Implement isBoundary() for combining sequences.

diff --git a/ext/unicode/unicode_iterators.c b/ext/unicode/unicode_iterators.c
index c8fc9c0714c6fc3697b12a92e9e8f136eebaec62..862c40e995d5cd48d5e6d1a1698ef0c81d87ab9c 100644 (file)
--- a/ext/unicode/unicode_iterators.c
+++ b/ext/unicode/unicode_iterators.c
@@ -448,7 +448,7 @@ static void text_iter_cs_rewind(text_iter_obj *object, long flags TSRMLS_DC)
 
 static void text_iter_cs_following(text_iter_obj *object, int32_t offset, long flags TSRMLS_DC)
 {
-       int32_t k, tmp;
+       int32_t k;
 
        if (offset < 0) {
                offset = 0;
@@ -501,7 +501,54 @@ static void text_iter_cs_following(text_iter_obj *object, int32_t offset, long f
 
 static zend_bool text_iter_cs_isBoundary(text_iter_obj *object, int32_t offset, long flags TSRMLS_DC)
 {
-       return 1;
+       UChar32 cp;
+       int32_t k, tmp;
+       zend_bool result;
+
+       if (offset < 0) {
+               offset = 0;
+       }
+
+       /*
+        * On invalid iterator we always want to start looking for the code unit
+        * offset from the beginning of the string.
+        */
+       if (object->u.cs.start_cp_offset == UBRK_DONE) {
+               object->u.cs.start_cp_offset = 0;
+               object->u.cs.start = 0;
+       }
+
+       /*
+        * Try to locate the code unit position relative to the last known codepoint
+        * offset.
+        */
+       k = object->u.cs.start;
+       if (offset > object->u.cs.start_cp_offset) {
+               U16_FWD_N(object->text, k, object->text_len, offset - object->u.cs.start_cp_offset);
+       } else {
+               U16_BACK_N(object->text, 0, k, object->u.cs.start_cp_offset - offset);
+       }
+
+       /* end of the text is always a boundary */
+       if (k == object->text_len) {
+               offset = u_countChar32(object->text, object->text_len);
+               result = 1;
+       } else {
+       /* if the next codepoint is a base character, it's a boundary */
+               tmp = k;
+               U16_NEXT(object->text, tmp, object->text_len, cp);
+               result = (u_getCombiningClass(cp) == 0);
+       }
+
+       if (k == object->u.cs.start) {
+               return result;
+       }
+
+       object->u.cs.start = k;
+       object->u.cs.start_cp_offset = offset;
+       object->u.cs.end = object->u.cs.start;
+
+       return result;
 }
 
 static text_iter_ops text_iter_cs_ops = {