util: add unicode_word_utf8

This originally use Xapian::Unicode::is_wordchar, but that forces
clients to link directly to libxapian, which seems like it might be
busywork if nothing else.

3 files changed, 57 insertions, 1 deletions

diff --git a/util/Makefile.local b/util/Makefile.local
index ba03230e..46f8af3a 100644
--- a/util/Makefile.local
+++ b/util/Makefile.local
@@ -5,7 +5,8 @@ extra_cflags += -I$(srcdir)/$(dir)
 
 libnotmuch_util_c_srcs := $(dir)/xutil.c $(dir)/error_util.c $(dir)/hex-escape.c \
 		  $(dir)/string-util.c $(dir)/talloc-extra.c $(dir)/zlib-extra.c \
-		$(dir)/util.c $(dir)/gmime-extra.c $(dir)/crypto.c
+		$(dir)/util.c $(dir)/gmime-extra.c $(dir)/crypto.c \
+		$(dir)/unicode-util.c
 
 libnotmuch_util_modules := $(libnotmuch_util_c_srcs:.c=.o)
 
diff --git a/util/unicode-util.c b/util/unicode-util.c
new file mode 100644
index 00000000..312e900f
--- /dev/null
+++ b/util/unicode-util.c
@@ -0,0 +1,43 @@
+#include "unicode-util.h"
+
+/* Based on Xapian::Unicode::is_wordchar, to avoid forcing clients to
+   link directly to libxapian.
+*/
+
+static bool
+unicode_is_wordchar (notmuch_unichar ch)
+{
+    switch (g_unichar_type (ch)) {
+    case G_UNICODE_UPPERCASE_LETTER:
+    case G_UNICODE_LOWERCASE_LETTER:
+    case G_UNICODE_TITLECASE_LETTER:
+    case G_UNICODE_MODIFIER_LETTER:
+    case G_UNICODE_OTHER_LETTER:
+    case G_UNICODE_NON_SPACING_MARK:
+    case G_UNICODE_ENCLOSING_MARK:
+    case G_UNICODE_SPACING_MARK:
+    case G_UNICODE_DECIMAL_NUMBER:
+    case G_UNICODE_LETTER_NUMBER:
+    case G_UNICODE_OTHER_NUMBER:
+    case G_UNICODE_CONNECT_PUNCTUATION:
+	return true;
+    default:
+	return false;
+    }
+}
+
+bool
+unicode_word_utf8 (const char *utf8_str)
+{
+    gunichar *decoded = g_utf8_to_ucs4_fast (utf8_str, -1, NULL);
+    const gunichar *p = decoded;
+    bool ret;
+
+    while (*p && unicode_is_wordchar (*p))
+	p++;
+
+    ret =  (*p == '\0');
+
+    g_free (decoded);
+    return ret;
+}
diff --git a/util/unicode-util.h b/util/unicode-util.h
new file mode 100644
index 00000000..32d1e6ef
--- /dev/null
+++ b/util/unicode-util.h
@@ -0,0 +1,12 @@
+#ifndef UNICODE_UTIL_H
+#define UNICODE_UTIL_H
+
+#include <stdbool.h>
+#include <gmodule.h>
+
+/* The utf8 encoded string would tokenize as a single word, according
+ * to xapian. */
+bool unicode_word_utf8 (const char *str);
+typedef gunichar notmuch_unichar;
+
+#endif