ICU-5456 modern Greek toUpperCase(), remove most diacritics

X-SVN-Rev: 39228

diff --git a/icu4c/source/common/ucase.cpp b/icu4c/source/common/ucase.cpp
index c3631a6dd646f7e2c771af44f646178085d4ddc6..fe4335ea459773e076b6e44df65dee879db141fd 100644 (file)
--- a/icu4c/source/common/ucase.cpp
+++ b/icu4c/source/common/ucase.cpp
@@ -621,6 +621,18 @@ ucase_getCaseLocale(const char *locale, int32_t *locCache) {
                 result=UCASE_LOC_LITHUANIAN;
             }
         }
+    } else if(is_e(c)) {
+        /* el or ell? */
+        c=*locale++;
+        if(is_l(c)) {
+            c=*locale++;
+            if(is_l(c)) {
+                c=*locale;
+            }
+            if(is_sep(c)) {
+                result=UCASE_LOC_GREEK;
+            }
+        }
     } else if(is_n(c)) {
         /* nl or nld? */
         c=*locale++;

diff --git a/icu4c/source/common/ucase.h b/icu4c/source/common/ucase.h
index 85b5c60bf89696d67d4e9bd88c9d5cc22e81bd9c..9bcb4c7a49988b28a3af1db9ed90eb56fcf4f1a5 100644 (file)
--- a/icu4c/source/common/ucase.h
+++ b/icu4c/source/common/ucase.h
@@ -64,6 +64,7 @@ enum {
     UCASE_LOC_ROOT,
     UCASE_LOC_TURKISH,
     UCASE_LOC_LITHUANIAN,
+    UCASE_LOC_GREEK,
     UCASE_LOC_DUTCH
 };
 
@@ -158,7 +159,7 @@ U_NAMESPACE_END
 U_CAPI int32_t U_EXPORT2
 ucase_getType(const UCaseProps *csp, UChar32 c);
 
-/** @return same as ucase_getType(), or <0 if c is case-ignorable */
+/** @return same as ucase_getType() and set bit 2 if c is case-ignorable */
 U_CAPI int32_t U_EXPORT2
 ucase_getTypeOrIgnorable(const UCaseProps *csp, UChar32 c);
 

diff --git a/icu4c/source/common/ucasemap.cpp b/icu4c/source/common/ucasemap.cpp
index dbb9f62e6b5944acfeced47dea6b4c7492c3a8bc..55195663df88744de157e241d53f61e9a1b10a7f 100644 (file)
--- a/icu4c/source/common/ucasemap.cpp
+++ b/icu4c/source/common/ucasemap.cpp
@@ -176,6 +176,15 @@ appendResult(uint8_t *dest, int32_t destIndex, int32_t destCapacity,
     return destIndex;
 }
 
+static inline int32_t
+appendUChar(uint8_t *dest, int32_t destIndex, int32_t destCapacity, UChar c) {
+    int32_t limit=destIndex+U8_LENGTH(c);
+    if(limit<destCapacity) {
+        U8_APPEND_UNSAFE(dest, destIndex, c);
+    }
+    return limit;
+}
+
 static UChar32 U_CALLCONV
 utf8_caseContextIterator(void *context, int8_t dir) {
     UCaseContext *csc=(UCaseContext *)context;
@@ -388,6 +397,141 @@ ucasemap_internalUTF8ToTitle(const UCaseMap *csm,
 
 #endif
 
+U_NAMESPACE_BEGIN
+namespace GreekUpper {
+
+UBool isFollowedByCasedLetter(const UCaseProps *csp, const uint8_t *s, int32_t i, int32_t length) {
+    while (i < length) {
+        UChar32 c;
+        U8_NEXT(s, i, length, c);
+        int32_t type = ucase_getTypeOrIgnorable(csp, c);
+        if ((type & UCASE_IGNORABLE) != 0) {
+            // Case-ignorable, continue with the loop.
+        } else if (type != UCASE_NONE) {
+            return TRUE;  // Followed by cased letter.
+        } else {
+            return FALSE;  // Uncased and not case-ignorable.
+        }
+    }
+    return FALSE;  // Not followed by cased letter.
+}
+
+// Keep this consistent with the UTF-16 version in ustrcase.cpp and the Java version in CaseMap.java.
+int32_t toUpper(const UCaseMap *csm,
+                uint8_t *dest, int32_t destCapacity,
+                const uint8_t *src, int32_t srcLength,
+                UErrorCode *pErrorCode) {
+    int32_t locCache = UCASE_LOC_GREEK;
+    int32_t destIndex=0;
+    uint32_t state = 0;
+    for (int32_t i = 0; i < srcLength;) {
+        int32_t nextIndex = i;
+        UChar32 c;
+        U8_NEXT(src, nextIndex, srcLength, c);
+        uint32_t nextState = 0;
+        int32_t type = ucase_getTypeOrIgnorable(csm->csp, c);
+        if ((type & UCASE_IGNORABLE) != 0) {
+            // c is case-ignorable
+            nextState |= (state & AFTER_CASED);
+        } else if (type != UCASE_NONE) {
+            // c is cased
+            nextState |= AFTER_CASED;
+        }
+        uint32_t data = getLetterData(c);
+        if (data > 0) {
+            uint32_t upper = data & UPPER_MASK;
+            // Add a dialytika to this iota or ypsilon vowel
+            // if we removed a tonos from the previous vowel,
+            // and that previous vowel did not also have (or gain) a dialytika.
+            // Adding one only to the final vowel in a longer sequence
+            // (which does not occur in normal writing) would require lookahead.
+            // Set the same flag as for preserving an existing dialytika.
+            if ((data & HAS_VOWEL) != 0 && (state & AFTER_VOWEL_WITH_ACCENT) != 0 &&
+                    (upper == 0x399 || upper == 0x3A5)) {
+                data |= HAS_DIALYTIKA;
+            }
+            int32_t numYpogegrammeni = 0;  // Map each one to a trailing, spacing, capital iota.
+            if ((data & HAS_YPOGEGRAMMENI) != 0) {
+                numYpogegrammeni = 1;
+            }
+            // Skip combining diacritics after this Greek letter.
+            int32_t nextNextIndex = nextIndex;
+            while (nextIndex < srcLength) {
+                UChar32 c2;
+                U8_NEXT(src, nextNextIndex, srcLength, c2);
+                uint32_t diacriticData = getDiacriticData(c2);
+                if (diacriticData != 0) {
+                    data |= diacriticData;
+                    if ((diacriticData & HAS_YPOGEGRAMMENI) != 0) {
+                        ++numYpogegrammeni;
+                    }
+                    nextIndex = nextNextIndex;
+                } else {
+                    break;  // not a Greek diacritic
+                }
+            }
+            if ((data & HAS_VOWEL_AND_ACCENT_AND_DIALYTIKA) == HAS_VOWEL_AND_ACCENT) {
+                nextState |= AFTER_VOWEL_WITH_ACCENT;
+            }
+            // Map according to Greek rules.
+            UBool addTonos = FALSE;
+            if (upper == 0x397 &&
+                    (data & HAS_ACCENT) != 0 &&
+                    numYpogegrammeni == 0 &&
+                    (state & AFTER_CASED) == 0 &&
+                    !isFollowedByCasedLetter(csm->csp, src, nextIndex, srcLength)) {
+                // Keep disjunctive "or" with (only) a tonos.
+                // We use the same "word boundary" conditions as for the Final_Sigma test.
+                if (i == nextIndex) {
+                    upper = 0x389;  // Preserve the precomposed form.
+                } else {
+                    addTonos = TRUE;
+                }
+            } else if ((data & HAS_DIALYTIKA) != 0) {
+                // Preserve a vowel with dialytika in precomposed form if it exists.
+                if (upper == 0x399) {
+                    upper = 0x3AA;
+                    data &= ~HAS_EITHER_DIALYTIKA;
+                } else if (upper == 0x3A5) {
+                    upper = 0x3AB;
+                    data &= ~HAS_EITHER_DIALYTIKA;
+                }
+            }
+            destIndex=appendUChar(dest, destIndex, destCapacity, (UChar)upper);
+            if ((data & HAS_EITHER_DIALYTIKA) != 0) {
+                destIndex=appendUChar(dest, destIndex, destCapacity, 0x308);  // restore or add a dialytika
+            }
+            if (addTonos) {
+                destIndex=appendUChar(dest, destIndex, destCapacity, 0x301);
+            }
+            while (numYpogegrammeni > 0) {
+                destIndex=appendUChar(dest, destIndex, destCapacity, 0x399);
+                --numYpogegrammeni;
+            }
+        } else {
+            const UChar *s;
+            UChar32 c2 = 0;
+            c=ucase_toFullUpper(csm->csp, c, NULL, NULL, &s, csm->locale, &locCache);
+            if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0x7f : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0x7f)) {
+                /* fast path version of appendResult() for ASCII results */
+                dest[destIndex++]=(uint8_t)c2;
+            } else {
+                destIndex=appendResult(dest, destIndex, destCapacity, c, s);
+            }
+        }
+        i = nextIndex;
+        state = nextState;
+    }
+
+    if(destIndex>destCapacity) {
+        *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
+    }
+    return destIndex;
+}
+
+}  // namespace GreekUpper
+U_NAMESPACE_END
+
 static int32_t U_CALLCONV
 ucasemap_internalUTF8ToLower(const UCaseMap *csm,
                              uint8_t *dest, int32_t destCapacity,
@@ -408,6 +552,10 @@ ucasemap_internalUTF8ToUpper(const UCaseMap *csm,
                              uint8_t *dest, int32_t destCapacity,
                              const uint8_t *src, int32_t srcLength,
                              UErrorCode *pErrorCode) {
+    int32_t locCache = csm->locCache;
+    if (ucase_getCaseLocale(csm->locale, &locCache) == UCASE_LOC_GREEK) {
+        return GreekUpper::toUpper(csm, dest, destCapacity, src, srcLength, pErrorCode);
+    }
     UCaseContext csc=UCASECONTEXT_INITIALIZER;
     csc.p=(void *)src;
     csc.limit=srcLength;

diff --git a/icu4c/source/common/ustr_imp.h b/icu4c/source/common/ustr_imp.h
index c30169592a01a2dab8ffc407eb76489560db9e85..34a69363a765d629236ef5e0eb04d7777e6e0d00 100644 (file)
--- a/icu4c/source/common/ustr_imp.h
+++ b/icu4c/source/common/ustr_imp.h
@@ -220,6 +220,44 @@ ucasemap_mapUTF8(const UCaseMap *csm,
                  UTF8CaseMapper *stringCaseMapper,
                  UErrorCode *pErrorCode);
 
+#ifdef __cplusplus
+
+U_NAMESPACE_BEGIN
+namespace GreekUpper {
+
+// Data bits.
+static const uint32_t UPPER_MASK = 0x3ff;
+static const uint32_t HAS_VOWEL = 0x1000;
+static const uint32_t HAS_YPOGEGRAMMENI = 0x2000;
+static const uint32_t HAS_ACCENT = 0x4000;
+static const uint32_t HAS_DIALYTIKA = 0x8000;
+// Further bits during data building and processing, not stored in the data map.
+static const uint32_t HAS_COMBINING_DIALYTIKA = 0x10000;
+static const uint32_t HAS_OTHER_GREEK_DIACRITIC = 0x20000;
+
+static const uint32_t HAS_VOWEL_AND_ACCENT = HAS_VOWEL | HAS_ACCENT;
+static const uint32_t HAS_VOWEL_AND_ACCENT_AND_DIALYTIKA =
+        HAS_VOWEL_AND_ACCENT | HAS_DIALYTIKA;
+static const uint32_t HAS_EITHER_DIALYTIKA = HAS_DIALYTIKA | HAS_COMBINING_DIALYTIKA;
+
+// State bits.
+static const uint32_t AFTER_CASED = 1;
+static const uint32_t AFTER_VOWEL_WITH_ACCENT = 2;
+
+uint32_t getLetterData(UChar32 c);
+
+/**
+ * Returns a non-zero value for each of the Greek combining diacritics
+ * listed in The Unicode Standard, version 8, chapter 7.2 Greek,
+ * plus some perispomeni look-alikes.
+ */
+uint32_t getDiacriticData(UChar32 c);
+
+}  // namespace GreekUpper
+U_NAMESPACE_END
+
+#endif  // __cplusplus
+
 U_CAPI int32_t U_EXPORT2 
 ustr_hashUCharsN(const UChar *str, int32_t length);
 

diff --git a/icu4c/source/common/ustrcase.cpp b/icu4c/source/common/ustrcase.cpp
index a8410dfeb669ebfaec0f7e39a1479d9c299717d1..fd5d2d6b03b3f8a84f5d345059958182e72c7110 100644 (file)
--- a/icu4c/source/common/ustrcase.cpp
+++ b/icu4c/source/common/ustrcase.cpp
@@ -89,6 +89,14 @@ appendResult(UChar *dest, int32_t destIndex, int32_t destCapacity,
     return destIndex;
 }
 
+static inline int32_t
+appendUChar(UChar *dest, int32_t destIndex, int32_t destCapacity, UChar c) {
+    if(destIndex<destCapacity) {
+        dest[destIndex]=c;
+    }
+    return destIndex+1;
+}
+
 static UChar32 U_CALLCONV
 utf16_caseContextIterator(void *context, int8_t dir) {
     UCaseContext *csc=(UCaseContext *)context;
@@ -295,6 +303,597 @@ ustrcase_internalToTitle(const UCaseMap *csm,
 
 #endif  // !UCONFIG_NO_BREAK_ITERATION
 
+U_NAMESPACE_BEGIN
+namespace GreekUpper {
+
+// Data generated by prototype code, see
+// http://site.icu-project.org/design/case/greek-upper
+// TODO: Move this data into ucase.icu.
+static const uint16_t data0370[] = {
+    // U+0370..03FF
+    0x0370,
+    0x0370,
+    0x0372,
+    0x0372,
+    0,
+    0,
+    0x0376,
+    0x0376,
+    0,
+    0,
+    0x037A,
+    0x03FD,
+    0x03FE,
+    0x03FF,
+    0,
+    0x037F,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0x0391 | HAS_VOWEL,
+    0x0392,
+    0x0393,
+    0x0394,
+    0x0395 | HAS_VOWEL,
+    0x0396,
+    0x0397 | HAS_VOWEL,
+    0x0398,
+    0x0399 | HAS_VOWEL,
+    0x039A,
+    0x039B,
+    0x039C,
+    0x039D,
+    0x039E,
+    0x039F | HAS_VOWEL,
+    0x03A0,
+    0x03A1,
+    0,
+    0x03A3,
+    0x03A4,
+    0x03A5 | HAS_VOWEL,
+    0x03A6,
+    0x03A7,
+    0x03A8,
+    0x03A9 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL | HAS_DIALYTIKA,
+    0x03A5 | HAS_VOWEL | HAS_DIALYTIKA,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0x0391 | HAS_VOWEL,
+    0x0392,
+    0x0393,
+    0x0394,
+    0x0395 | HAS_VOWEL,
+    0x0396,
+    0x0397 | HAS_VOWEL,
+    0x0398,
+    0x0399 | HAS_VOWEL,
+    0x039A,
+    0x039B,
+    0x039C,
+    0x039D,
+    0x039E,
+    0x039F | HAS_VOWEL,
+    0x03A0,
+    0x03A1,
+    0x03A3,
+    0x03A3,
+    0x03A4,
+    0x03A5 | HAS_VOWEL,
+    0x03A6,
+    0x03A7,
+    0x03A8,
+    0x03A9 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL | HAS_DIALYTIKA,
+    0x03A5 | HAS_VOWEL | HAS_DIALYTIKA,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03CF,
+    0x0392,
+    0x0398,
+    0x03D2,
+    0x03D2 | HAS_ACCENT,
+    0x03D2 | HAS_DIALYTIKA,
+    0x03A6,
+    0x03A0,
+    0x03CF,
+    0x03D8,
+    0x03D8,
+    0x03DA,
+    0x03DA,
+    0x03DC,
+    0x03DC,
+    0x03DE,
+    0x03DE,
+    0x03E0,
+    0x03E0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0x039A,
+    0x03A1,
+    0x03F9,
+    0x037F,
+    0x03F4,
+    0x0395 | HAS_VOWEL,
+    0,
+    0x03F7,
+    0x03F7,
+    0x03F9,
+    0x03FA,
+    0x03FA,
+    0x03FC,
+    0x03FD,
+    0x03FE,
+    0x03FF,
+};
+
+static const uint16_t data1F00[] = {
+    // U+1F00..1FFF
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL,
+    0x0395 | HAS_VOWEL,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0,
+    0x0395 | HAS_VOWEL,
+    0x0395 | HAS_VOWEL,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0,
+    0x0397 | HAS_VOWEL,
+    0x0397 | HAS_VOWEL,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL,
+    0x0397 | HAS_VOWEL,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL,
+    0x039F | HAS_VOWEL,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0,
+    0x039F | HAS_VOWEL,
+    0x039F | HAS_VOWEL,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0,
+    0x03A5 | HAS_VOWEL,
+    0x03A5 | HAS_VOWEL,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0x03A5 | HAS_VOWEL,
+    0,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL,
+    0x03A9 | HAS_VOWEL,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL,
+    0x03A9 | HAS_VOWEL,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_ACCENT,
+    0x0391 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0,
+    0x0399 | HAS_VOWEL,
+    0,
+    0,
+    0,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0395 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_ACCENT,
+    0x0397 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0,
+    0,
+    0,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0,
+    0,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0x0399 | HAS_VOWEL | HAS_ACCENT,
+    0,
+    0,
+    0,
+    0,
+    0x03A5 | HAS_VOWEL,
+    0x03A5 | HAS_VOWEL,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0x03A1,
+    0x03A1,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT | HAS_DIALYTIKA,
+    0x03A5 | HAS_VOWEL,
+    0x03A5 | HAS_VOWEL,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A5 | HAS_VOWEL | HAS_ACCENT,
+    0x03A1,
+    0,
+    0,
+    0,
+    0,
+    0,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x039F | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_ACCENT,
+    0x03A9 | HAS_VOWEL | HAS_YPOGEGRAMMENI,
+    0,
+    0,
+    0,
+};
+
+// U+2126 Ohm sign
+static const uint16_t data2126 = 0x03A9 | HAS_VOWEL;
+
+uint32_t getLetterData(UChar32 c) {
+    if (c < 0x370 || 0x2126 < c || (0x3ff < c && c < 0x1f00)) {
+        return 0;
+    } else if (c <= 0x3ff) {
+        return data0370[c - 0x370];
+    } else if (c <= 0x1fff) {
+        return data1F00[c - 0x1f00];
+    } else if (c == 0x2126) {
+        return data2126;
+    } else {
+        return 0;
+    }
+}
+
+uint32_t getDiacriticData(UChar32 c) {
+    switch (c) {
+    case 0x0300:  // varia
+    case 0x0301:  // tonos = oxia
+    case 0x0342:  // perispomeni
+    case 0x0302:  // circumflex can look like perispomeni
+    case 0x0303:  // tilde can look like perispomeni
+    case 0x0311:  // inverted breve can look like perispomeni
+        return HAS_ACCENT;
+    case 0x0308:  // dialytika = diaeresis
+        return HAS_COMBINING_DIALYTIKA;
+    case 0x0344:  // dialytika tonos
+        return HAS_COMBINING_DIALYTIKA | HAS_ACCENT;
+    case 0x0345:  // ypogegrammeni = iota subscript
+        return HAS_YPOGEGRAMMENI;
+    case 0x0304:  // macron
+    case 0x0306:  // breve
+    case 0x0313:  // comma above
+    case 0x0314:  // reversed comma above
+    case 0x0343:  // koronis
+        return HAS_OTHER_GREEK_DIACRITIC;
+    default:
+        return 0;
+    }
+}
+
+UBool isFollowedByCasedLetter(const UCaseProps *csp, const UChar *s, int32_t i, int32_t length) {
+    while (i < length) {
+        UChar32 c;
+        U16_NEXT(s, i, length, c);
+        int32_t type = ucase_getTypeOrIgnorable(csp, c);
+        if ((type & UCASE_IGNORABLE) != 0) {
+            // Case-ignorable, continue with the loop.
+        } else if (type != UCASE_NONE) {
+            return TRUE;  // Followed by cased letter.
+        } else {
+            return FALSE;  // Uncased and not case-ignorable.
+        }
+    }
+    return FALSE;  // Not followed by cased letter.
+}
+
+/**
+ * Greek string uppercasing with a state machine.
+ * Probably simpler than a stateless function that has to figure out complex context-before
+ * for each character.
+ * TODO: Try to re-consolidate one way or another with the non-Greek function.
+ */
+int32_t toUpper(const UCaseMap *csm,
+                UChar *dest, int32_t destCapacity,
+                const UChar *src, int32_t srcLength,
+                UErrorCode *pErrorCode) {
+    int32_t locCache = UCASE_LOC_GREEK;
+    int32_t destIndex=0;
+    uint32_t state = 0;
+    for (int32_t i = 0; i < srcLength;) {
+        int32_t nextIndex = i;
+        UChar32 c;
+        U16_NEXT(src, nextIndex, srcLength, c);
+        uint32_t nextState = 0;
+        int32_t type = ucase_getTypeOrIgnorable(csm->csp, c);
+        if ((type & UCASE_IGNORABLE) != 0) {
+            // c is case-ignorable
+            nextState |= (state & AFTER_CASED);
+        } else if (type != UCASE_NONE) {
+            // c is cased
+            nextState |= AFTER_CASED;
+        }
+        uint32_t data = getLetterData(c);
+        if (data > 0) {
+            uint32_t upper = data & UPPER_MASK;
+            // Add a dialytika to this iota or ypsilon vowel
+            // if we removed a tonos from the previous vowel,
+            // and that previous vowel did not also have (or gain) a dialytika.
+            // Adding one only to the final vowel in a longer sequence
+            // (which does not occur in normal writing) would require lookahead.
+            // Set the same flag as for preserving an existing dialytika.
+            if ((data & HAS_VOWEL) != 0 && (state & AFTER_VOWEL_WITH_ACCENT) != 0 &&
+                    (upper == 0x399 || upper == 0x3A5)) {
+                data |= HAS_DIALYTIKA;
+            }
+            int32_t numYpogegrammeni = 0;  // Map each one to a trailing, spacing, capital iota.
+            if ((data & HAS_YPOGEGRAMMENI) != 0) {
+                numYpogegrammeni = 1;
+            }
+            // Skip combining diacritics after this Greek letter.
+            while (nextIndex < srcLength) {
+                uint32_t diacriticData = getDiacriticData(src[nextIndex]);
+                if (diacriticData != 0) {
+                    data |= diacriticData;
+                    if ((diacriticData & HAS_YPOGEGRAMMENI) != 0) {
+                        ++numYpogegrammeni;
+                    }
+                    ++nextIndex;
+                } else {
+                    break;  // not a Greek diacritic
+                }
+            }
+            if ((data & HAS_VOWEL_AND_ACCENT_AND_DIALYTIKA) == HAS_VOWEL_AND_ACCENT) {
+                nextState |= AFTER_VOWEL_WITH_ACCENT;
+            }
+            // Map according to Greek rules.
+            UBool addTonos = FALSE;
+            if (upper == 0x397 &&
+                    (data & HAS_ACCENT) != 0 &&
+                    numYpogegrammeni == 0 &&
+                    (state & AFTER_CASED) == 0 &&
+                    !isFollowedByCasedLetter(csm->csp, src, nextIndex, srcLength)) {
+                // Keep disjunctive "or" with (only) a tonos.
+                // We use the same "word boundary" conditions as for the Final_Sigma test.
+                if (i == nextIndex) {
+                    upper = 0x389;  // Preserve the precomposed form.
+                } else {
+                    addTonos = TRUE;
+                }
+            } else if ((data & HAS_DIALYTIKA) != 0) {
+                // Preserve a vowel with dialytika in precomposed form if it exists.
+                if (upper == 0x399) {
+                    upper = 0x3AA;
+                    data &= ~HAS_EITHER_DIALYTIKA;
+                } else if (upper == 0x3A5) {
+                    upper = 0x3AB;
+                    data &= ~HAS_EITHER_DIALYTIKA;
+                }
+            }
+            destIndex=appendUChar(dest, destIndex, destCapacity, (UChar)upper);
+            if ((data & HAS_EITHER_DIALYTIKA) != 0) {
+                destIndex=appendUChar(dest, destIndex, destCapacity, 0x308);  // restore or add a dialytika
+            }
+            if (addTonos) {
+                destIndex=appendUChar(dest, destIndex, destCapacity, 0x301);
+            }
+            while (numYpogegrammeni > 0) {
+                destIndex=appendUChar(dest, destIndex, destCapacity, 0x399);
+                --numYpogegrammeni;
+            }
+        } else {
+            const UChar *s;
+            UChar32 c2 = 0;
+            c=ucase_toFullUpper(csm->csp, c, NULL, NULL, &s, csm->locale, &locCache);
+            if((destIndex<destCapacity) && (c<0 ? (c2=~c)<=0xffff : UCASE_MAX_STRING_LENGTH<c && (c2=c)<=0xffff)) {
+                /* fast path version of appendResult() for BMP results */
+                dest[destIndex++]=(UChar)c2;
+            } else {
+                destIndex=appendResult(dest, destIndex, destCapacity, c, s);
+            }
+        }
+        i = nextIndex;
+        state = nextState;
+    }
+
+    if(destIndex>destCapacity) {
+        *pErrorCode=U_BUFFER_OVERFLOW_ERROR;
+    }
+    return destIndex;
+}
+
+}  // namespace GreekUpper
+U_NAMESPACE_END
+
 /* functions available in the common library (for unistr_case.cpp) */
 
 U_CFUNC int32_t U_CALLCONV
@@ -317,6 +916,10 @@ ustrcase_internalToUpper(const UCaseMap *csm,
                          UChar *dest, int32_t destCapacity,
                          const UChar *src, int32_t srcLength,
                          UErrorCode *pErrorCode) {
+    int32_t locCache = csm->locCache;
+    if (ucase_getCaseLocale(csm->locale, &locCache) == UCASE_LOC_GREEK) {
+        return GreekUpper::toUpper(csm, dest, destCapacity, src, srcLength, pErrorCode);
+    }
     UCaseContext csc=UCASECONTEXT_INITIALIZER;
     csc.p=(void *)src;
     csc.limit=srcLength;

diff --git a/icu4c/source/test/intltest/strcase.cpp b/icu4c/source/test/intltest/strcase.cpp
index 1d07df2b216d4c44313fa0214be913424385e030..b0e617b1dd78be30651f7fc988dedd9d29eee6ee 100644 (file)
--- a/icu4c/source/test/intltest/strcase.cpp
+++ b/icu4c/source/test/intltest/strcase.cpp
@@ -18,6 +18,7 @@
 *   Test file for string casing C++ API functions.
 */
 
+#include "unicode/std_string.h"
 #include "unicode/uchar.h"
 #include "unicode/ures.h"
 #include "unicode/uloc.h"
@@ -30,6 +31,8 @@
 #include "unicode/tstdtmod.h"
 #include "cmemory.h"
 
+StringCaseTest::StringCaseTest() : GREEK_LOCALE_("el") {}
+
 StringCaseTest::~StringCaseTest() {}
 
 void
@@ -43,6 +46,7 @@ StringCaseTest::runIndexedTest(int32_t index, UBool exec, const char *&name, cha
     TESTCASE_AUTO(TestCasing);
 #endif
     TESTCASE_AUTO(TestFullCaseFoldingIterator);
+    TESTCASE_AUTO(TestGreekUpper);
     TESTCASE_AUTO_END;
 }
 
@@ -573,3 +577,113 @@ StringCaseTest::TestFullCaseFoldingIterator() {
         errln("error: FullCaseFoldingIterator yielded only %d (cp, full) pairs", (int)count);
     }
 }
+
+void
+StringCaseTest::assertGreekUpper(const char *s, const char *expected) {
+    UnicodeString s16 = UnicodeString(s).unescape();
+    UnicodeString expected16 = UnicodeString(expected).unescape();
+    UnicodeString msg = UnicodeString("UnicodeString::toUpper/Greek(\"") + s16 + "\")";
+    UnicodeString result16(s16);
+    result16.toUpper(GREEK_LOCALE_);
+    assertEquals(msg, expected16, result16);
+
+#if U_HAVE_STD_STRING
+    UErrorCode errorCode = U_ZERO_ERROR;
+    LocalUCaseMapPointer csm(ucasemap_open("el", 0, &errorCode));
+    assertSuccess("ucasemap_open", errorCode);
+    std::string s8;
+    s16.toUTF8String(s8);
+    msg = UnicodeString("ucasemap_utf8ToUpper/Greek(\"") + s16 + "\")";
+    char dest[1000];
+    int32_t length = ucasemap_utf8ToUpper(csm.getAlias(), dest, UPRV_LENGTHOF(dest),
+                                          s8.data(), s8.length(), &errorCode);
+    assertSuccess("ucasemap_utf8ToUpper", errorCode);
+    StringPiece result8(dest, length);
+    UnicodeString result16From8 = UnicodeString::fromUTF8(result8);
+    assertEquals(msg, expected16, result16From8);
+#endif
+}
+
+void
+StringCaseTest::TestGreekUpper() {
+    // See UCharacterCaseTest.java for human-readable strings.
+
+    // http://bugs.icu-project.org/trac/ticket/5456
+    assertGreekUpper("\\u03AC\\u03B4\\u03B9\\u03BA\\u03BF\\u03C2, "
+                     "\\u03BA\\u03B5\\u03AF\\u03BC\\u03B5\\u03BD\\u03BF, "
+                     "\\u03AF\\u03C1\\u03B9\\u03B4\\u03B1",
+                     "\\u0391\\u0394\\u0399\\u039A\\u039F\\u03A3, "
+                     "\\u039A\\u0395\\u0399\\u039C\\u0395\\u039D\\u039F, "
+                     "\\u0399\\u03A1\\u0399\\u0394\\u0391");
+    // https://bugzilla.mozilla.org/show_bug.cgi?id=307039
+    // https://bug307039.bmoattachments.org/attachment.cgi?id=194893
+    assertGreekUpper("\\u03A0\\u03B1\\u03C4\\u03AC\\u03C4\\u03B1",
+                     "\\u03A0\\u0391\\u03A4\\u0391\\u03A4\\u0391");
+    assertGreekUpper("\\u0391\\u03AD\\u03C1\\u03B1\\u03C2, "
+                     "\\u039C\\u03C5\\u03C3\\u03C4\\u03AE\\u03C1\\u03B9\\u03BF, "
+                     "\\u03A9\\u03C1\\u03B1\\u03AF\\u03BF",
+                     "\\u0391\\u0395\\u03A1\\u0391\\u03A3, "
+                     "\\u039C\\u03A5\\u03A3\\u03A4\\u0397\\u03A1\\u0399\\u039F, "
+                     "\\u03A9\\u03A1\\u0391\\u0399\\u039F");
+    assertGreekUpper("\\u039C\\u03B1\\u0390\\u03BF\\u03C5, \\u03A0\\u03CC\\u03C1\\u03BF\\u03C2, "
+                     "\\u03A1\\u03CD\\u03B8\\u03BC\\u03B9\\u03C3\\u03B7",
+                     "\\u039C\\u0391\\u03AA\\u039F\\u03A5, \\u03A0\\u039F\\u03A1\\u039F\\u03A3, "
+                     "\\u03A1\\u03A5\\u0398\\u039C\\u0399\\u03A3\\u0397");
+    assertGreekUpper("\\u03B0, \\u03A4\\u03B7\\u03C1\\u03CE, \\u039C\\u03AC\\u03B9\\u03BF\\u03C2",
+                     "\\u03AB, \\u03A4\\u0397\\u03A1\\u03A9, \\u039C\\u0391\\u03AA\\u039F\\u03A3");
+    assertGreekUpper("\\u03AC\\u03C5\\u03BB\\u03BF\\u03C2",
+                     "\\u0391\\u03AB\\u039B\\u039F\\u03A3");
+    assertGreekUpper("\\u0391\\u03AB\\u039B\\u039F\\u03A3",
+                     "\\u0391\\u03AB\\u039B\\u039F\\u03A3");
+    assertGreekUpper("\\u0386\\u03BA\\u03BB\\u03B9\\u03C4\\u03B1 "
+                     "\\u03C1\\u03AE\\u03BC\\u03B1\\u03C4\\u03B1 \\u03AE "
+                     "\\u03AC\\u03BA\\u03BB\\u03B9\\u03C4\\u03B5\\u03C2 "
+                     "\\u03BC\\u03B5\\u03C4\\u03BF\\u03C7\\u03AD\\u03C2",
+                     "\\u0391\\u039A\\u039B\\u0399\\u03A4\\u0391 "
+                     "\\u03A1\\u0397\\u039C\\u0391\\u03A4\\u0391 \\u0397\\u0301 "
+                     "\\u0391\\u039A\\u039B\\u0399\\u03A4\\u0395\\u03A3 "
+                     "\\u039C\\u0395\\u03A4\\u039F\\u03A7\\u0395\\u03A3");
+    // http://www.unicode.org/udhr/d/udhr_ell_monotonic.html
+    assertGreekUpper("\\u0395\\u03C0\\u03B5\\u03B9\\u03B4\\u03AE \\u03B7 "
+                     "\\u03B1\\u03BD\\u03B1\\u03B3\\u03BD\\u03CE\\u03C1\\u03B9\\u03C3\\u03B7 "
+                     "\\u03C4\\u03B7\\u03C2 \\u03B1\\u03BE\\u03B9\\u03BF\\u03C0\\u03C1\\u03AD"
+                     "\\u03C0\\u03B5\\u03B9\\u03B1\\u03C2",
+                     "\\u0395\\u03A0\\u0395\\u0399\\u0394\\u0397 \\u0397 "
+                     "\\u0391\\u039D\\u0391\\u0393\\u039D\\u03A9\\u03A1\\u0399\\u03A3\\u0397 "
+                     "\\u03A4\\u0397\\u03A3 \\u0391\\u039E\\u0399\\u039F\\u03A0\\u03A1\\u0395"
+                     "\\u03A0\\u0395\\u0399\\u0391\\u03A3");
+    assertGreekUpper("\\u03BD\\u03BF\\u03BC\\u03B9\\u03BA\\u03BF\\u03CD \\u03AE "
+                     "\\u03B4\\u03B9\\u03B5\\u03B8\\u03BD\\u03BF\\u03CD\\u03C2",
+                     "\\u039D\\u039F\\u039C\\u0399\\u039A\\u039F\\u03A5 \\u0397\\u0301 "
+                     "\\u0394\\u0399\\u0395\\u0398\\u039D\\u039F\\u03A5\\u03A3");
+    // http://unicode.org/udhr/d/udhr_ell_polytonic.html
+    assertGreekUpper("\\u1F18\\u03C0\\u03B5\\u03B9\\u03B4\\u1F74 \\u1F21 "
+                     "\\u1F00\\u03BD\\u03B1\\u03B3\\u03BD\\u1F7D\\u03C1\\u03B9\\u03C3\\u03B7",
+                     "\\u0395\\u03A0\\u0395\\u0399\\u0394\\u0397 \\u0397 "
+                     "\\u0391\\u039D\\u0391\\u0393\\u039D\\u03A9\\u03A1\\u0399\\u03A3\\u0397");
+    assertGreekUpper("\\u03BD\\u03BF\\u03BC\\u03B9\\u03BA\\u03BF\\u1FE6 \\u1F22 "
+                     "\\u03B4\\u03B9\\u03B5\\u03B8\\u03BD\\u03BF\\u1FE6\\u03C2",
+                     "\\u039D\\u039F\\u039C\\u0399\\u039A\\u039F\\u03A5 \\u0397\\u0301 "
+                     "\\u0394\\u0399\\u0395\\u0398\\u039D\\u039F\\u03A5\\u03A3");
+    // From Google bug report
+    assertGreekUpper("\\u039D\\u03AD\\u03BF, "
+                     "\\u0394\\u03B7\\u03BC\\u03B9\\u03BF\\u03C5\\u03C1\\u03B3\\u03AF\\u03B1",
+                     "\\u039D\\u0395\\u039F, "
+                     "\\u0394\\u0397\\u039C\\u0399\\u039F\\u03A5\\u03A1\\u0393\\u0399\\u0391");
+    // http://crbug.com/234797
+    assertGreekUpper("\\u0395\\u03BB\\u03AC\\u03C4\\u03B5 \\u03BD\\u03B1 \\u03C6\\u03AC\\u03C4\\u03B5 "
+                     "\\u03C4\\u03B1 \\u03BA\\u03B1\\u03BB\\u03CD\\u03C4\\u03B5\\u03C1\\u03B1 "
+                     "\\u03C0\\u03B1\\u03CA\\u03B4\\u03AC\\u03BA\\u03B9\\u03B1!",
+                     "\\u0395\\u039B\\u0391\\u03A4\\u0395 \\u039D\\u0391 \\u03A6\\u0391\\u03A4\\u0395 "
+                     "\\u03A4\\u0391 \\u039A\\u0391\\u039B\\u03A5\\u03A4\\u0395\\u03A1\\u0391 "
+                     "\\u03A0\\u0391\\u03AA\\u0394\\u0391\\u039A\\u0399\\u0391!");
+    assertGreekUpper("\\u039C\\u03B1\\u0390\\u03BF\\u03C5, \\u03C4\\u03C1\\u03CC\\u03BB\\u03B5\\u03CA",
+                     "\\u039C\\u0391\\u03AA\\u039F\\u03A5, \\u03A4\\u03A1\\u039F\\u039B\\u0395\\u03AA");
+    assertGreekUpper("\\u03A4\\u03BF \\u03AD\\u03BD\\u03B1 \\u03AE \\u03C4\\u03BF "
+                     "\\u03AC\\u03BB\\u03BB\\u03BF.",
+                     "\\u03A4\\u039F \\u0395\\u039D\\u0391 \\u0397\\u0301 \\u03A4\\u039F "
+                     "\\u0391\\u039B\\u039B\\u039F.");
+    // http://multilingualtypesetting.co.uk/blog/greek-typesetting-tips/
+    assertGreekUpper("\\u03C1\\u03C9\\u03BC\\u03AD\\u03B9\\u03BA\\u03B1",
+                     "\\u03A1\\u03A9\\u039C\\u0395\\u03AA\\u039A\\u0391");
+}

diff --git a/icu4c/source/test/intltest/ustrtest.h b/icu4c/source/test/intltest/ustrtest.h
index 9abd68b6aebb69dd5f31ae3e506229b25dcf9d57..6598f03aa06ee0cfeae721a8f0ab284813df0d51 100644 (file)
--- a/icu4c/source/test/intltest/ustrtest.h
+++ b/icu4c/source/test/intltest/ustrtest.h
@@ -9,6 +9,7 @@
 #ifndef UNICODESTRINGTEST_H
 #define UNICODESTRINGTEST_H
 
+#include "unicode/locid.h"
 #include "unicode/unistr.h"
 #include "intltest.h"
 
@@ -95,7 +96,7 @@ public:
 
 class StringCaseTest: public IntlTest {
 public:
-    StringCaseTest() {}
+    StringCaseTest();
     virtual ~StringCaseTest();
 
     void runIndexedTest(int32_t index, UBool exec, const char *&name, char *par=0);
@@ -108,6 +109,12 @@ public:
                         void *iter, const char *localeID, uint32_t options);
     void TestCasing();
     void TestFullCaseFoldingIterator();
+    void TestGreekUpper();
+
+private:
+    void assertGreekUpper(const char *s, const char *expected);
+
+    Locale GREEK_LOCALE_;
 };
 
 #endif