From 5b5d6cca9a4202a78f15edba53715e2a0d291d6b Mon Sep 17 00:00:00 2001
From: ksqsf <i@ksqsf.moe>
Date: Fri, 6 Mar 2026 20:41:53 +0100
Subject: [PATCH] feat(syllabifier): trim leading delimiters

---
 src/rime/algo/syllabifier.cc   | 14 ++++++++--
 src/rime/gear/abc_segmentor.cc |  2 +-
 test/syllabifier_test.cc       | 51 ++++++++++++++++++++++++++++++++++
 3 files changed, 63 insertions(+), 4 deletions(-)

diff --git a/src/rime/algo/syllabifier.cc b/src/rime/algo/syllabifier.cc
index d3577628ac..b7ed6b7973 100644
--- a/src/rime/algo/syllabifier.cc
+++ b/src/rime/algo/syllabifier.cc
@@ -53,12 +53,19 @@ int Syllabifier::BuildSyllableGraph(const string& input,
 
     if (current_pos > farthest)
       farthest = current_pos;
-    DLOG(INFO) << "current_pos: " << current_pos;
+
+    // consume leading delimiters
+    size_t begin_pos = current_pos;
+    while (begin_pos < input.length() &&
+           delimiters_.find(input[begin_pos]) != string::npos)
+      ++begin_pos;
+    DLOG(INFO) << "current_pos: " << current_pos
+               << ", begin_pos: " << begin_pos;
 
     // see where we can go by advancing a syllable
     vector<Prism::Match> matches;
     set<SyllableId> exact_match_syllables;
-    auto current_input = input.substr(current_pos);
+    auto current_input = input.substr(begin_pos);
     prism.CommonPrefixSearch(current_input, &matches);
     if (corrector_) {
       for (auto& m : matches) {
@@ -78,12 +85,13 @@ int Syllabifier::BuildSyllableGraph(const string& input,
       }
     }
 
+    size_t leading_gap = begin_pos - current_pos;
     if (!matches.empty()) {
       auto& end_vertices(graph->edges[current_pos]);
       for (const auto& m : matches) {
         if (m.length == 0)
           continue;
-        size_t end_pos = current_pos + m.length;
+        size_t end_pos = current_pos + leading_gap + m.length;
         // consume trailing delimiters
         while (end_pos < input.length() &&
                delimiters_.find(input[end_pos]) != string::npos)
diff --git a/src/rime/gear/abc_segmentor.cc b/src/rime/gear/abc_segmentor.cc
index f99bd55f24..1e9deb8883 100644
--- a/src/rime/gear/abc_segmentor.cc
+++ b/src/rime/gear/abc_segmentor.cc
@@ -44,7 +44,7 @@ bool AbcSegmentor::Proceed(Segmentation* segmentation) {
   bool expecting_an_initial = true;
   for (; k < input.length(); ++k) {
     bool is_letter = alphabet_.find(input[k]) != string::npos;
-    bool is_delimiter = (k != j) && (delimiter_.find(input[k]) != string::npos);
+    bool is_delimiter = (k != 0) && (delimiter_.find(input[k]) != string::npos);
     if (!is_letter && !is_delimiter)
       break;
     bool is_initial = initials_.find(input[k]) != string::npos;
diff --git a/test/syllabifier_test.cc b/test/syllabifier_test.cc
index 6dc4430d34..fa748ff39d 100644
--- a/test/syllabifier_test.cc
+++ b/test/syllabifier_test.cc
@@ -160,3 +160,54 @@ TEST_F(RimeSyllabifierTest, TransposedSyllableGraph) {
   ASSERT_FALSE(NULL == g.indices[0][syllable_id_["chan"]][0]);
   EXPECT_EQ(4, g.indices[0][syllable_id_["chan"]][0]->end_pos);
 }
+
+TEST_F(RimeSyllabifierTest, TrimLeadingDelimiters) {
+  rime::Syllabifier s(" '");
+  rime::SyllableGraph g;
+  const rime::string input("''a");
+  s.BuildSyllableGraph(input, *prism_, &g);
+  EXPECT_EQ(input.length(), g.input_length);
+  EXPECT_EQ(input.length(), g.interpreted_length);
+  EXPECT_EQ(2, g.vertices.size());
+  ASSERT_FALSE(g.vertices.end() == g.vertices.find(3));
+  EXPECT_EQ(rime::kNormalSpelling, g.vertices[1]);
+  rime::SpellingMap& sp(g.edges[0][3]);
+  EXPECT_EQ(1, sp.size());
+  ASSERT_FALSE(sp.end() == sp.find(syllable_id_["a"]));
+  EXPECT_EQ(rime::kNormalSpelling, sp[0].type);
+  EXPECT_EQ(0.0, sp[0].credibility);
+}
+
+TEST_F(RimeSyllabifierTest, TrimTrailingDelimiters) {
+  rime::Syllabifier s(" '");
+  rime::SyllableGraph g;
+  const rime::string input("a''");
+  s.BuildSyllableGraph(input, *prism_, &g);
+  EXPECT_EQ(input.length(), g.input_length);
+  EXPECT_EQ(input.length(), g.interpreted_length);
+  EXPECT_EQ(2, g.vertices.size());
+  ASSERT_FALSE(g.vertices.end() == g.vertices.find(3));
+  EXPECT_EQ(rime::kNormalSpelling, g.vertices[1]);
+  rime::SpellingMap& sp(g.edges[0][3]);
+  EXPECT_EQ(1, sp.size());
+  ASSERT_FALSE(sp.end() == sp.find(syllable_id_["a"]));
+  EXPECT_EQ(rime::kNormalSpelling, sp[0].type);
+  EXPECT_EQ(0.0, sp[0].credibility);
+}
+
+TEST_F(RimeSyllabifierTest, TrimBothLeadingAndTrailingDelimiters) {
+  rime::Syllabifier s(" '");
+  rime::SyllableGraph g;
+  const rime::string input("''a''");
+  s.BuildSyllableGraph(input, *prism_, &g);
+  EXPECT_EQ(input.length(), g.input_length);
+  EXPECT_EQ(input.length(), g.interpreted_length);
+  EXPECT_EQ(2, g.vertices.size());
+  ASSERT_FALSE(g.vertices.end() == g.vertices.find(5));
+  EXPECT_EQ(rime::kNormalSpelling, g.vertices[1]);
+  rime::SpellingMap& sp(g.edges[0][5]);
+  EXPECT_EQ(1, sp.size());
+  ASSERT_FALSE(sp.end() == sp.find(syllable_id_["a"]));
+  EXPECT_EQ(rime::kNormalSpelling, sp[0].type);
+  EXPECT_EQ(0.0, sp[0].credibility);
+}