Support full-width characters in romaji table

BUG= TEST= REF_BUG=31444698 REF_CL=132973961,133089311 REF_TIME=2016-09-13T17:49:10+09:00 REF_TIME_RAW=1473756550 +0900
google · Sep 13, 2016 · 2315f95 · 2315f95
1 parent 08ba3f9
commit 2315f95
Show file tree

Hide file tree

Showing 6 changed files with 64 additions and 8 deletions.
diff --git a/src/base/util.h b/src/base/util.h
@@ -187,6 +187,10 @@ class Util {
   static char32 UTF8ToUCS4(const char *begin,
                            const char *end,
                            size_t *mblen);
+  static char32 UTF8ToUCS4(StringPiece s) {
+    size_t mblen = 0;
+    return UTF8ToUCS4(s.data(), s.data() + s.size(), &mblen);
+  }
 
   // Converts a UCS4 code point to UTF8 string.
   static void UCS4ToUTF8(char32 c, string *output);

diff --git a/src/composer/composer.cc b/src/composer/composer.cc
@@ -328,9 +328,17 @@ void Composer::ApplyTemporaryInputMode(const string &input, bool caps_locked) {
   const config::Config::ShiftKeyModeSwitch switch_mode =
       config_->shift_key_mode_switch();
 
-  // Input is NOT an ASCII code.
+  // When input is not an ASCII code, reset the input mode to the one before
+  // temporary input mode.
   if (Util::OneCharLen(input.c_str()) != 1) {
-    SetInputMode(comeback_input_mode_);
+    // Call SetInputMode() only when the current input mode is temporary, which
+    // is detected by the if-condition below.  Without this check,
+    // SetInputMode() is called always for multi-byte charactesrs.  This causes
+    // a bug that multi-byte characters is inserted to a new chunk because
+    // |is_new_input_| is set to true in SetInputMode(); see b/31444698.
+    if (comeback_input_mode_ != input_mode_) {
+      SetInputMode(comeback_input_mode_);
+    }
     return;
   }
 

diff --git a/src/composer/composer_test.cc b/src/composer/composer_test.cc
@@ -1247,6 +1247,52 @@ TEST_F(ComposerTest, ApplyTemporaryInputMode) {
   }
 }
 
+TEST_F(ComposerTest, FullWidthCharRules_b31444698) {
+  // Construct the following romaji table:
+  //
+  // 1<tab><tab>{?}あ<tab>NewChunk NoTransliteration
+  // {?}あ1<tab><tab>{?}い<tab>
+  // か<tab><tab>{?}か<tab>NewChunk NoTransliteration
+  // {?}かか<tab><tab>{?}き<tab>
+  const int kAttrs =
+      TableAttribute::NEW_CHUNK | TableAttribute::NO_TRANSLITERATION;
+  table_->AddRuleWithAttributes("1", "",
+                                "\x7B\x3F\x7D\xE3\x81\x82",  // "{?}あ"
+                                kAttrs);
+  table_->AddRule("\x7B\x3F\x7D\xE3\x81\x82\x31",  // "{?}あ1"
+                  "",
+                  "\x7B\x3F\x7D\xE3\x81\x84");  // "{?}い"
+  table_->AddRuleWithAttributes("\xE3\x81\x8B",  // "か"
+                                "",
+                                "\x7B\x3F\x7D\xE3\x81\x8B",  // "{?}か"
+                                kAttrs);
+  table_->AddRule("\x7B\x3F\x7D\xE3\x81\x8B\xE3\x81\x8B",  // "{?}かか"
+                  "",
+                  "\x7B\x3F\x7D\xE3\x81\x8D");  // "{?}き"
+
+  // Test if "11" is transliterated to "い"
+  ASSERT_TRUE(InsertKeyWithMode("1", commands::HIRAGANA, composer_.get()));
+  EXPECT_EQ("\xE3\x81\x82",  // "あ"
+            GetPreedit(composer_.get()));
+  ASSERT_TRUE(InsertKeyWithMode("1", commands::HIRAGANA, composer_.get()));
+  EXPECT_EQ("\xE3\x81\x84",  // "い"
+            GetPreedit(composer_.get()));
+
+  composer_->Reset();
+
+  // b/31444698.  Test if "かか" is transliterated to "き"
+  ASSERT_TRUE(InsertKeyWithMode("\xE3\x81\x8B",  // "か"
+                                commands::HIRAGANA,
+                                composer_.get()));
+  EXPECT_EQ("\xE3\x81\x8B",  // "か"
+            GetPreedit(composer_.get()));
+  ASSERT_TRUE(InsertKeyWithMode("\xE3\x81\x8B",  // "か"
+                                commands::HIRAGANA,
+                                composer_.get()));
+  EXPECT_EQ("\xE3\x81\x8D",  // "き"
+            GetPreedit(composer_.get()));
+}
+
 TEST_F(ComposerTest, CopyFrom) {
   // "あ"
   table_->AddRule("a", "\xE3\x81\x82", "");

diff --git a/src/composer/key_parser.cc b/src/composer/key_parser.cc
@@ -190,10 +190,6 @@ class KeyParserData {
 
 bool KeyParser::ParseKey(const string &key_string,
                          KeyEvent *key_event) {
-  if (Util::GetFormType(key_string) != mozc::Util::HALF_WIDTH) {
-    LOG(ERROR) << "key should be half-width";
-    return false;
-  }
   vector<string> keys;
   Util::SplitStringUsing(key_string, " ", &keys);
   if (keys.empty()) {

diff --git a/src/composer/key_parser_test.cc b/src/composer/key_parser_test.cc
@@ -50,8 +50,10 @@ TEST(KeyParserTest, KeyCode) {
   EXPECT_EQ('A', key_event.key_code());
 
   // "あ" (not half width)
+  const char32 kHiraganaA = 0x3042;
   key_event.Clear();
-  EXPECT_FALSE(KeyParser::ParseKey("\xE3\x81\x82", &key_event));
+  EXPECT_TRUE(KeyParser::ParseKey("\xE3\x81\x82", &key_event));
+  EXPECT_EQ(kHiraganaA, key_event.key_code());
 }
 
 TEST(KeyParserTest, ModifierKeys) {

diff --git a/src/data/version/mozc_version_template.bzl b/src/data/version/mozc_version_template.bzl
@@ -30,7 +30,7 @@
 
 MAJOR=2
 MINOR=18
-BUILD=2611
+BUILD=2612
 REVISION=102
 # This version represents the version of Mozc IME engine (converter, predictor,
 # etc.).  This version info is included both in the Mozc server and in the Mozc