1 files changed, 74 insertions, 12 deletions

diff --git a/src/search.c b/src/search.c
index 601cc55..ce4a338 100644
--- a/src/search.c
+++ b/src/search.c
@@ -166,6 +166,53 @@ teco_state_search_initial(teco_machine_main_t *ctx, GError **error)
 	return TRUE;
 }
 
+/**
+ * Escape an arbitrary string, so it can be embedded into an
+ * Advanced Regular Expression.
+ *
+ * This is similar to g_regex_escape_string() but in contrast
+ * the result can be used within character classes as well.
+ *
+ * @param str String to escape.
+ * @param len Length of string.
+ * @return Regular expression as a newly allocated null-terminated string.
+ */
+static gchar *
+teco_regex_escape(const gchar *str, gsize len)
+{
+	/*
+	 * Considering the expected size of strings to escape,
+	 * it's probably not worth to calculate a minimal size.
+	 */
+	gchar *escaped = g_malloc(len*4+1);
+	gchar *p = escaped;
+
+	/*
+	 * This works in UTF-8 as well since none of the escaped characters
+	 * can be a continuation byte.
+	 */
+	while (len > 0) {
+		if (!*str) {
+			/* there is no shorter __unambiguous__ way */
+			*p++ = '\\';
+			*p++ = '0';
+			*p++ = '0';
+			*p++ = '0';
+		} else if (strchr("^.*+?(|[-]{\\$", *str)) {
+			*p++ = '\\';
+			*p++ = *str;
+		} else {
+			*p++ = *str;
+		}
+
+		str++;
+		len--;
+	}
+	*p = '\0';
+
+	return escaped;
+}
+
 typedef enum {
 	TECO_SEARCH_STATE_START,
 	TECO_SEARCH_STATE_CTL,
@@ -230,7 +277,7 @@ teco_class2regexp(teco_search_state_t *state, teco_string_t *pattern,
 					return g_strdup("");
 				gsize len = codepage == SC_CP_UTF8
 						? g_utf8_next_char(pattern->data) - pattern->data : 1;
-				gchar *escaped = g_regex_escape_string(pattern->data, len);
+				gchar *escaped = teco_regex_escape(pattern->data, len);
 				pattern->data += len;
 				pattern->len -= len;
 				return escaped;
@@ -325,7 +372,7 @@ teco_class2regexp(teco_search_state_t *state, teco_string_t *pattern,
 				pattern->data += len;
 				pattern->len -= len;
 				*state = TECO_SEARCH_STATE_START;
-				return g_regex_escape_string(str.data ? : "", str.len);
+				return teco_regex_escape(str.data ? : "", str.len);
 			}
 			break;
 		}
@@ -378,14 +425,29 @@ teco_pattern2regexp(teco_string_t *pattern, teco_machine_qregspec_t *qreg_machin
 	g_auto(teco_string_t) re = {NULL, 0};
 
 	do {
-		/*
-		 * Previous character was caret.
-		 * Make sure it is handled like a control character.
-		 * This is necessary even though we have string building activated,
-		 * to support constructs like ^Q^Q (typed with carets) in order to
-		 * quote pattern matching characters.
-		 */
 		if (state == TECO_SEARCH_STATE_CTL) {
+			if (*pattern->data == '~') {
+				/* rest of pattern is a regular expression */
+				teco_string_append(&re, pattern->data+1, pattern->len-1);
+				/*
+				 * FIXME: In terex, it actually could contain null bytes.
+				 */
+				if (teco_string_contains(re, '\0')) {
+					g_set_error_literal(error, TECO_ERROR, TECO_ERROR_FAILED,
+					                    "Regular expression must not contain null-byte"
+					                    " - use \\0 instead");
+					return NULL;
+				}
+				return g_steal_pointer(&re.data) ? : g_strdup("");
+			}
+
+			/*
+			 * Previous character was caret.
+			 * Make sure it is handled like a control character.
+			 * This is necessary even though we have string building activated,
+			 * to support constructs like ^Q^Q (typed with carets) in order to
+			 * quote pattern matching characters.
+			 */
 			*pattern->data = TECO_CTL_KEY(g_ascii_toupper(*pattern->data));
 			state = TECO_SEARCH_STATE_START;
 		}
@@ -444,7 +506,7 @@ teco_pattern2regexp(teco_string_t *pattern, teco_machine_qregspec_t *qreg_machin
 			gsize len = codepage == SC_CP_UTF8
 					? g_utf8_next_char(pattern->data) - pattern->data : 1;
 			/* the allocation could theoretically be avoided by escaping char-wise */
-			g_autofree gchar *escaped = g_regex_escape_string(pattern->data, len);
+			g_autofree gchar *escaped = teco_regex_escape(pattern->data, len);
 			teco_string_append(&re, escaped, strlen(escaped));
 			pattern->data += len;
 			pattern->len -= len;
@@ -736,8 +798,8 @@ teco_do_search_backwards(regex_t *re, gsize from, gsize to, gint *count, GError
 	gsize to_block = to-from;
 
 	while (to_block > 0) {
-		gsize from_block = teco_search_block_size > 0
-					? MAX(0, to_block - teco_search_block_size) : 0;
+		gsize from_block = teco_search_block_size > 0 && to_block >= teco_search_block_size
+				 	? to_block - teco_search_block_size : 0;
 		/* how many bytes have been consumed in the current block */
 		gsize offset = 0;