Use string pools in some dictionaries, names, and strings.

BUG=pdfium:597

Review-Url: https://codereview.chromium.org/2345063002

10 files changed, 95 insertions, 48 deletions

diff --git a/core/fpdfapi/fpdf_parser/cfdf_document.cpp b/core/fpdfapi/fpdf_parser/cfdf_document.cpp
index c602f634ad..128c513c60 100644
--- a/core/fpdfapi/fpdf_parser/cfdf_document.cpp
+++ b/core/fpdfapi/fpdf_parser/cfdf_document.cpp
@@ -14,18 +14,21 @@ CFDF_Document::CFDF_Document()
     : CPDF_IndirectObjectHolder(),
       m_pRootDict(nullptr),
       m_pFile(nullptr),
-      m_bOwnFile(FALSE) {}
+      m_bOwnFile(FALSE),
+      m_pByteStringPool(WrapUnique(new CFX_ByteStringPool)) {}
 
 CFDF_Document::~CFDF_Document() {
   if (m_bOwnFile && m_pFile)
     m_pFile->Release();
+  m_pByteStringPool.Clear();  // Make weak.
 }
 
 CFDF_Document* CFDF_Document::CreateNewDoc() {
   CFDF_Document* pDoc = new CFDF_Document;
-  pDoc->m_pRootDict = new CPDF_Dictionary;
+  pDoc->m_pRootDict = new CPDF_Dictionary(pDoc->GetByteStringPool());
   pDoc->AddIndirectObject(pDoc->m_pRootDict);
-  pDoc->m_pRootDict->SetFor("FDF", new CPDF_Dictionary);
+  pDoc->m_pRootDict->SetFor("FDF",
+                            new CPDF_Dictionary(pDoc->GetByteStringPool()));
   return pDoc;
 }
 
diff --git a/core/fpdfapi/fpdf_parser/cpdf_dictionary.cpp b/core/fpdfapi/fpdf_parser/cpdf_dictionary.cpp
index 1a84fcb142..0b415d370d 100644
--- a/core/fpdfapi/fpdf_parser/cpdf_dictionary.cpp
+++ b/core/fpdfapi/fpdf_parser/cpdf_dictionary.cpp
@@ -19,7 +19,8 @@
 #include "third_party/base/stl_util.h"
 #include "third_party/base/logging.h"
 
-CPDF_Dictionary::CPDF_Dictionary() {}
+CPDF_Dictionary::CPDF_Dictionary(const CFX_WeakPtr<CFX_ByteStringPool>& pPool)
+    : m_pPool(pPool) {}
 
 CPDF_Dictionary::~CPDF_Dictionary() {
   // Mark the object as deleted so that it will not be deleted again
@@ -61,7 +62,7 @@ CPDF_Object* CPDF_Dictionary::CloneNonCyclic(
     bool bDirect,
     std::set<const CPDF_Object*>* pVisited) const {
   pVisited->insert(this);
-  CPDF_Dictionary* pCopy = new CPDF_Dictionary();
+  CPDF_Dictionary* pCopy = new CPDF_Dictionary(m_pPool);
   for (const auto& it : *this) {
     CPDF_Object* value = it.second;
     if (!pdfium::ContainsKey(*pVisited, value)) {
@@ -174,7 +175,7 @@ void CPDF_Dictionary::SetFor(const CFX_ByteString& key, CPDF_Object* pObj) {
   auto it = m_Map.find(key);
   if (it == m_Map.end()) {
     if (pObj)
-      m_Map.insert(std::make_pair(key, pObj));
+      m_Map.insert(std::make_pair(MaybeIntern(key), pObj));
     return;
   }
 
@@ -211,7 +212,7 @@ void CPDF_Dictionary::ReplaceKey(const CFX_ByteString& oldkey,
     new_it->second->Release();
     new_it->second = old_it->second;
   } else {
-    m_Map.insert(std::make_pair(newkey, old_it->second));
+    m_Map.insert(std::make_pair(MaybeIntern(newkey), old_it->second));
   }
   m_Map.erase(old_it);
 }
@@ -222,12 +223,12 @@ void CPDF_Dictionary::SetIntegerFor(const CFX_ByteString& key, int i) {
 
 void CPDF_Dictionary::SetNameFor(const CFX_ByteString& key,
                                  const CFX_ByteString& name) {
-  SetFor(key, new CPDF_Name(name));
+  SetFor(key, new CPDF_Name(MaybeIntern(name)));
 }
 
 void CPDF_Dictionary::SetStringFor(const CFX_ByteString& key,
                                    const CFX_ByteString& str) {
-  SetFor(key, new CPDF_String(str, FALSE));
+  SetFor(key, new CPDF_String(MaybeIntern(str), FALSE));
 }
 
 void CPDF_Dictionary::SetReferenceFor(const CFX_ByteString& key,
@@ -265,3 +266,7 @@ void CPDF_Dictionary::SetMatrixFor(const CFX_ByteString& key,
   pArray->AddNumber(matrix.f);
   SetFor(key, pArray);
 }
+
+CFX_ByteString CPDF_Dictionary::MaybeIntern(const CFX_ByteString& str) {
+  return m_pPool ? m_pPool->Intern(str) : str;
+}
diff --git a/core/fpdfapi/fpdf_parser/cpdf_document.cpp b/core/fpdfapi/fpdf_parser/cpdf_document.cpp
index 1694566f1b..7e75a437a8 100644
--- a/core/fpdfapi/fpdf_parser/cpdf_document.cpp
+++ b/core/fpdfapi/fpdf_parser/cpdf_document.cpp
@@ -383,14 +383,15 @@ void ProcessNonbCJK(CPDF_Dictionary* pBaseDict,
   pBaseDict->SetFor("Widths", pWidths);
 }
 
-CPDF_Dictionary* CalculateFontDesc(CFX_ByteString basefont,
+CPDF_Dictionary* CalculateFontDesc(CPDF_Document* pDoc,
+                                   CFX_ByteString basefont,
                                    int flags,
                                    int italicangle,
                                    int ascend,
                                    int descend,
                                    CPDF_Array* bbox,
                                    int32_t stemV) {
-  CPDF_Dictionary* pFontDesc = new CPDF_Dictionary;
+  CPDF_Dictionary* pFontDesc = new CPDF_Dictionary(pDoc->GetByteStringPool());
   pFontDesc->SetNameFor("Type", "FontDescriptor");
   pFontDesc->SetNameFor("FontName", basefont);
   pFontDesc->SetIntegerFor("Flags", flags);
@@ -413,7 +414,8 @@ CPDF_Document::CPDF_Document(std::unique_ptr<CPDF_Parser> pParser)
       m_iFirstPageNo(0),
       m_dwFirstPageObjNum(0),
       m_pDocPage(new CPDF_DocPageData(this)),
-      m_pDocRender(new CPDF_DocRenderData(this)) {
+      m_pDocRender(new CPDF_DocRenderData(this)),
+      m_pByteStringPool(WrapUnique(new CFX_ByteStringPool)) {
   if (pParser)
     SetLastObjNum(m_pParser->GetLastObjNum());
 }
@@ -421,6 +423,7 @@ CPDF_Document::CPDF_Document(std::unique_ptr<CPDF_Parser> pParser)
 CPDF_Document::~CPDF_Document() {
   delete m_pDocPage;
   CPDF_ModuleMgr::Get()->GetPageModule()->ClearStockFont(this);
+  m_pByteStringPool.Clear();  // Make weak.
 }
 
 CPDF_Object* CPDF_Document::ParseIndirectObject(uint32_t objnum) {
@@ -687,20 +690,21 @@ CPDF_Image* CPDF_Document::LoadImageF(CPDF_Object* pObj) {
 
 void CPDF_Document::CreateNewDoc() {
   ASSERT(!m_pRootDict && !m_pInfoDict);
-  m_pRootDict = new CPDF_Dictionary;
+  m_pRootDict = new CPDF_Dictionary(m_pByteStringPool);
   m_pRootDict->SetNameFor("Type", "Catalog");
   AddIndirectObject(m_pRootDict);
-  CPDF_Dictionary* pPages = new CPDF_Dictionary;
+
+  CPDF_Dictionary* pPages = new CPDF_Dictionary(m_pByteStringPool);
   pPages->SetNameFor("Type", "Pages");
   pPages->SetNumberFor("Count", 0);
   pPages->SetFor("Kids", new CPDF_Array);
   m_pRootDict->SetReferenceFor("Pages", this, AddIndirectObject(pPages));
-  m_pInfoDict = new CPDF_Dictionary;
+  m_pInfoDict = new CPDF_Dictionary(m_pByteStringPool);
   AddIndirectObject(m_pInfoDict);
 }
 
 CPDF_Dictionary* CPDF_Document::CreateNewPage(int iPage) {
-  CPDF_Dictionary* pDict = new CPDF_Dictionary;
+  CPDF_Dictionary* pDict = new CPDF_Dictionary(m_pByteStringPool);
   pDict->SetNameFor("Type", "Page");
   uint32_t dwObjNum = AddIndirectObject(pDict);
   if (InsertNewPage(this, iPage, pDict, m_PageList) < 0) {
@@ -743,7 +747,7 @@ size_t CPDF_Document::CalculateEncodingDict(int charset,
   }
   if (i == FX_ArraySize(g_FX_CharsetUnicodes))
     return i;
-  CPDF_Dictionary* pEncodingDict = new CPDF_Dictionary;
+  CPDF_Dictionary* pEncodingDict = new CPDF_Dictionary(m_pByteStringPool);
   pEncodingDict->SetNameFor("BaseEncoding", "WinAnsiEncoding");
   CPDF_Array* pArray = new CPDF_Array;
   pArray->AddInteger(128);
@@ -765,7 +769,7 @@ CPDF_Dictionary* CPDF_Document::ProcessbCJK(
     FX_BOOL bVert,
     CFX_ByteString basefont,
     std::function<void(FX_WCHAR, FX_WCHAR, CPDF_Array*)> Insert) {
-  CPDF_Dictionary* pFontDict = new CPDF_Dictionary;
+  CPDF_Dictionary* pFontDict = new CPDF_Dictionary(m_pByteStringPool);
   CFX_ByteString cmap;
   CFX_ByteString ordering;
   int supplement = 0;
@@ -815,7 +819,7 @@ CPDF_Dictionary* CPDF_Document::ProcessbCJK(
   pFontDict->SetNameFor("Type", "Font");
   pFontDict->SetNameFor("Subtype", "CIDFontType2");
   pFontDict->SetNameFor("BaseFont", basefont);
-  CPDF_Dictionary* pCIDSysInfo = new CPDF_Dictionary;
+  CPDF_Dictionary* pCIDSysInfo = new CPDF_Dictionary(m_pByteStringPool);
   pCIDSysInfo->SetStringFor("Registry", "Adobe");
   pCIDSysInfo->SetStringFor("Ordering", ordering);
   pCIDSysInfo->SetIntegerFor("Supplement", supplement);
@@ -840,7 +844,7 @@ CPDF_Font* CPDF_Document::AddFont(CFX_Font* pFont, int charset, FX_BOOL bVert) {
       CalculateFlags(pFont->IsBold(), pFont->IsItalic(), pFont->IsFixedWidth(),
                      false, false, charset == FXFONT_SYMBOL_CHARSET);
 
-  CPDF_Dictionary* pBaseDict = new CPDF_Dictionary;
+  CPDF_Dictionary* pBaseDict = new CPDF_Dictionary(m_pByteStringPool);
   pBaseDict->SetNameFor("Type", "Font");
   std::unique_ptr<CFX_UnicodeEncoding> pEncoding(
       new CFX_UnicodeEncoding(pFont));
@@ -907,7 +911,7 @@ CPDF_Font* CPDF_Document::AddFont(CFX_Font* pFont, int charset, FX_BOOL bVert) {
     }
   }
   CPDF_Dictionary* pFontDesc =
-      CalculateFontDesc(basefont, flags, italicangle, pFont->GetAscent(),
+      CalculateFontDesc(this, basefont, flags, italicangle, pFont->GetAscent(),
                         pFont->GetDescent(), pBBox, nStemV);
   pFontDict->SetReferenceFor("FontDescriptor", this,
                              AddIndirectObject(pFontDesc));
@@ -972,7 +976,7 @@ CPDF_Font* CPDF_Document::AddWindowsFont(LOGFONTA* pLogFont,
                  ptm->otmrcFontBox.right, ptm->otmrcFontBox.top};
   FX_Free(tm_buf);
   basefont.Replace(" ", "");
-  CPDF_Dictionary* pBaseDict = new CPDF_Dictionary;
+  CPDF_Dictionary* pBaseDict = new CPDF_Dictionary(m_pByteStringPool);
   pBaseDict->SetNameFor("Type", "Font");
   CPDF_Dictionary* pFontDict = pBaseDict;
   if (!bCJK) {
@@ -1002,8 +1006,8 @@ CPDF_Font* CPDF_Document::AddWindowsFont(LOGFONTA* pLogFont,
   for (int i = 0; i < 4; i++)
     pBBox->AddInteger(bbox[i]);
   CPDF_Dictionary* pFontDesc =
-      CalculateFontDesc(basefont, flags, italicangle, ascend, descend, pBBox,
-                        pLogFont->lfWeight / 5);
+      CalculateFontDesc(this, basefont, flags, italicangle, ascend, descend,
+                        pBBox, pLogFont->lfWeight / 5);
   pFontDesc->SetIntegerFor("CapHeight", capheight);
   pFontDict->SetReferenceFor("FontDescriptor", this,
                              AddIndirectObject(pFontDesc));
diff --git a/core/fpdfapi/fpdf_parser/cpdf_object_unittest.cpp b/core/fpdfapi/fpdf_parser/cpdf_object_unittest.cpp
index dad98cd28e..1548b1914a 100644
--- a/core/fpdfapi/fpdf_parser/cpdf_object_unittest.cpp
+++ b/core/fpdfapi/fpdf_parser/cpdf_object_unittest.cpp
@@ -66,7 +66,7 @@ class PDFObjectsTest : public testing::Test {
     m_ArrayObj->InsertAt(0, new CPDF_Number(8902));
     m_ArrayObj->InsertAt(1, new CPDF_Name("address"));
     // Dictionary object.
-    m_DictObj = new CPDF_Dictionary;
+    m_DictObj = new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
     m_DictObj->SetFor("bool", new CPDF_Boolean(false));
     m_DictObj->SetFor("num", new CPDF_Number(0.23f));
     // Stream object.
@@ -74,7 +74,7 @@ class PDFObjectsTest : public testing::Test {
     size_t buf_len = FX_ArraySize(content);
     uint8_t* buf = reinterpret_cast<uint8_t*>(malloc(buf_len));
     memcpy(buf, content, buf_len);
-    m_StreamDictObj = new CPDF_Dictionary;
+    m_StreamDictObj = new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
     m_StreamDictObj->SetFor("key1", new CPDF_String(L" test dict"));
     m_StreamDictObj->SetFor("key2", new CPDF_Number(-1));
     CPDF_Stream* stream_obj = new CPDF_Stream(buf, buf_len, m_StreamDictObj);
@@ -553,7 +553,7 @@ TEST(PDFArrayTest, GetTypeAt) {
     CPDF_Dictionary* vals[3];
     ScopedArray arr(new CPDF_Array);
     for (size_t i = 0; i < 3; ++i) {
-      vals[i] = new CPDF_Dictionary;
+      vals[i] = new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
       for (size_t j = 0; j < 3; ++j) {
         std::string key("key");
         char buf[33];
@@ -580,7 +580,7 @@ TEST(PDFArrayTest, GetTypeAt) {
     CPDF_Stream* stream_vals[3];
     ScopedArray arr(new CPDF_Array);
     for (size_t i = 0; i < 3; ++i) {
-      vals[i] = new CPDF_Dictionary;
+      vals[i] = new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
       for (size_t j = 0; j < 3; ++j) {
         std::string key("key");
         char buf[33];
@@ -625,11 +625,13 @@ TEST(PDFArrayTest, GetTypeAt) {
     arr_val->AddNumber(1);
     arr_val->AddNumber(2);
     arr->InsertAt(11, arr_val);
-    CPDF_Dictionary* dict_val = new CPDF_Dictionary;
+    CPDF_Dictionary* dict_val =
+        new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
     dict_val->SetFor("key1", new CPDF_String("Linda", false));
     dict_val->SetFor("key2", new CPDF_String("Zoe", false));
     arr->InsertAt(12, dict_val);
-    CPDF_Dictionary* stream_dict = new CPDF_Dictionary;
+    CPDF_Dictionary* stream_dict =
+        new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
     stream_dict->SetFor("key1", new CPDF_String("John", false));
     stream_dict->SetFor("key2", new CPDF_String("King", false));
     uint8_t data[] = "A stream for test";
@@ -768,7 +770,7 @@ TEST(PDFArrayTest, CloneDirectObject) {
 
 TEST(PDFDictionaryTest, CloneDirectObject) {
   CPDF_IndirectObjectHolder objects_holder;
-  ScopedDict dict(new CPDF_Dictionary);
+  ScopedDict dict(new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>()));
   dict->SetReferenceFor("foo", &objects_holder, 1234);
   ASSERT_EQ(1U, dict->GetCount());
   CPDF_Object* obj = dict->GetObjectFor("foo");
@@ -790,7 +792,8 @@ TEST(PDFObjectTest, CloneCheckLoop) {
     // Create an object with a reference loop.
     ScopedArray arr_obj(new CPDF_Array);
     // Dictionary object.
-    CPDF_Dictionary* dict_obj = new CPDF_Dictionary;
+    CPDF_Dictionary* dict_obj =
+        new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
     dict_obj->SetFor("arr", arr_obj.get());
     arr_obj->InsertAt(0, dict_obj);
 
@@ -808,7 +811,8 @@ TEST(PDFObjectTest, CloneCheckLoop) {
   {
     CPDF_IndirectObjectHolder objects_holder;
     // Create an object with a reference loop.
-    CPDF_Dictionary* dict_obj = new CPDF_Dictionary;
+    CPDF_Dictionary* dict_obj =
+        new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>());
     CPDF_Array* arr_obj = new CPDF_Array;
     objects_holder.AddIndirectObject(dict_obj);
     EXPECT_EQ(1u, dict_obj->GetObjNum());
diff --git a/core/fpdfapi/fpdf_parser/cpdf_stream.cpp b/core/fpdfapi/fpdf_parser/cpdf_stream.cpp
index cc584b78b3..3b047205c4 100644
--- a/core/fpdfapi/fpdf_parser/cpdf_stream.cpp
+++ b/core/fpdfapi/fpdf_parser/cpdf_stream.cpp
@@ -93,7 +93,7 @@ void CPDF_Stream::SetData(const uint8_t* pData, uint32_t size) {
     FXSYS_memcpy(m_pDataBuf.get(), pData, size);
   m_dwSize = size;
   if (!m_pDict)
-    m_pDict.reset(new CPDF_Dictionary);
+    m_pDict.reset(new CPDF_Dictionary(CFX_WeakPtr<CFX_ByteStringPool>()));
   m_pDict->SetIntegerFor("Length", size);
   m_pDict->RemoveFor("Filter");
   m_pDict->RemoveFor("DecodeParms");
diff --git a/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.cpp b/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.cpp
index 4bfaf99ed9..a29137cd43 100644
--- a/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.cpp
+++ b/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.cpp
@@ -37,10 +37,15 @@ struct SearchTagRecord {
 int CPDF_SyntaxParser::s_CurrentRecursionDepth = 0;
 
 CPDF_SyntaxParser::CPDF_SyntaxParser()
+    : CPDF_SyntaxParser(CFX_WeakPtr<CFX_ByteStringPool>()) {}
+
+CPDF_SyntaxParser::CPDF_SyntaxParser(
+    const CFX_WeakPtr<CFX_ByteStringPool>& pPool)
     : m_MetadataObjnum(0),
       m_pFileAccess(nullptr),
       m_pFileBuf(nullptr),
-      m_BufSize(CPDF_ModuleMgr::kFileBufSize) {}
+      m_BufSize(CPDF_ModuleMgr::kFileBufSize),
+      m_pPool(pPool) {}
 
 CPDF_SyntaxParser::~CPDF_SyntaxParser() {
   FX_Free(m_pFileBuf);
@@ -409,15 +414,14 @@ CPDF_Object* CPDF_SyntaxParser::GetObject(CPDF_IndirectObjectHolder* pObjList,
     CFX_ByteString str = ReadString();
     if (m_pCryptoHandler && bDecrypt)
       m_pCryptoHandler->Decrypt(objnum, gennum, str);
-    return new CPDF_String(str, FALSE);
+    return new CPDF_String(MaybeIntern(str), FALSE);
   }
 
   if (word == "<") {
     CFX_ByteString str = ReadHexString();
     if (m_pCryptoHandler && bDecrypt)
       m_pCryptoHandler->Decrypt(objnum, gennum, str);
-
-    return new CPDF_String(str, TRUE);
+    return new CPDF_String(MaybeIntern(str), TRUE);
   }
 
   if (word == "[") {
@@ -429,8 +433,8 @@ CPDF_Object* CPDF_SyntaxParser::GetObject(CPDF_IndirectObjectHolder* pObjList,
   }
 
   if (word[0] == '/') {
-    return new CPDF_Name(
-        PDF_NameDecode(CFX_ByteStringC(m_WordBuffer + 1, m_WordSize - 1)));
+    return new CPDF_Name(MaybeIntern(
+        PDF_NameDecode(CFX_ByteStringC(m_WordBuffer + 1, m_WordSize - 1))));
   }
 
   if (word == "<<") {
@@ -438,7 +442,7 @@ CPDF_Object* CPDF_SyntaxParser::GetObject(CPDF_IndirectObjectHolder* pObjList,
     FX_FILESIZE dwSignValuePos = 0;
 
     std::unique_ptr<CPDF_Dictionary, ReleaseDeleter<CPDF_Dictionary>> pDict(
-        new CPDF_Dictionary);
+        new CPDF_Dictionary(m_pPool));
     while (1) {
       CFX_ByteString key = GetNextWord(nullptr);
       if (key.IsEmpty())
@@ -531,14 +535,14 @@ CPDF_Object* CPDF_SyntaxParser::GetObjectForStrict(
     CFX_ByteString str = ReadString();
     if (m_pCryptoHandler)
       m_pCryptoHandler->Decrypt(objnum, gennum, str);
-    return new CPDF_String(str, FALSE);
+    return new CPDF_String(MaybeIntern(str), FALSE);
   }
 
   if (word == "<") {
     CFX_ByteString str = ReadHexString();
     if (m_pCryptoHandler)
       m_pCryptoHandler->Decrypt(objnum, gennum, str);
-    return new CPDF_String(str, TRUE);
+    return new CPDF_String(MaybeIntern(str), TRUE);
   }
 
   if (word == "[") {
@@ -551,13 +555,13 @@ CPDF_Object* CPDF_SyntaxParser::GetObjectForStrict(
   }
 
   if (word[0] == '/') {
-    return new CPDF_Name(
-        PDF_NameDecode(CFX_ByteStringC(m_WordBuffer + 1, m_WordSize - 1)));
+    return new CPDF_Name(MaybeIntern(
+        PDF_NameDecode(CFX_ByteStringC(m_WordBuffer + 1, m_WordSize - 1))));
   }
 
   if (word == "<<") {
     std::unique_ptr<CPDF_Dictionary, ReleaseDeleter<CPDF_Dictionary>> pDict(
-        new CPDF_Dictionary);
+        new CPDF_Dictionary(m_pPool));
     while (1) {
       FX_FILESIZE SavedPos = m_Pos;
       CFX_ByteString key = GetNextWord(nullptr);
@@ -987,3 +991,7 @@ void CPDF_SyntaxParser::SetEncrypt(
     std::unique_ptr<CPDF_CryptoHandler> pCryptoHandler) {
   m_pCryptoHandler = std::move(pCryptoHandler);
 }
+
+CFX_ByteString CPDF_SyntaxParser::MaybeIntern(const CFX_ByteString& str) {
+  return m_pPool ? m_pPool->Intern(str) : str;
+}
diff --git a/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.h b/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.h
index 8ca7e33824..5838fb7144 100644
--- a/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.h
+++ b/core/fpdfapi/fpdf_parser/cpdf_syntax_parser.h
@@ -9,6 +9,8 @@
 
 #include <memory>
 
+#include "core/fxcrt/include/cfx_string_pool_template.h"
+#include "core/fxcrt/include/cfx_weak_ptr.h"
 #include "core/fxcrt/include/fx_basic.h"
 
 class CPDF_CryptoHandler;
@@ -21,6 +23,7 @@ class IFX_FileRead;
 class CPDF_SyntaxParser {
  public:
   CPDF_SyntaxParser();
+  explicit CPDF_SyntaxParser(const CFX_WeakPtr<CFX_ByteStringPool>& pPool);
   ~CPDF_SyntaxParser();
 
   void InitParser(IFX_FileRead* pFileAccess, uint32_t HeaderOffset);
@@ -64,7 +67,6 @@ class CPDF_SyntaxParser {
   static int s_CurrentRecursionDepth;
 
   uint32_t GetDirectNum();
-
   FX_BOOL GetNextChar(uint8_t& ch);
   FX_BOOL GetCharAtBackward(FX_FILESIZE pos, uint8_t& ch);
   void GetNextWordInternal(bool* bIsNumber);
@@ -80,6 +82,8 @@ class CPDF_SyntaxParser {
                           uint32_t objnum,
                           uint32_t gennum);
 
+  CFX_ByteString MaybeIntern(const CFX_ByteString& str);
+
   FX_FILESIZE m_Pos;
   int m_MetadataObjnum;
   IFX_FileRead* m_pFileAccess;
@@ -91,6 +95,7 @@ class CPDF_SyntaxParser {
   std::unique_ptr<CPDF_CryptoHandler> m_pCryptoHandler;
   uint8_t m_WordBuffer[257];
   uint32_t m_WordSize;
+  CFX_WeakPtr<CFX_ByteStringPool> m_pPool;
 };
 
 #endif  // CORE_FPDFAPI_FPDF_PARSER_CPDF_SYNTAX_PARSER_H_
diff --git a/core/fpdfapi/fpdf_parser/include/cfdf_document.h b/core/fpdfapi/fpdf_parser/include/cfdf_document.h
index 834aecd1ba..25366b0cd4 100644
--- a/core/fpdfapi/fpdf_parser/include/cfdf_document.h
+++ b/core/fpdfapi/fpdf_parser/include/cfdf_document.h
@@ -9,6 +9,8 @@
 
 #include "core/fpdfapi/fpdf_parser/include/cpdf_indirect_object_holder.h"
 #include "core/fpdfapi/fpdf_parser/include/cpdf_object.h"
+#include "core/fxcrt/include/cfx_string_pool_template.h"
+#include "core/fxcrt/include/cfx_weak_ptr.h"
 #include "core/fxcrt/include/fx_basic.h"
 
 class CPDF_Dictionary;
@@ -23,6 +25,9 @@ class CFDF_Document : public CPDF_IndirectObjectHolder {
 
   FX_BOOL WriteBuf(CFX_ByteTextBuf& buf) const;
   CPDF_Dictionary* GetRoot() const { return m_pRootDict; }
+  CFX_WeakPtr<CFX_ByteStringPool> GetByteStringPool() const {
+    return m_pByteStringPool;
+  }
 
  protected:
   CFDF_Document();
@@ -31,6 +36,7 @@ class CFDF_Document : public CPDF_IndirectObjectHolder {
   CPDF_Dictionary* m_pRootDict;
   IFX_FileRead* m_pFile;
   FX_BOOL m_bOwnFile;
+  CFX_WeakPtr<CFX_ByteStringPool> m_pByteStringPool;
 };
 
 #endif  // CORE_FPDFAPI_FPDF_PARSER_INCLUDE_CFDF_DOCUMENT_H_
diff --git a/core/fpdfapi/fpdf_parser/include/cpdf_dictionary.h b/core/fpdfapi/fpdf_parser/include/cpdf_dictionary.h
index fc69ce2f4b..501b287dfc 100644
--- a/core/fpdfapi/fpdf_parser/include/cpdf_dictionary.h
+++ b/core/fpdfapi/fpdf_parser/include/cpdf_dictionary.h
@@ -11,6 +11,8 @@
 #include <set>
 
 #include "core/fpdfapi/fpdf_parser/include/cpdf_object.h"
+#include "core/fxcrt/include/cfx_string_pool_template.h"
+#include "core/fxcrt/include/cfx_weak_ptr.h"
 #include "core/fxcrt/include/fx_coordinates.h"
 #include "core/fxcrt/include/fx_string.h"
 
@@ -21,7 +23,7 @@ class CPDF_Dictionary : public CPDF_Object {
   using iterator = std::map<CFX_ByteString, CPDF_Object*>::iterator;
   using const_iterator = std::map<CFX_ByteString, CPDF_Object*>::const_iterator;
 
-  CPDF_Dictionary();
+  explicit CPDF_Dictionary(const CFX_WeakPtr<CFX_ByteStringPool>& pPool);
 
   // CPDF_Object.
   Type GetType() const override;
@@ -78,13 +80,17 @@ class CPDF_Dictionary : public CPDF_Object {
   const_iterator begin() const { return m_Map.begin(); }
   const_iterator end() const { return m_Map.end(); }
 
+  CFX_WeakPtr<CFX_ByteStringPool> GetByteStringPool() const { return m_pPool; }
+
  protected:
   ~CPDF_Dictionary() override;
 
+  CFX_ByteString MaybeIntern(const CFX_ByteString& str);
   CPDF_Object* CloneNonCyclic(
       bool bDirect,
       std::set<const CPDF_Object*>* visited) const override;
 
+  CFX_WeakPtr<CFX_ByteStringPool> m_pPool;
   std::map<CFX_ByteString, CPDF_Object*> m_Map;
 };
 
diff --git a/core/fpdfapi/fpdf_parser/include/cpdf_document.h b/core/fpdfapi/fpdf_parser/include/cpdf_document.h
index 1cc32dc3b2..383c6faeb4 100644
--- a/core/fpdfapi/fpdf_parser/include/cpdf_document.h
+++ b/core/fpdfapi/fpdf_parser/include/cpdf_document.h
@@ -13,6 +13,8 @@
 #include "core/fpdfapi/fpdf_parser/include/cpdf_indirect_object_holder.h"
 #include "core/fpdfapi/fpdf_parser/include/cpdf_object.h"
 #include "core/fpdfdoc/include/cpdf_linklist.h"
+#include "core/fxcrt/include/cfx_string_pool_template.h"
+#include "core/fxcrt/include/cfx_weak_ptr.h"
 #include "core/fxcrt/include/fx_basic.h"
 
 class CFX_Font;
@@ -47,6 +49,9 @@ class CPDF_Document : public CPDF_IndirectObjectHolder {
   CPDF_Parser* GetParser() const { return m_pParser.get(); }
   CPDF_Dictionary* GetRoot() const { return m_pRootDict; }
   CPDF_Dictionary* GetInfo() const { return m_pInfoDict; }
+  CFX_WeakPtr<CFX_ByteStringPool> GetByteStringPool() const {
+    return m_pByteStringPool;
+  }
 
   void DeletePage(int iPage);
   int GetPageCount() const;
@@ -134,6 +139,7 @@ class CPDF_Document : public CPDF_IndirectObjectHolder {
   std::unique_ptr<JBig2_DocumentContext> m_pCodecContext;
   std::unique_ptr<CPDF_LinkList> m_pLinksContext;
   CFX_ArrayTemplate<uint32_t> m_PageList;
+  CFX_WeakPtr<CFX_ByteStringPool> m_pByteStringPool;
 };
 
 #endif  // CORE_FPDFAPI_FPDF_PARSER_INCLUDE_CPDF_DOCUMENT_H_