 @@ function renderTsv(pages, args) {
   }
   return rows.join("\n") + "\n";
 }
-function extractPdfToTextBytes(bytes, argv = []) {
+function* extractPdfToTextBytesSteps(bytes, argv = []) {
   const args = parseArgs(argv);
   const stderrParts = [];
   if (args.invalidEolWarning) {
@@ -44974,12 +45043,14 @@ ${[...SUPPORTED_ENCODINGS].join("\n")}
   const mode = args.raw ? "raw" : args.layout ? "layout" : "logical";
   const pages = [];
   for (let p = firstPage; p <= lastPage; p++) {
+    yield;
     const page = doc.getPage(p - 1);
     let pageCropBoxPt;
     if (args.cropbox) {
       let cur = page.pageDict;
       const visited = /* @__PURE__ */ new Set();
       while (cur && !visited.has(cur)) {
+        yield;
         visited.add(cur);
         const cb = doc.cos.resolveArray(dictGet(cur, "CropBox"));
         if (cb && cb.items.length >= 4) {
@@ -45058,6 +45129,7 @@ ${escapeXml(tsvText)}</pre>
   const eolChar = args.eol === "dos" ? "\r\n" : args.eol === "mac" ? "\r" : "\n";
   let textOut = "";
   for (const { pageNumber, extracted } of pages) {
+    yield;
     let pageText = formatExtractedPageText(extracted, {
       mode,
       rejoinHyphens: mode === "logical" && args.removeHyphens,
@@ -45069,6 +45141,7 @@ ${escapeXml(tsvText)}</pre>
       const annots = extractPageAnnotations(doc.cos, doc.getPage(pageNumber - 1).pageDict);
       const uris = [];
       for (const ann of annots) {
+        yield;
         if (ann.uri && !pageText.includes(ann.uri) && !uris.includes(ann.uri)) {
           uris.push(ann.uri);
         }
@@ -45123,14 +45196,34 @@ ${escapeXml(textOut)}</pre>
     outputPath: args.outputFile ?? "-"
   };
 }
-async function runPdftotextCli(argv, files, stdinBytes = new Uint8Array(0)) {
+function extractPdfToTextBytes(bytes, argv = []) {
+  const steps = extractPdfToTextBytesSteps(bytes, argv);
+  for (; ; ) {
+    const step = steps.next();
+    if (step.done) return step.value;
+  }
+}
+function* extractPdfToTextBytesCooperativelySteps(bytes, argv = [], signal) {
+  const steps = extractPdfToTextBytesSteps(bytes, argv);
+  try {
+    for (; ; ) {
+      yield;
+      const step = steps.next();
+      if (step.done)
+        return step.value;
+    }
+  } finally {
+    steps.return({ exitCode: 1, output: "", stderr: "", outputPath: "-" });
+  }
+}
+function* runPdftotextCliSteps(argv, files, stdinBytes = new Uint8Array(0), signal) {
   const args = parseArgs(argv);
   if (args.error || args.listenc || args.version || args.help) {
-    return extractPdfToTextBytes(new Uint8Array(0), argv);
+    return yield* extractPdfToTextBytesCooperativelySteps(new Uint8Array(0), argv, signal);
   }
   const target = args.inputFile ?? "-";
   if (target === "-") {
-    return extractPdfToTextBytes(stdinBytes, argv);
+    return yield* extractPdfToTextBytesCooperativelySteps(stdinBytes, argv, signal);
   }
   const fileBytes = files.get(target);
   if (!fileBytes) {
@@ -45142,9 +45235,13 @@ async function runPdftotextCli(argv, files, stdinBytes = new Uint8Array(0)) {
       outputPath: args.outputFile ?? "-"
     };
   }
-  return extractPdfToTextBytes(fileBytes, argv);
+  return yield* extractPdfToTextBytesCooperativelySteps(fileBytes, argv, signal);
+}
+async function runPdftotextCli(argv, files, stdinBytes = new Uint8Array(0), signal) {
+  return drainSteps(runPdftotextCliSteps(argv, files, stdinBytes, signal), signal);
 }
 async function pdftotext(context) {
+  let cooperativeWork = 63;
   const invocation = createOutputOperation(context, { write: async () => {
   } });
   try {
@@ -45152,7 +45249,7 @@ async function pdftotext(context) {
     const argv = [...carrier.args];
     const parsed = parseArgs(argv);
     if (parsed.error || parsed.listenc || parsed.version || parsed.help) {
-      const res2 = extractPdfToTextBytes(new Uint8Array(0), argv);
+      const res2 = await drainSteps(extractPdfToTextBytesCooperativelySteps(new Uint8Array(0), argv, invocation.signal), invocation.signal);
       if (res2.stderr) {
         await writeBytes(context.stderr, new TextEncoder().encode(res2.stderr), invocation.signal);
       }
@@ -45175,6 +45272,7 @@ async function pdftotext(context) {
       const chunks = [];
       let total = 0;
       for await (const chunk of readBytes(context.stdin, invocation.signal)) {
+        if (++cooperativeWork % 64 === 0) await yieldTurn(context.signal);
         chunks.push(chunk);
         total += chunk.byteLength;
         chargeBytes(chunk.byteLength);
@@ -45182,6 +45280,7 @@ async function pdftotext(context) {
       pdfBytes = new Uint8Array(total);
       let offset = 0;
       for (const c of chunks) {
+        if (++cooperativeWork % 64 === 0) await yieldTurn(context.signal);
         pdfBytes.set(c, offset);
         offset += c.byteLength;
       }
@@ -45199,7 +45298,7 @@ async function pdftotext(context) {
         return { exitCode: 1 };
       }
     }
-    const res = extractPdfToTextBytes(pdfBytes, argv);
+    const res = await drainSteps(extractPdfToTextBytesCooperativelySteps(pdfBytes, argv, invocation.signal), invocation.signal);
     if (res.stderr) {
       await writeBytes(context.stderr, new TextEncoder().encode(res.stderr), invocation.signal);
     }
@@ -45229,13 +45328,14 @@ async function pdftotext(context) {
     await invocation.close();
   }
 }
-function createPdftotextCommand(_options = {}) {
+function createPdftotextCommand(options = {}) {
+  const maxInputBytes = InputByteBudget.limit(options.limits?.maxInputBytes);
   return Object.freeze({
     name: "pdftotext",
     runtimeIdentity: commandRuntimeIdentity,
     description: "Extract PDF text, layout, XHTML bounding boxes, and TSV via @poe-code/pdf-ast",
     execute(context) {
-      return pdftotext(context);
+      return new InputByteBudget(maxInputBytes).run(context, pdftotext);
     }
   });
 }
@@ -45258,10 +45358,8 @@ function bytesToBase64(bytes) {
   }
   return out;
 }
-async function runPdftohtmlCli(argv, files) {
-  return runPdftohtmlCliSync(argv, files);
-}
-function runPdftohtmlCliSync(argv, files) {
+function* runPdftohtmlCliSteps(argv, files, signal) {
+  let cooperativeWork = 63;
   let xmlMode = false;
   let toStdout = false;
   let ignoreImages = false;
@@ -45274,6 +45372,8 @@ function runPdftohtmlCliSync(argv, files) {
   let password = "";
   const positionals = [];
   for (let i = 0; i < argv.length; i++) {
+    if (++cooperativeWork % 64 === 0)
+      yield;
     const arg = argv[i];
     if (arg === "-v" || arg === "--version") {
       return { exitCode: 0, stdout: "pdftohtml version 24.08.0\n", stderr: "" };
@@ -45343,6 +45443,7 @@ function runPdftohtmlCliSync(argv, files) {
   }
   const pageRefToNum = /* @__PURE__ */ new Map();
   for (let i = 0; i < doc.pageCount; i++) {
+    yield;
     pageRefToNum.set(doc.getPage(i).ref.objectNumber, i + 1);
   }
   const resolveOutlineDestPage = (itemDict) => {
@@ -45354,11 +45455,13 @@ function runPdftohtmlCliSync(argv, files) {
     let cur = firstNode;
     while (cur) {
       if (cur.kind === "ref") {
-        if (visited.has(cur.objectNumber)) break;
+        if (visited.has(cur.objectNumber))
+          break;
         visited.add(cur.objectNumber);
       }
       const dict = doc.cos.resolveDict(cur);
-      if (!dict) break;
+      if (!dict)
+        break;
       const tNode = doc.cos.resolve(dictGet(dict, "Title"));
       const title = tNode?.kind === "string" ? decodePdfString(tNode) : "";
       const pageNumber = resolveOutlineDestPage(dict);
@@ -45381,6 +45484,7 @@ function runPdftohtmlCliSync(argv, files) {
       `<pdf2xml producer="@poe-code/pdf-ast" version="24.08.0">`
     ];
     for (let p = firstPage; p <= endPage; p++) {
+      yield;
       const page = doc.getPage(p - 1);
       const { width, height } = page.getSize();
       const extracted = page.extractPage();
@@ -45391,6 +45495,8 @@ function runPdftohtmlCliSync(argv, files) {
       const fontSubDict = resDict ? doc.cos.resolveDict(dictGet(resDict, "Font")) : void 0;
       if (fontSubDict) {
         for (const entry of fontSubDict.entries) {
+          if (++cooperativeWork % 64 === 0)
+            yield;
           const fDict = doc.cos.resolveDict(entry.value);
           const bfNode = fDict ? doc.cos.resolve(dictGet(fDict, "BaseFont") ?? dictGet(fDict, "Name")) : void 0;
           if (bfNode?.kind === "name") {
@@ -45408,6 +45514,7 @@ function runPdftohtmlCliSync(argv, files) {
       lines.push(`  <page number="${p}" position="absolute" top="0" left="0" height="${Math.round(height * zoom)}" width="${Math.round(width * zoom)}">`);
       const imageLines = [];
       for (let imgIdx = 0; imgIdx < pageImages.length; imgIdx++) {
+        yield;
         const img = pageImages[imgIdx];
         const imgBytes = imageFmt === "jpg" ? encodeJpeg(img.bitmap) : encodePng(img.bitmap);
         const imgFile = `page${p}_${imgIdx + 1}.${imageFmt}`;
@@ -45420,7 +45527,11 @@ function runPdftohtmlCliSync(argv, files) {
       }
       const textLines = [];
       for (const block of extracted.blocks) {
+        if (++cooperativeWork % 64 === 0)
+          yield;
         for (const line of block.lines) {
+          if (++cooperativeWork % 64 === 0)
+            yield;
           const [x0, y0, x1, y1] = line.bbox;
           const top = Math.max(0, Math.round((height - y1) * zoom));
           const left = Math.max(0, Math.round(x0 * zoom));
@@ -45436,13 +45547,9 @@ function runPdftohtmlCliSync(argv, files) {
           if (specId === void 0) {
             specId = fontSpecIdByKey.size;
             fontSpecIdByKey.set(specKey, specId);
-            fontSpecLines.push(
-              `    <fontspec id="${specId}" size="${scaledSize}" family="${escapeHtmlXml(family)}" color="#000000"/>`
-            );
+            fontSpecLines.push(`    <fontspec id="${specId}" size="${scaledSize}" family="${escapeHtmlXml(family)}" color="#000000"/>`);
           }
-          const matchingLink = annots.find(
-            (a) => a.uri && Math.min(x1, Math.max(a.rect[0], a.rect[2])) > Math.max(x0, Math.min(a.rect[0], a.rect[2])) && Math.min(y1, Math.max(a.rect[1], a.rect[3])) > Math.max(y0, Math.min(a.rect[1], a.rect[3]))
-          );
+          const matchingLink = annots.find((a) => a.uri && Math.min(x1, Math.max(a.rect[0], a.rect[2])) > Math.max(x0, Math.min(a.rect[0], a.rect[2])) && Math.min(y1, Math.max(a.rect[1], a.rect[3])) > Math.max(y0, Math.min(a.rect[1], a.rect[3])));
           let styledText = escapeHtmlXml(line.text);
           const lowerFam = family.toLowerCase();
           if (lowerFam.includes("italic") || lowerFam.includes("oblique")) {
@@ -45483,6 +45590,7 @@ function runPdftohtmlCliSync(argv, files) {
       `<body>`
     ];
     for (let p = firstPage; p <= endPage; p++) {
+      yield;
       const page = doc.getPage(p - 1);
       const { width, height } = page.getSize();
       const extracted = page.extractPage();
@@ -45490,6 +45598,7 @@ function runPdftohtmlCliSync(argv, files) {
       const pageImages = ignoreImages ? [] : extractDocumentImages(doc.cos, { firstPage: p, lastPage: p });
       lines.push(`<div class="page" id="page${p}" style="position:relative;width:${Math.round(width * zoom)}pt;height:${Math.round(height * zoom)}pt;">`);
       for (let imgIdx = 0; imgIdx < pageImages.length; imgIdx++) {
+        yield;
         const img = pageImages[imgIdx];
         const imgBytes = imageFmt === "jpg" ? encodeJpeg(img.bitmap) : encodePng(img.bitmap);
         const imgFile = `page${p}_${imgIdx + 1}.${imageFmt}`;
@@ -45501,13 +45610,15 @@ function runPdftohtmlCliSync(argv, files) {
         lines.push(`  <img src="${src}" width="${img.width}" height="${img.height}"/>`);
       }
       for (const block of extracted.blocks) {
+        if (++cooperativeWork % 64 === 0)
+          yield;
         for (const line of block.lines) {
+          if (++cooperativeWork % 64 === 0)
+            yield;
           const [x0, y0, x1, y1] = line.bbox;
           const top = Math.max(0, Math.round((height - y1) * zoom));
           const left = Math.max(0, Math.round(x0 * zoom));
-          const matchingLink = annots.find(
-            (a) => a.uri && Math.min(x1, Math.max(a.rect[0], a.rect[2])) > Math.max(x0, Math.min(a.rect[0], a.rect[2])) && Math.min(y1, Math.max(a.rect[1], a.rect[3])) > Math.max(y0, Math.min(a.rect[1], a.rect[3]))
-          );
+          const matchingLink = annots.find((a) => a.uri && Math.min(x1, Math.max(a.rect[0], a.rect[2])) > Math.max(x0, Math.min(a.rect[0], a.rect[2])) && Math.min(y1, Math.max(a.rect[1], a.rect[3])) > Math.max(y0, Math.min(a.rect[1], a.rect[3])));
           const firstWord = line.words[0];
           const rawFontName = firstWord?.fontName ?? firstWord?.glyphs[0]?.fontName ?? "Helvetica";
           const cleanFontKey = rawFontName.startsWith("/") ? rawFontName.slice(1) : rawFontName;
@@ -45559,7 +45670,19 @@ function runPdftohtmlCliSync(argv, files) {
   files.set(outPath, new TextEncoder().encode(outputText));
   return { exitCode: 0, stdout: "", stderr: "" };
 }
+async function runPdftohtmlCli(argv, files, signal) {
+  return drainSteps(runPdftohtmlCliSteps(argv, files, signal), signal);
+}
+function runPdftohtmlCliSync(argv, files, signal) {
+  const steps = runPdftohtmlCliSteps(argv, files, signal);
+  let next = steps.next();
+  while (!next.done) {
+    next = steps.next();
+  }
+  return next.value;
+}
 async function pdftohtml(context) {
+  let cooperativeWork = 63;
   const invocation = createOutputOperation(context, { write: async () => {
   } });
   try {
@@ -45576,6 +45699,7 @@ async function pdftohtml(context) {
     };
     let inputOperand;
     for (let i = 0; i < argv.length; i++) {
+      if (++cooperativeWork % 64 === 0) await yieldTurn(context.signal);
       const arg = argv[i];
       if (["-f", "-l", "-zoom", "-fmt", "-enc", "-upw", "-opw"].includes(arg)) {
         i++;
@@ -45588,6 +45712,7 @@ async function pdftohtml(context) {
       const chunks = [];
       let total = 0;
       for await (const chunk of readBytes(context.stdin, invocation.signal)) {
+        if (++cooperativeWork % 64 === 0) await yieldTurn(context.signal);
         chunks.push(chunk);
         total += chunk.byteLength;
         chargeBytes(chunk.byteLength);
@@ -45596,6 +45721,7 @@ async function pdftohtml(context) {
         const buf = new Uint8Array(total);
         let off = 0;
         for (const c of chunks) {
+          if (++cooperativeWork % 64 === 0) await yieldTurn(context.signal);
           buf.set(c, off);
           off += c.byteLength;
         }
@@ -45603,6 +45729,7 @@ async function pdftohtml(context) {
       }
     }
     for (const token of argv) {
+      if (++cooperativeWork % 64 === 0) await yieldTurn(context.signal);
       if (token.startsWith("-") && token !== "-") continue;
       try {
         const bytes = await context.fs.readFile(resolveVfsPath(token), { signal: invocation.signal });
@@ -45611,8 +45738,9 @@ async function pdftohtml(context) {
       } catch {
       }
     }
+    context.inputBudget?.check(0);
     const existingSnap = new Map(vfsFiles);
-    const res = await runPdftohtmlCli(argv, vfsFiles);
+    const res = await runPdftohtmlCli(argv, vfsFiles, invocation.signal);
     if (res.stderr) {
       await writeBytes(context.stderr, new TextEncoder().encode(res.stderr), invocation.signal);
     }
@@ -45623,6 +45751,7 @@ async function pdftohtml(context) {
       await writeBytes(stdout.output, outBytes, invocation.signal);
     }
     for (const [key, val] of vfsFiles.entries()) {
+      if (++cooperativeWork % 64 === 0) await yieldTurn(context.signal);
       if (existingSnap.get(key) !== val) {
         chargeBytes(val.byteLength);
         const abs = resolveVfsPath(key);
@@ -45634,13 +45763,14 @@ async function pdftohtml(context) {
     await invocation.close();
   }
 }
-function createPdftohtmlCommand(_options = {}) {
+function createPdftohtmlCommand(options = {}) {
+  const maxInputBytes = InputByteBudget.limit(options.limits?.maxInputBytes);
   return Object.freeze({
     name: "pdftohtml",
     runtimeIdentity: commandRuntimeIdentity,
     description: "Convert PDF pages into HTML or XML layout documents via @poe-code/pdf-ast",
     execute(context) {
-      return pdftohtml(context);
+      return new InputByteBudget(maxInputBytes).run(context, pdftohtml);
     }
   });
 }