Related: fdo#49208 use UText adaptor to underlying rtl::OUString

Don't convert to icu UnicodeString, retain text as shallow-copy of
original via rtl::OUString and use UText adaptor. Allows use of
equals to do super fast compare that new string is the same as the
old string

Change-Id: Ie9a3dc981b22a6866f3712c786331a1d6fcf153a

2 files changed, 27 insertions, 24 deletions

diff --git a/i18npool/inc/breakiterator_unicode.hxx b/i18npool/inc/breakiterator_unicode.hxx
index 57e0aadc7fc4..a9273f14b24e 100644
--- a/i18npool/inc/breakiterator_unicode.hxx
+++ b/i18npool/inc/breakiterator_unicode.hxx
@@ -84,11 +84,22 @@ protected:
     const sal_Char *cBreakIterator, *wordRule, *lineRule;
     Boundary result; // for word break iterator
 
-    struct BI_Data {
-        UnicodeString aICUText;
+    struct BI_Data
+    {
+        rtl::OUString aICUText;
+        UText *ut;
         icu::BreakIterator *aBreakIterator;
 
-        BI_Data() : aICUText(), aBreakIterator(NULL) {}
+        BI_Data()
+            : ut(NULL)
+            , aBreakIterator(NULL)
+        {
+        }
+        ~BI_Data()
+        {
+            utext_close(ut);
+        }
+
     } character, word, sentence, line, *icuBI;
 
     com::sun::star::lang::Locale aLocale;
diff --git a/i18npool/source/breakiterator/breakiterator_unicode.cxx b/i18npool/source/breakiterator/breakiterator_unicode.cxx
index 4005780a27d3..34e6918f5182 100644
--- a/i18npool/source/breakiterator/breakiterator_unicode.cxx
+++ b/i18npool/source/breakiterator/breakiterator_unicode.cxx
@@ -95,24 +95,6 @@ class OOoRuleBasedBreakIterator : public RuleBasedBreakIterator {
 
 };
 
-namespace
-{
-    bool isEqual(const UnicodeString &rOne, const rtl::OUString &rOther)
-    {
-        sal_Int32 nLength = rOne.length();
-        if (nLength != rOther.getLength())
-            return false;
-
-        //fdo#49208 operator== is implemented by compareTo etc in icu which is
-        //horrifically slow when all you want to know is that they're the same
-        //or not
-        const UChar *pOne = rOne.getBuffer();
-        // UChar != sal_Unicode in MinGW
-        const UChar *pOther = reinterpret_cast<const UChar *>(rOther.getStr());
-        return memcmp(pOne, pOther, nLength * sizeof(UChar)) == 0;
-    }
-}
-
 // loading ICU breakiterator on demand.
 void SAL_CALL BreakIterator_Unicode::loadICUBreakIterator(const com::sun::star::lang::Locale& rLocale,
         sal_Int16 rBreakType, sal_Int16 rWordType, const sal_Char *rule, const OUString& rText) throw(uno::RuntimeException)
@@ -218,12 +200,22 @@ void SAL_CALL BreakIterator_Unicode::loadICUBreakIterator(const com::sun::star::
         }
     }
 
-    if (newBreak || !isEqual(icuBI->aICUText, rText))
+    if (newBreak || !icuBI->aICUText.equals(rText))
     {
         // UChar != sal_Unicode in MinGW
         const UChar *pText = reinterpret_cast<const UChar *>(rText.getStr());
-        icuBI->aICUText=UnicodeString(pText, rText.getLength());
-        icuBI->aBreakIterator->setText(icuBI->aICUText);
+
+        icuBI->ut = utext_openUChars(icuBI->ut, pText, rText.getLength(), &status);
+
+        if (!U_SUCCESS(status))
+            throw ERROR;
+
+        icuBI->aBreakIterator->setText(icuBI->ut, status);
+
+        if (!U_SUCCESS(status))
+            throw ERROR;
+
+        icuBI->aICUText = rText;
     }
 }