summaryrefslogtreecommitdiff
path: root/i18npool
diff options
context:
space:
mode:
Diffstat (limited to 'i18npool')
-rw-r--r--i18npool/qa/cppunit/test_breakiterator.cxx51
-rw-r--r--i18npool/source/breakiterator/breakiteratorImpl.cxx200
2 files changed, 191 insertions, 60 deletions
diff --git a/i18npool/qa/cppunit/test_breakiterator.cxx b/i18npool/qa/cppunit/test_breakiterator.cxx
index 0f42fe7b2164..7b4df321879e 100644
--- a/i18npool/qa/cppunit/test_breakiterator.cxx
+++ b/i18npool/qa/cppunit/test_breakiterator.cxx
@@ -57,11 +57,13 @@ public:
void testLineBreaking();
void testGraphemeIteration();
void testWeak();
+ void testAsian();
CPPUNIT_TEST_SUITE(TestBreakIterator);
CPPUNIT_TEST(testLineBreaking);
CPPUNIT_TEST(testGraphemeIteration);
CPPUNIT_TEST(testWeak);
+ CPPUNIT_TEST(testAsian);
CPPUNIT_TEST_SUITE_END();
private:
@@ -104,7 +106,7 @@ void TestBreakIterator::testGraphemeIteration()
aLocale.Country = ::rtl::OUString(RTL_CONSTASCII_USTRINGPARAM("IN"));
{
- sal_Unicode BA_HALANT_LA[] = { 0x09AC, 0x09CD, 0x09AF };
+ const sal_Unicode BA_HALANT_LA[] = { 0x09AC, 0x09CD, 0x09AF };
::rtl::OUString aTest1(BA_HALANT_LA, SAL_N_ELEMENTS(BA_HALANT_LA));
sal_Int32 nDone=0;
@@ -118,7 +120,7 @@ void TestBreakIterator::testGraphemeIteration()
}
{
- sal_Unicode HA_HALANT_NA_VOWELSIGNI[] = { 0x09B9, 0x09CD, 0x09A3, 0x09BF };
+ const sal_Unicode HA_HALANT_NA_VOWELSIGNI[] = { 0x09B9, 0x09CD, 0x09A3, 0x09BF };
::rtl::OUString aTest1(HA_HALANT_NA_VOWELSIGNI, SAL_N_ELEMENTS(HA_HALANT_NA_VOWELSIGNI));
sal_Int32 nDone=0;
@@ -132,7 +134,7 @@ void TestBreakIterator::testGraphemeIteration()
}
{
- sal_Unicode TA_HALANT_MA_HALANT_YA [] = { 0x09A4, 0x09CD, 0x09AE, 0x09CD, 0x09AF };
+ const sal_Unicode TA_HALANT_MA_HALANT_YA [] = { 0x09A4, 0x09CD, 0x09AE, 0x09CD, 0x09AF };
::rtl::OUString aTest1(TA_HALANT_MA_HALANT_YA, SAL_N_ELEMENTS(TA_HALANT_MA_HALANT_YA));
sal_Int32 nDone=0;
@@ -156,10 +158,12 @@ void TestBreakIterator::testWeak()
aLocale.Country = ::rtl::OUString(RTL_CONSTASCII_USTRINGPARAM("US"));
{
- sal_Unicode WEAKS[] =
+ const sal_Unicode WEAKS[] =
{
0x0001, 0x0002,
0x0020, 0x00A0,
+ 0x2150, 0x215F, //Number Forms, fractions
+ 0x2160, 0x2180, //Number Forms, roman numerals
0x2200, 0x22FF, //Mathematical Operators
0x27C0, 0x27EF, //Miscellaneous Mathematical Symbols-A
0x2980, 0x29FF, //Miscellaneous Mathematical Symbols-B
@@ -184,6 +188,45 @@ void TestBreakIterator::testWeak()
}
}
+//A test to ensure that certain ranges and codepoints that are categorized as
+//asian remain as asian, so that existing docs that depend on this don't silently
+//change font for those asian chars.
+//See https://bugs.freedesktop.org/show_bug.cgi?id=38095
+void TestBreakIterator::testAsian()
+{
+ lang::Locale aLocale;
+ aLocale.Language = ::rtl::OUString(RTL_CONSTASCII_USTRINGPARAM("en"));
+ aLocale.Country = ::rtl::OUString(RTL_CONSTASCII_USTRINGPARAM("US"));
+
+ {
+ const sal_Unicode ASIANS[] =
+ {
+ //some typical CJK chars
+ 0x4E00, 0x62FF,
+ //The full HalfWidth and FullWidth block has historically been
+ //designated as taking the CJK font :-(
+ //HalfWidth and FullWidth forms of ASCII 0-9, categorized under
+ //UAX24 as "Common" i.e. by that logic WEAK
+ 0xFF10, 0xFF19,
+ //HalfWidth and FullWidth forms of ASCII A-z, categorized under
+ //UAX25 as "Latin", i.e. by that logic LATIN
+ 0xFF21, 0xFF5A
+ };
+ ::rtl::OUString aAsians(ASIANS, SAL_N_ELEMENTS(ASIANS));
+
+ for (sal_Int32 i = 0; i < aAsians.getLength(); ++i)
+ {
+ sal_Int16 nScript = m_xBreak->getScriptType(aAsians, i);
+ rtl::OStringBuffer aMsg;
+ aMsg.append(RTL_CONSTASCII_STRINGPARAM("Char 0x"));
+ aMsg.append(static_cast<sal_Int32>(aAsians.getStr()[i]), 16);
+ aMsg.append(RTL_CONSTASCII_STRINGPARAM(" should have been asian"));
+ CPPUNIT_ASSERT_MESSAGE(aMsg.getStr(),
+ nScript == i18n::ScriptType::ASIAN);
+ }
+ }
+}
+
TestBreakIterator::TestBreakIterator()
{
m_xContext = cppu::defaultBootstrap_InitialComponentContext();
diff --git a/i18npool/source/breakiterator/breakiteratorImpl.cxx b/i18npool/source/breakiterator/breakiteratorImpl.cxx
index 3cc974870c3d..9c6d09d7d6ec 100644
--- a/i18npool/source/breakiterator/breakiteratorImpl.cxx
+++ b/i18npool/source/breakiterator/breakiteratorImpl.cxx
@@ -443,67 +443,155 @@ sal_Int16 SAL_CALL BreakIteratorImpl::getWordType( const OUString& /*Text*/,
return 0;
}
-static sal_Int16 scriptTypes[] = {
- ScriptType::WEAK, ScriptType::WEAK, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX,
- ScriptType::ASIAN, ScriptType::LATIN, ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::LATIN, ScriptType::LATIN,
-// 15
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::ASIAN, ScriptType::ASIAN, ScriptType::COMPLEX,
- ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN,
-// 30
- ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::LATIN, ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
-// 45
- ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN,
- ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
-// 60
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::ASIAN, ScriptType::ASIAN,
-// 75
- ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
-// 90
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::WEAK, ScriptType::WEAK, ScriptType::COMPLEX,
-// 105
- ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::ASIAN,
-// 120
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::WEAK, ScriptType::WEAK,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
-// 135
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
- ScriptType::COMPLEX,
- ScriptType::WEAK};
-
-#define scriptListCount SAL_N_ELEMENTS(scriptTypes)
+namespace
+{
+ //See unicode/uscript.h
+ static sal_Int16 scriptTypes[] =
+ {
+ ScriptType::WEAK, ScriptType::WEAK, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX,
+ ScriptType::ASIAN, ScriptType::LATIN, ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::LATIN, ScriptType::LATIN,
+ // 15
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::ASIAN, ScriptType::ASIAN, ScriptType::COMPLEX,
+ ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN,
+ // 30
+ ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::LATIN, ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ // 45
+ ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN,
+ ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ // 60
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::ASIAN, ScriptType::ASIAN,
+ // 75
+ ScriptType::COMPLEX, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::LATIN, ScriptType::LATIN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ // 90
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::WEAK, ScriptType::WEAK, ScriptType::COMPLEX,
+ // 105
+ ScriptType::ASIAN, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::ASIAN,
+ // 120
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::WEAK, ScriptType::WEAK,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ // 135
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX, ScriptType::COMPLEX,
+ ScriptType::COMPLEX,
+ ScriptType::WEAK
+ };
+
+# define scriptTypesCount SAL_N_ELEMENTS(scriptTypes)
+
+ sal_Int16 getScriptClassByUAX24Script(sal_uInt32 currentChar)
+ {
+ sal_Int16 nRet;
+ int32_t script = u_getIntPropertyValue(currentChar, UCHAR_SCRIPT);
+ if (script < 0)
+ nRet = ScriptType::WEAK;
+ else if (static_cast<size_t>(script) >= SAL_N_ELEMENTS(scriptTypes))
+ nRet = ScriptType::COMPLEX; // anything new is going to be pretty wild
+ else
+ nRet = scriptTypes[script];
+ return nRet;
+ }
+
+ struct UBlock2Script
+ {
+ UBlockCode from;
+ UBlockCode to;
+ sal_Int16 script;
+ };
+
+ static UBlock2Script scriptList[] =
+ {
+ {UBLOCK_NO_BLOCK, UBLOCK_NO_BLOCK, ScriptType::WEAK},
+ {UBLOCK_BASIC_LATIN, UBLOCK_ARMENIAN, ScriptType::LATIN},
+ {UBLOCK_HEBREW, UBLOCK_MYANMAR, ScriptType::COMPLEX},
+ {UBLOCK_GEORGIAN, UBLOCK_GEORGIAN, ScriptType::LATIN},
+ {UBLOCK_HANGUL_JAMO, UBLOCK_HANGUL_JAMO, ScriptType::ASIAN},
+ {UBLOCK_ETHIOPIC, UBLOCK_ETHIOPIC, ScriptType::COMPLEX},
+ {UBLOCK_CHEROKEE, UBLOCK_RUNIC, ScriptType::LATIN},
+ {UBLOCK_KHMER, UBLOCK_MONGOLIAN, ScriptType::COMPLEX},
+ {UBLOCK_LATIN_EXTENDED_ADDITIONAL, UBLOCK_GREEK_EXTENDED, ScriptType::LATIN},
+ {UBLOCK_NUMBER_FORMS, UBLOCK_NUMBER_FORMS, ScriptType::WEAK},
+ {UBLOCK_CJK_RADICALS_SUPPLEMENT, UBLOCK_HANGUL_SYLLABLES, ScriptType::ASIAN},
+ {UBLOCK_CJK_COMPATIBILITY_IDEOGRAPHS, UBLOCK_CJK_COMPATIBILITY_IDEOGRAPHS, ScriptType::ASIAN},
+ {UBLOCK_ARABIC_PRESENTATION_FORMS_A, UBLOCK_ARABIC_PRESENTATION_FORMS_A, ScriptType::COMPLEX},
+ {UBLOCK_CJK_COMPATIBILITY_FORMS, UBLOCK_CJK_COMPATIBILITY_FORMS, ScriptType::ASIAN},
+ {UBLOCK_ARABIC_PRESENTATION_FORMS_B, UBLOCK_ARABIC_PRESENTATION_FORMS_B, ScriptType::COMPLEX},
+ {UBLOCK_HALFWIDTH_AND_FULLWIDTH_FORMS, UBLOCK_HALFWIDTH_AND_FULLWIDTH_FORMS, ScriptType::ASIAN},
+ {UBLOCK_CJK_UNIFIED_IDEOGRAPHS_EXTENSION_B, UBLOCK_CJK_COMPATIBILITY_IDEOGRAPHS_SUPPLEMENT, ScriptType::ASIAN},
+ {UBLOCK_CJK_STROKES, UBLOCK_CJK_STROKES, ScriptType::ASIAN},
+ {UBLOCK_LATIN_EXTENDED_C, UBLOCK_LATIN_EXTENDED_D, ScriptType::LATIN}
+ };
+
+ #define scriptListCount SAL_N_ELEMENTS(scriptTypes)
+
+ //always sets rScriptType
+ //
+ //returns true for characters historically explicitly assigned to
+ //latin/weak/asian
+ //
+ //returns false for characters that historically implicitly assigned to
+ //weak as unknown
+ bool getCompatibilityScriptClassByBlock(sal_uInt32 currentChar, sal_Int16 &rScriptType)
+ {
+ bool bKnown = true;
+ //handle specific characters always as weak:
+ // 0x01 - this breaks a word
+ // 0x02 - this can be inside a word
+ // 0x20 & 0xA0 - Bug 102975, declare western space and non-break space as WEAK char.
+ if( 0x01 == currentChar || 0x02 == currentChar || 0x20 == currentChar || 0xA0 == currentChar)
+ rScriptType = ScriptType::WEAK;
+ // workaround for Coptic
+ else if ( 0x2C80 <= currentChar && 0x2CE3 >= currentChar)
+ rScriptType = ScriptType::LATIN;
+ else
+ {
+ UBlockCode block=ublock_getCode(currentChar);
+ size_t i = 0;
+ while (i < scriptListCount)
+ {
+ if (block <= scriptList[i].to)
+ break;
+ ++i;
+ }
+ if (i < scriptListCount && block >= scriptList[i].from)
+ rScriptType = scriptList[i].script;
+ else
+ {
+ rScriptType = ScriptType::WEAK;
+ bKnown = false;
+ }
+ }
+ return bKnown;
+ }
+}
sal_Int16 BreakIteratorImpl::getScriptClass(sal_uInt32 currentChar)
{
- static sal_uInt32 lastChar = 0;
- static sal_Int16 nRet = 0;
+ static sal_uInt32 lastChar = 0;
+ static sal_Int16 nRet = 0;
- if (currentChar != lastChar) {
- lastChar = currentChar;
+ if (currentChar != lastChar)
+ {
+ lastChar = currentChar;
- int32_t script = u_getIntPropertyValue(currentChar, UCHAR_SCRIPT);
- if (script < 0)
- nRet = ScriptType::WEAK;
- else if (static_cast<size_t>(script) >= SAL_N_ELEMENTS(scriptTypes))
- nRet = ScriptType::COMPLEX; // anything new is going to be pretty wild
- else
- nRet = scriptTypes[script];
- }
- return nRet;
+ if (!getCompatibilityScriptClassByBlock(currentChar, nRet))
+ nRet = getScriptClassByUAX24Script(currentChar);
+ }
+
+ return nRet;
}
static inline sal_Bool operator == (const Locale& l1, const Locale& l2) {