Refactor regex and implement UTF8 mode for GRegex

- with non-unicode byte to Private Use Area A mapping
- move all ifdefs to iregex.h file only

src/core/iregex-gregex.c

@@ -0,0 +1,137 @@
+#include <string.h>
+
+#include "iregex.h"
+
+const gchar *
+make_valid_utf8(const gchar *text, gboolean *free_ret)
+{
+	GString *str;
+	const gchar *ptr;
+	if (g_utf8_validate(text, -1, NULL)) {
+		if (free_ret)
+			*free_ret = FALSE;
+		return text;
+	}
+
+	str = g_string_sized_new(strlen(text) + 12);
+
+	ptr = text;
+	while (*ptr) {
+		gunichar c = g_utf8_get_char_validated(ptr, -1);
+		/* the unicode is invalid */
+		if (c == (gunichar)-1 || c == (gunichar)-2) {
+			/* encode the byte into PUA-A */
+			g_string_append_unichar(str, (gunichar) (0xfff00 | (*ptr & 0xff)));
+			ptr++;
+		} else {
+			g_string_append_unichar(str, c);
+			ptr = g_utf8_next_char(ptr);
+		}
+	}
+
+	if (free_ret)
+		*free_ret = TRUE;
+	return g_string_free(str, FALSE);
+}
+
+Regex *
+i_regex_new (const gchar *pattern,
+             GRegexCompileFlags compile_options,
+             GRegexMatchFlags match_options,
+             GError **error)
+{
+	const gchar *valid_pattern;
+	gboolean free_valid_pattern;
+	Regex *ret = NULL;
+
+	valid_pattern = make_valid_utf8(pattern, &free_valid_pattern);
+	ret = g_regex_new(valid_pattern, compile_options, match_options, error);
+
+	if (free_valid_pattern)
+		g_free_not_null((gchar *)valid_pattern);
+
+	return ret;
+}
+
+void
+i_regex_unref (Regex *regex)
+{
+	g_regex_unref(regex);
+}
+
+/* if new_string is present, the caller must free new_string.
+   otherwise, g_match_info_get_string must not be used. */
+gboolean
+i_regex_match (const Regex *regex,
+               const gchar *string,
+               GRegexMatchFlags match_options,
+               MatchInfo **match_info,
+               const gchar **new_string)
+{
+	gboolean ret;
+	gboolean free_valid_string;
+	const gchar *valid_string = make_valid_utf8(string, &free_valid_string);
+
+	ret = g_regex_match(regex, valid_string, match_options, match_info);
+	if (free_valid_string) {
+		if (new_string)
+			*new_string = valid_string;
+		else
+			g_free_not_null((gchar *)valid_string);
+	}
+	return ret;
+}
+
+gsize
+strlen_pua_oddly(const char *str)
+{
+	const gchar *ptr;
+	gsize ret = 0;
+	ptr = str;
+
+	while (*ptr) {
+		const gchar *old;
+		gunichar c = g_utf8_get_char(ptr);
+		old = ptr;
+		ptr = g_utf8_next_char(ptr);
+
+		/* it is our PUA encoded byte */
+		if ((c & 0xfff00) == 0xfff00)
+			ret++;
+		else
+			ret += ptr - old;
+	}
+
+	return ret;
+}
+
+gboolean
+i_match_info_fetch_pos (const MatchInfo *match_info,
+                        gint match_num,
+                        gint *start_pos,
+                        gint *end_pos,
+                        const gchar *new_string)
+{
+	gint tmp_start, tmp_end, new_start_pos;
+	gboolean ret;
+
+	if (!new_string || (!start_pos && !end_pos))
+		return g_match_info_fetch_pos(match_info, match_num, start_pos, end_pos);
+
+	ret = g_match_info_fetch_pos(match_info, match_num, &tmp_start, &tmp_end);
+	if (start_pos || end_pos) {
+		gchar *to_start = g_strndup(new_string, tmp_start);
+		new_start_pos = strlen_pua_oddly(to_start);
+		g_free_not_null(to_start);
+
+		if (start_pos)
+			*start_pos = new_start_pos;
+
+		if (end_pos) {
+			gchar *to_end = g_strndup(new_string + tmp_start, tmp_end - tmp_start);
+			*end_pos = new_start_pos + strlen_pua_oddly(to_end);
+			g_free_not_null(to_end);
+		}
+	}
+	return ret;
+}