diff --git a/.gitignore b/.gitignore
index 651cad7394..7e4a4b1d5a 100644
--- a/.gitignore
+++ b/.gitignore
@@ -12,3 +12,4 @@ env.bat
 node_modules/
 *~
 .*.swp
+.cache/
diff --git a/src/rime/dict/dict_compiler.cc b/src/rime/dict/dict_compiler.cc
index 22fe0cc691..c448f0499f 100644
--- a/src/rime/dict/dict_compiler.cc
+++ b/src/rime/dict/dict_compiler.cc
@@ -236,22 +236,24 @@ bool DictCompiler::BuildTable(int table_index,
     for (const auto& s : collector.syllabary) {
       syllable_to_id[s] = syllable_id++;
     }
-    for (RawDictEntry& r : collector.entries) {
+    for (const auto& r : collector.entries) {
       Code code;
-      for (const auto& s : r.raw_code) {
+      for (const auto& s : r->raw_code) {
         code.push_back(syllable_to_id[s]);
       }
-      DictEntryList* ls = vocabulary.LocateEntries(code);
+      auto ls = vocabulary.LocateEntries(code);
       if (!ls) {
         LOG(ERROR) << "Error locating entries in vocabulary.";
         continue;
       }
-      auto e = New<DictEntry>();
+      auto e = New<ShortDictEntry>();
       e->code.swap(code);
-      e->text.swap(r.text);
-      e->weight = log(r.weight > 0 ? r.weight : DBL_EPSILON);
+      e->text.swap(r->text);
+      e->weight = log(r->weight > 0 ? r->weight : DBL_EPSILON);
       ls->push_back(e);
     }
+    // release memory in time to reduce peak memory usage
+    vector<of<RawDictEntry>>().swap(collector.entries);
     if (settings->sort_order() != "original") {
       vocabulary.SortHomophones();
     }
diff --git a/src/rime/dict/entry_collector.cc b/src/rime/dict/entry_collector.cc
index ad5857b454..f3232c470d 100644
--- a/src/rime/dict/entry_collector.cc
+++ b/src/rime/dict/entry_collector.cc
@@ -206,7 +206,7 @@ void EntryCollector::CreateEntry(const string &word,
     words[e.text][code_str] += e.weight;
     total_weight[e.text] += e.weight;
   }
-  entries.push_back(e);
+  entries.emplace_back(New<RawDictEntry>(e));
   ++num_entries;
 }
 
@@ -240,10 +240,10 @@ void EntryCollector::Dump(const string& file_name) const {
     out << "# - " << syllable << std::endl;
   }
   out << std::endl;
-  for (const RawDictEntry& e : entries) {
-    out << e.text << '\t'
-        << e.raw_code.ToString() << '\t'
-        << e.weight << std::endl;
+  for (const auto &e : entries) {
+    out << e->text << '\t'
+        << e->raw_code.ToString() << '\t'
+        << e->weight << std::endl;
   }
   out.close();
 }
diff --git a/src/rime/dict/entry_collector.h b/src/rime/dict/entry_collector.h
index c0ce2c7ea8..77bc7ae2ee 100644
--- a/src/rime/dict/entry_collector.h
+++ b/src/rime/dict/entry_collector.h
@@ -35,7 +35,7 @@ class EntryCollector : public PhraseCollector {
  public:
   Syllabary syllabary;
   bool build_syllabary = true;
-  vector<RawDictEntry> entries;
+  vector<of<RawDictEntry>> entries;
   size_t num_entries = 0;
   ReverseLookupTable stems;
 
diff --git a/src/rime/dict/table.cc b/src/rime/dict/table.cc
index 5d19539217..f0059c2405 100644
--- a/src/rime/dict/table.cc
+++ b/src/rime/dict/table.cc
@@ -518,7 +518,7 @@ table::TailIndex* Table::BuildTailIndex(const Code& prefix,
   return index;
 }
 
-Array<table::Entry>* Table::BuildEntryArray(const DictEntryList& entries) {
+Array<table::Entry>* Table::BuildEntryArray(const ShortDictEntryList& entries) {
   auto array = CreateArray<table::Entry>(entries.size());
   if (!array) {
     return NULL;
@@ -531,7 +531,7 @@ Array<table::Entry>* Table::BuildEntryArray(const DictEntryList& entries) {
   return array;
 }
 
-bool Table::BuildEntryList(const DictEntryList& src,
+bool Table::BuildEntryList(const ShortDictEntryList& src,
                            List<table::Entry>* dest) {
   if (!dest)
     return false;
@@ -549,7 +549,7 @@ bool Table::BuildEntryList(const DictEntryList& src,
   return true;
 }
 
-bool Table::BuildEntry(const DictEntry& dict_entry, table::Entry* entry) {
+bool Table::BuildEntry(const ShortDictEntry& dict_entry, table::Entry* entry) {
   if (!entry)
     return false;
   if (!AddString(dict_entry.text, &entry->text, dict_entry.weight)) {
diff --git a/src/rime/dict/table.h b/src/rime/dict/table.h
index 3c14bddd62..1e4c018c42 100644
--- a/src/rime/dict/table.h
+++ b/src/rime/dict/table.h
@@ -166,9 +166,9 @@ class Table : public MappedFile {
                                    const Vocabulary& vocabulary);
   bool BuildPhraseIndex(Code code, const Vocabulary& vocabulary,
                         map<string, int>* index_data);
-  Array<table::Entry>* BuildEntryArray(const DictEntryList& entries);
-  bool BuildEntryList(const DictEntryList& src, List<table::Entry>* dest);
-  bool BuildEntry(const DictEntry& dict_entry, table::Entry* entry);
+  Array<table::Entry>* BuildEntryArray(const ShortDictEntryList& entries);
+  bool BuildEntryList(const ShortDictEntryList& src, List<table::Entry>* dest);
+  bool BuildEntry(const ShortDictEntry& dict_entry, table::Entry* entry);
 
   string GetString(const table::StringType& x);
   bool AddString(const string& src, table::StringType* dest,
diff --git a/src/rime/dict/vocabulary.cc b/src/rime/dict/vocabulary.cc
index b4786d0c36..18a8577e5d 100644
--- a/src/rime/dict/vocabulary.cc
+++ b/src/rime/dict/vocabulary.cc
@@ -5,6 +5,7 @@
 // 2011-07-24 GONG Chen <chen.sst@gmail.com>
 //
 #include <algorithm>
+#include <iterator>
 #include <sstream>
 #include <utility>
 #include <rime/dict/vocabulary.h>
@@ -59,6 +60,18 @@ string Code::ToString() const {
   return stream.str();
 }
 
+inline ShortDictEntry DictEntry::ToShort() const {
+  return {text, code, weight};
+}
+
+bool ShortDictEntry::operator< (const ShortDictEntry& other) const {
+  // Sort different entries sharing the same code by weight desc.
+  if (weight != other.weight)
+    return weight > other.weight;
+  // reduce carbon emission
+  return 0;  //text < other.text;
+}
+
 bool DictEntry::operator< (const DictEntry& other) const {
   // Sort different entries sharing the same code by weight desc.
   if (weight != other.weight)
@@ -72,16 +85,34 @@ inline bool dereference_less(const T& a, const T& b) {
   return *a < *b;
 }
 
+template <typename C>
+inline void sort(C &container) {
+  std::sort(std::begin(container), std::end(container), dereference_less<typename C::value_type>);
+}
+
+template <typename C>
+inline void sort_range(C &container, size_t start, size_t count) {
+  if (start >= container.size())
+    return;
+  auto i(std::begin(container) + start);
+  auto j(start + count >= container.size() ? std::end(container) : i + count);
+  std::sort(i, j, dereference_less<typename C::value_type>);
+}
+
+void ShortDictEntryList::Sort() {
+  sort(*this);
+}
+
+void ShortDictEntryList::SortRange(size_t start, size_t count) {
+  sort_range(*this, start, count);
+}
+
 void DictEntryList::Sort() {
-  std::sort(begin(), end(), dereference_less<DictEntryList::value_type>);
+  sort(*this);
 }
 
 void DictEntryList::SortRange(size_t start, size_t count) {
-  if (start >= size())
-    return;
-  iterator i(begin() + start);
-  iterator j(start + count >= size() ? end() : i + count);
-  std::sort(i, j, dereference_less<DictEntryList::value_type>);
+  sort_range(*this, start, count);
 }
 
 void DictEntryFilterBinder::AddFilter(DictEntryFilter filter) {
@@ -96,7 +127,7 @@ void DictEntryFilterBinder::AddFilter(DictEntryFilter filter) {
   }
 }
 
-DictEntryList* Vocabulary::LocateEntries(const Code& code) {
+ShortDictEntryList* Vocabulary::LocateEntries(const Code& code) {
   Vocabulary* v = this;
   size_t n = code.size();
   for (size_t i = 0; i < n; ++i) {
diff --git a/src/rime/dict/vocabulary.h b/src/rime/dict/vocabulary.h
index c6a985c6dd..563622f09d 100644
--- a/src/rime/dict/vocabulary.h
+++ b/src/rime/dict/vocabulary.h
@@ -30,20 +30,36 @@ class Code : public vector<SyllableId> {
   string ToString() const;
 };
 
+struct ShortDictEntry {
+  string text;
+  Code code;  // multi-syllable code from prism
+  double weight = 0.0;
+
+  ShortDictEntry() = default;
+  bool operator< (const ShortDictEntry& other) const;
+};
+
 struct DictEntry {
   string text;
   string comment;
   string preedit;
-  double weight = 0.0;
-  int commit_count = 0;
   Code code;  // multi-syllable code from prism
   string custom_code;  // user defined code
+  double weight = 0.0;
+  int commit_count = 0;
   int remaining_code_length = 0;
 
   DictEntry() = default;
+  ShortDictEntry ToShort() const;
   bool operator< (const DictEntry& other) const;
 };
 
+class ShortDictEntryList : public vector<of<ShortDictEntry>> {
+ public:
+  void Sort();
+  void SortRange(size_t start, size_t count);
+};
+
 class DictEntryList : public vector<of<DictEntry>> {
  public:
   void Sort();
@@ -64,13 +80,13 @@ class DictEntryFilterBinder {
 class Vocabulary;
 
 struct VocabularyPage {
-  DictEntryList entries;
+  ShortDictEntryList entries;
   an<Vocabulary> next_level;
 };
 
 class Vocabulary : public map<int, VocabularyPage> {
  public:
-  DictEntryList* LocateEntries(const Code& code);
+  ShortDictEntryList* LocateEntries(const Code& code);
   void SortHomophones();
 };
 
diff --git a/test/table_test.cc b/test/table_test.cc
index 3724dd25a9..d78e02d654 100644
--- a/test/table_test.cc
+++ b/test/table_test.cc
@@ -44,7 +44,7 @@ rime::the<rime::Table> RimeTableTest::table_;
 
 void RimeTableTest::PrepareSampleVocabulary(rime::Syllabary& syll,
                                             rime::Vocabulary& voc) {
-  auto d = rime::New<rime::DictEntry>();
+  auto d = rime::New<rime::ShortDictEntry>();
   syll.insert("0");
   // no entries for '0', however
   syll.insert("1");
@@ -52,26 +52,26 @@ void RimeTableTest::PrepareSampleVocabulary(rime::Syllabary& syll,
   d->text = "yi";
   d->weight = 1.0;
   voc[1].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   syll.insert("2");
   d->code.back() = 2;
   d->text = "er";
   voc[2].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   d->text = "liang";
   voc[2].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   d->text = "lia";
   voc[2].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   syll.insert("3");
   d->code.back() = 3;
   d->text = "san";
   voc[3].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   d->text = "sa";
   voc[3].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   syll.insert("4");
   auto lv2 = rime::New<rime::Vocabulary>();
   voc[1].next_level = lv2;
@@ -84,11 +84,11 @@ void RimeTableTest::PrepareSampleVocabulary(rime::Syllabary& syll,
   d->code.push_back(3);
   d->text = "yi-er-san";
   (*lv3)[3].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   d->code.push_back(4);
   d->text = "yi-er-san-si";
   (*lv4)[-1].entries.push_back(d);
-  d = rime::New<rime::DictEntry>(*d);
+  d = rime::New<rime::ShortDictEntry>(*d);
   d->code.resize(3);
   d->code.push_back(2);
   d->code.push_back(1);