From ad97c5b816b01055711ebfe6018c6adc7d17cdc0 Mon Sep 17 00:00:00 2001
From: Calixte Denizet <calixte.denizet@gmail.com>
Date: Fri, 7 Nov 2025 15:34:08 +0100
Subject: [PATCH] Update the page labels tree when a pdf is extracted (bug
 1997379)

---
 src/core/catalog.js           |  16 +++--
 src/core/editor/pdf_editor.js | 123 +++++++++++++++++++++++++++++++++-
 test/pdfs/.gitignore          |   1 +
 test/pdfs/labelled_pages.pdf  | Bin 0 -> 6717 bytes
 test/unit/api_spec.js         |  34 ++++++++++
 5 files changed, 169 insertions(+), 5 deletions(-)
 create mode 100755 test/pdfs/labelled_pages.pdf

diff --git a/src/core/catalog.js b/src/core/catalog.js
index e5946c50f..5b5b319c7 100644
--- a/src/core/catalog.js
+++ b/src/core/catalog.js
@@ -735,6 +735,16 @@ class Catalog {
     return rawDests;
   }
 
+  get rawPageLabels() {
+    const obj = this.#catDict.getRaw("PageLabels");
+    if (!obj) {
+      return null;
+    }
+
+    const numberTree = new NumberTree(obj, this.xref);
+    return numberTree.getAll();
+  }
+
   get pageLabels() {
     let obj = null;
     try {
@@ -749,8 +759,8 @@ class Catalog {
   }
 
   #readPageLabels() {
-    const obj = this.#catDict.getRaw("PageLabels");
-    if (!obj) {
+    const nums = this.rawPageLabels;
+    if (!nums) {
       return null;
     }
 
@@ -758,8 +768,6 @@ class Catalog {
     let style = null,
       prefix = "";
 
-    const numberTree = new NumberTree(obj, this.xref);
-    const nums = numberTree.getAll();
     let currentLabel = "",
       currentIndex = 1;
 
diff --git a/src/core/editor/pdf_editor.js b/src/core/editor/pdf_editor.js
index 7df909156..5241863a7 100644
--- a/src/core/editor/pdf_editor.js
+++ b/src/core/editor/pdf_editor.js
@@ -25,6 +25,7 @@ import { StringStream } from "../stream.js";
 import { stringToAsciiOrUTF16BE } from "../core_utils.js";
 
 const MAX_LEAVES_PER_PAGES_NODE = 16;
+const MAX_IN_NAME_TREE_NODE = 64;
 
 class PageData {
   constructor(page, documentData) {
@@ -39,6 +40,7 @@ class PageData {
 class DocumentData {
   constructor(document) {
     this.document = document;
+    this.pageLabels = null;
     this.pagesMap = new RefSetCache();
     this.oldRefMapping = new RefSetCache();
   }
@@ -61,6 +63,7 @@ class PDFEditor {
     this.version = "1.7";
     this.title = title;
     this.author = author;
+    this.pageLabels = null;
   }
 
   /**
@@ -253,6 +256,8 @@ class PDFEditor {
     await Promise.all(promises);
     promises.length = 0;
 
+    this.#collectPageLabels();
+
     for (const page of this.oldPages) {
       promises.push(this.#postCollectPageData(page));
     }
@@ -270,7 +275,12 @@ class PDFEditor {
    * @param {DocumentData} documentData
    * @return {Promise<void>}
    */
-  async #collectDocumentData(documentData) {}
+  async #collectDocumentData(documentData) {
+    const { document } = documentData;
+    await document.pdfManager
+      .ensureCatalog("rawPageLabels")
+      .then(pageLabels => (documentData.pageLabels = pageLabels));
+  }
 
   /**
    * Post process the collected page data.
@@ -306,6 +316,56 @@ class PDFEditor {
     pageData.annotations = newAnnotations.length > 0 ? newAnnotations : null;
   }
 
+  async #collectPageLabels() {
+    // We can only preserve page labels when editing a single PDF file.
+    // This is consistent with behavior in Adobe Acrobat.
+    if (!this.hasSingleFile) {
+      return;
+    }
+    const {
+      documentData: { document, pageLabels },
+    } = this.oldPages[0];
+    if (!pageLabels) {
+      return;
+    }
+    const numPages = document.numPages;
+    const oldPageLabels = [];
+    const oldPageIndices = new Set(
+      this.oldPages.map(({ page: { pageIndex } }) => pageIndex)
+    );
+    let currentLabel = null;
+    let stFirstIndex = -1;
+    for (let i = 0; i < numPages; i++) {
+      const newLabel = pageLabels.get(i);
+      if (newLabel) {
+        currentLabel = newLabel;
+        stFirstIndex = currentLabel.has("St") ? i : -1;
+      }
+      if (!oldPageIndices.has(i)) {
+        continue;
+      }
+      if (stFirstIndex !== -1) {
+        const st = currentLabel.get("St");
+        currentLabel = currentLabel.clone();
+        currentLabel.set("St", st + (i - stFirstIndex));
+        stFirstIndex = -1;
+      }
+      oldPageLabels.push(currentLabel);
+    }
+    currentLabel = oldPageLabels[0];
+    let currentIndex = 0;
+    const newPageLabels = (this.pageLabels = [[0, currentLabel]]);
+    for (let i = 0, ii = oldPageLabels.length; i < ii; i++) {
+      const label = oldPageLabels[i];
+      if (label === currentLabel) {
+        continue;
+      }
+      currentIndex = i;
+      currentLabel = label;
+      newPageLabels.push([currentIndex, currentLabel]);
+    }
+  }
+
   /**
    * Create a copy of a page.
    * @param {number} pageIndex
@@ -423,6 +483,66 @@ class PDFEditor {
     }
   }
 
+  /**
+   * Create a name or number tree from the given map.
+   * @param {Array<[string, any]>} map
+   * @returns {Ref}
+   */
+  #makeNameNumTree(map, areNames) {
+    const allEntries = map.sort(
+      areNames
+        ? ([keyA], [keyB]) => keyA.localeCompare(keyB)
+        : ([keyA], [keyB]) => keyA - keyB
+    );
+    const maxLeaves =
+      MAX_IN_NAME_TREE_NODE <= 1 ? allEntries.length : MAX_IN_NAME_TREE_NODE;
+    const [treeRef, treeDict] = this.newDict;
+    const stack = [{ dict: treeDict, entries: allEntries }];
+    const valueType = areNames ? "Names" : "Nums";
+
+    while (stack.length > 0) {
+      const { dict, entries } = stack.pop();
+      if (entries.length <= maxLeaves) {
+        dict.set("Limits", [entries[0][0], entries.at(-1)[0]]);
+        dict.set(valueType, entries.flat());
+        continue;
+      }
+      const entriesChunks = [];
+      const chunkSize = Math.max(
+        maxLeaves,
+        Math.ceil(entries.length / maxLeaves)
+      );
+      for (let i = 0; i < entries.length; i += chunkSize) {
+        entriesChunks.push(entries.slice(i, i + chunkSize));
+      }
+      const entriesRefs = [];
+      dict.set("Kids", entriesRefs);
+      for (const chunk of entriesChunks) {
+        const [entriesRef, entriesDict] = this.newDict;
+        entriesRefs.push(entriesRef);
+        entriesDict.set("Limits", [chunk[0][0], chunk.at(-1)[0]]);
+        stack.push({ dict: entriesDict, entries: chunk });
+      }
+    }
+    return treeRef;
+  }
+
+  /**
+   * Create the page labels tree if it exists.
+   */
+  #makePageLabelsTree() {
+    const { pageLabels } = this;
+    if (!pageLabels || pageLabels.length === 0) {
+      return;
+    }
+    const { rootDict } = this;
+    const pageLabelsRef = this.#makeNameNumTree(
+      this.pageLabels,
+      /* areNames = */ false
+    );
+    rootDict.set("PageLabels", pageLabelsRef);
+  }
+
   /**
    * Create the root dictionary.
    * @returns {Promise<void>}
@@ -432,6 +552,7 @@ class PDFEditor {
     rootDict.setIfName("Type", "Catalog");
     rootDict.set("Version", this.version);
     this.#makePageTree();
+    this.#makePageLabelsTree();
   }
 
   /**
diff --git a/test/pdfs/.gitignore b/test/pdfs/.gitignore
index 91091a44e..9fc4acaf5 100644
--- a/test/pdfs/.gitignore
+++ b/test/pdfs/.gitignore
@@ -757,3 +757,4 @@
 !doc_1_3_pages.pdf
 !doc_2_3_pages.pdf
 !doc_3_3_pages.pdf
+!labelled_pages.pdf
diff --git a/test/pdfs/labelled_pages.pdf b/test/pdfs/labelled_pages.pdf
new file mode 100755
index 0000000000000000000000000000000000000000..68e389f40c5d467f478b2899378d375682f64ce3
GIT binary patch
literal 6717
zcmeHMdvFuS8JDrKV~xc)E+vmTun~q5W9fG9C7nhRmL(gTSXh2g8;px|y0eij>EzQX
zHcY02ha@}#hO}uW88a;zNFmUMn2=0TXfbI^2-JWRm=sz_NRxpQNT#KQv{2}t{1PC~
z#51J+$C{Bwce~%~_x*PF+ua!+SBU``S>23J|9b30otEYR1QOq@vs#JrKv?w2f#sqA
z5K#_T6EYDMfI&ntFe8eX4-g`1084Quq7>8vnnD0E<_5c6Cx!)82T1nOX&G|ZhD-EI
zg6Q$dp(wWKl!CEPI0`7jBa1@7uShaLq}@)G1cHhv6D2{PBKC_AWv+UwtB8Y6k~1?b
z!%-}4c93S;;dGH4Wnz)TN!zWEb3il46&Dw|93>o07m+k%iyUSrD&m}`Vu!tfC<_Z>
zdp!=EQU5nefHq_Qs)%wi+@vf86jYsgLXFc*R7sKo7#MG_#>fDMCTf;NM53-rY$R&y
z5x}uw1fq&8`a(MG(z84-WP5q<Zjtwv3jEh?>(+GiJq&s@Ig2Ok(`hkK16>Coiz+@@
zX_v)Dofhh}GiDT5l<2gBFk<MaVd#=!0&p(MadrlsXE~fPmLaME4R7<x>-7p=Z;n^f
zVafj9BoLgE-kJV%`b4^Q^<tZAMOR0MGgp7;?P;Y~AdLKQW&Y{(-`~i}%4zo8T(Clu
zs=1kwet#0@<G>7%$1N&8!Ke5DQeCbhDq>0y^C=3DDgNy7HHqauUJORDYeR4F0~3FU
zgX+ZcL8)n&Pz=J20|uP36e*J0>!A^g<1dRy{A=RyZv)}R?D37*is>eX6+p9bQ&nPA
ziphQv6A&}4N>VV#0PIhz>gBNcv9+EsYQ`Y;1mcZ>b$U$mu2S;ev96#0-<)+SOUKT-
z>h0KBS7+j}v!1XP-UH5$Ht$HC7RN4llrY;RS!zlGa#R6`9a?HA2I7){r1~e$3ck5o
z2_-H$>z;r7@!u6re)`tog%|Uw==sc7K1n^ZddlT@GmN|M%gK6m&#M>qU+Y}2^c<zL
z554o!JLmV7&(}YawslqI#+0hGS+D)-=^y`c^5i9*Gdj}doB&Omv-V8(OgY`Ua^-Um
zuXS4;6YH){Px<1J(s#QjKK3?nq#bi_J-0IDp|rClQyxvbDq41JTqUG@8pwO#jfJu9
z(^pe#PcGPdHmh|}$MuW9=(OCp`1!TD)n`JhUc21Y`{YquQy(y%VV5}R4a+NcbzRuA
z^M%XDneU%(IJ4hhxu4nQpFXYUR`0X1D-RvZ`?&qz`nL5ODvmwb=6LzzC(g9|%hRyy
zg`IPvZ~nYF_h$C$wVPqbmTQ|@FFA9oKl?f}w<|L1NNaItSI^%w-xIi(4|TuC<>ho9
zY2BWiMs@t;vqCXi_<8Py%uRV2&-bSMG+Ou_bJ?U>U8(u!3*DdOahrpa4lZ$Bb=nS|
ztF1eGWqOZ>D_OiZ$y$sOmWvV7M{<LOg4v9d$BdLa`tV-e3YhiOxdq|Wu1H?d+QpUC
z8*?h5Mz`U=Ybv*YIp=R%iVjR#=j|zceQp+bZZdb|M(#)d$^63v%RfrLw3cjR*f(}O
zyD7Bot?Rd5`OD^0vwJie?SoU3d7ZG#4Ax(WBZ=x5uf)qNcX_{$gXP6A!>n`KBR+qN
zr~qDU3WRO?ulDZOgMeVu*D`MCjyT1ofzsu&SiO9HjemKIpA+<R>{@HPr9Bi0;bH}}
zhl1g#rQN3Y;o?=a;Q#7tLJ#6ylop%bp*Dazw+A>SSp=LB8T=3;Tyz*&6C=PzQvo0$
z9s-ah5HcXnLLv)d2g^6TU9IG-ve0O$a+M6&!%sH-Qbmba2%@d6&Dcg6CAo<}9LEum
zBuLVLdl;h2!iulm5ROjoS2C!_B}V;nAfg1MFi`dScqyjX^!lNaQMEk~SST`NC>%A$
zgEac35Yg_75XcCLWI}>}NG1}KgK^gcKOqLi5U!x2SPmtX!<f3=L!wE|g+fUsBMEri
zCWjlaIf@tpK*a#tvA886(74Qnt2R3ZpD`e4pkNI~o@6WxHya`ggjV8?p2GmebvSXk
zj&~MGbWj!&ChAD&d54ZcN+#2CmnQ?&alG#xdg2x&xki$L_I}1W{IbOR6fldW4SYa(
z43LZw8A)hj*#5v0MG~|LeoLbyhkT0N7l{M|exJ${0t0C!hMN-l$FpW}N&c9c-7=Rw
z77GX#$om*S#jpmJ=XnE7lE}b|e%9b)IbJlI8MDBe<8HX!iL!U;Eeqq#P1r9E>ZQy)
z6iJRYm}r4D&;l|Wc#_A;SyD7ZhC(9h4-e~2kWJ8wBe8g7Nr?GH`I|s>TmJz)Rze*&
z3FD}LkjCHQQHo~>VrYtpH>o4)Tm1|N_$`_lH!>=4$Ke`>Yg7n~s`EIz#^D+j0;B3Y
z&aQg~mv(rQfzNAg`Zm0UNVbbfIIbFT?}f~RdoP5~)#u|g@A#SL-5X~TEr1?~sv)X`
zs20+DJg;~6wH+mTXQ#;j{0&vTc1y?Zbgcsfz;|~}t6I2m$Fz+v{%ldODWkSUTiM)v
z-w(8puJ753T$D>H<SX+%Z)P6)09=~+lEXcrW#yl;-l{t=p?pzM0R+O-!WAi+oO+W#
zSxgCsU?axF42Q(2_X8t_)a`-sD2(&m1@HgyShTF5>|~^yRO@$qJ`YKnWXa<!tT92)
zM^9}(OneY2f4^ehf=IdXaK*fW%GG_4>T@41SkqSn$&=IxO09)IJaMhJ;sD)y;zUk(
zQDNne<^|;Z)`O=`rKCNm3H{pe+hjo{oUJ9liJ<?#HxYN;KCtxNcMdeyk88r61G?kR
zK{H`yNn=C$%=Bk0;6UBnyq!~Bmoo1E^NO^uQ1%zv(yRH$_dT!=Oi7(?%RE?X&U^9Z
a<{j5Fvb~xMO*$nBP6@XX2(d%=68{4*B&}Ki

literal 0
HcmV?d00001

diff --git a/test/unit/api_spec.js b/test/unit/api_spec.js
index 0729875b6..892166732 100644
--- a/test/unit/api_spec.js
+++ b/test/unit/api_spec.js
@@ -5542,5 +5542,39 @@ small scripts as well as for`);
         await loadingTask.destroy();
       });
     });
+
+    describe("Page labels", function () {
+      it("extract page and check labels", async function () {
+        let loadingTask = getDocument(
+          buildGetDocumentParams("labelled_pages.pdf")
+        );
+        const pdfDoc = await loadingTask.promise;
+        let labels = await pdfDoc.getPageLabels();
+        expect(labels).toEqual([
+          "i" /* Page 0 */,
+          "ii" /* Page 1 */,
+          "iii" /* Page 2 */,
+          "iv" /* Page 3 */,
+          "1" /* Page 4 */,
+          "2" /* Page 5 */,
+          "3" /* Page 6 */,
+          "a" /* Page 7 */,
+          "b" /* Page 8 */,
+          "4" /* Page 9 */,
+          "5" /* Page 10 */,
+        ]);
+
+        const data = await pdfDoc.extractPages({
+          document: null,
+          includePages: [0, 1, 5, 7, 10],
+        });
+        await loadingTask.destroy();
+        loadingTask = getDocument(data);
+        const newPdfDoc = await loadingTask.promise;
+        labels = await newPdfDoc.getPageLabels();
+        expect(labels).toEqual(["i", "ii", "1", "a", "5"]);
+        await loadingTask.destroy();
+      });
+    });
   });
 });