pattern match characters support ^Q/^R now as well

* makes it possible, albeit cumbersome, to escape pattern match characters
* For instance, to search for ^Q, you now have to type
  S^Q^Q^Q^Q$.
  To search for ^E you have to type
  S^Q^Q^Q^E$.
  But the last character cannot be typed with carets currently (FIXME?).
  For pattern-only characters, two ^Q should be sufficient as in
  S^Q^Q^X$.
* Perhaps it would be more elegant to abolish the difference between string building
  and pattern matching characters to avoid double quoting.
  But then all string building constructs like ^EQq should operate at the pattern level
  as well (ie. match the contents of register q verbatim instead of being interpreted as a pattern).
  TECOC and TECO-64 don't do that either.
  If we leave everything as it is, at least a new string building construct should be added for
  auto-quoting patterns (analoguous to ^EN and ^E@).

4 files changed, 57 insertions, 11 deletions

diff --git a/doc/sciteco.7.template b/doc/sciteco.7.template
index ca93fa6..81deac0 100644
--- a/doc/sciteco.7.template
+++ b/doc/sciteco.7.template
@@ -1789,6 +1789,14 @@ The following pattern match constructs are supported for matching
 one character in different character classes
 (caret-notations refer to the corresponding control characters):
 .TP
+.BI ^Q c
+.TQ
+.BI ^R c
+Escape character \fIc\fP.
+Since these are interpreted as string building characters as well,
+you may have to type two or three \fB^Q\fP in a row to escape a
+pattern match character.
+.TP
 .SCITECO_TOPIC ^S ^EB
 .B ^S
 .TQ
diff --git a/src/parser.c b/src/parser.c
index 7ba9876..8cb26e7 100644
--- a/src/parser.c
+++ b/src/parser.c
@@ -603,6 +603,7 @@ teco_state_stringbuilding_ctle_input(teco_machine_stringbuilding_t *ctx, gunicha
 	case 'N':  next = &teco_state_stringbuilding_ctle_n; break;
 	default:
 		if (ctx->result) {
+			/* also makes sure that search patterns can start with ^E */
 			gchar buf[1+6] = {TECO_CTL_KEY('E')};
 			gsize len = g_unichar_to_utf8(chr, buf+1);
 			teco_machine_stringbuilding_append(ctx, buf, 1+len);
diff --git a/src/search.c b/src/search.c
index ed3a00c..c9a2ba0 100644
--- a/src/search.c
+++ b/src/search.c
@@ -115,6 +115,8 @@ teco_state_search_initial(teco_machine_main_t *ctx, GError **error)
 
 typedef enum {
 	TECO_SEARCH_STATE_START,
+	TECO_SEARCH_STATE_CTL,
+	TECO_SEARCH_STATE_ESCAPE,
 	TECO_SEARCH_STATE_NOT,
 	TECO_SEARCH_STATE_CTL_E,
 	TECO_SEARCH_STATE_ANYQ,
@@ -320,6 +322,18 @@ teco_pattern2regexp(teco_string_t *pattern, guint codepage, gboolean single_expr
 
 	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) {
+			*pattern->data = TECO_CTL_KEY(g_ascii_toupper(*pattern->data));
+			state = TECO_SEARCH_STATE_START;
+		}
+
+		/*
 		 * First check whether it is a class.
 		 * This will not treat individual characters
 		 * as classes, so we do not convert them to regexp
@@ -347,21 +361,36 @@ teco_pattern2regexp(teco_string_t *pattern, guint codepage, gboolean single_expr
 		switch (state) {
 		case TECO_SEARCH_STATE_START:
 			switch (*pattern->data) {
-			case TECO_CTL_KEY('X'): teco_string_append_c(&re, '.'); break;
-			case TECO_CTL_KEY('N'): state = TECO_SEARCH_STATE_NOT; break;
-			default: {
-				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);
-				teco_string_append(&re, escaped, strlen(escaped));
-				pattern->data += len;
-				pattern->len -= len;
+			case '^':
+				state = TECO_SEARCH_STATE_CTL;
+				break;
+			case TECO_CTL_KEY('Q'):
+			case TECO_CTL_KEY('R'):
+				state = TECO_SEARCH_STATE_ESCAPE;
+				break;
+			case TECO_CTL_KEY('X'):
+				teco_string_append_c(&re, '.');
+				break;
+			case TECO_CTL_KEY('N'):
+				state = TECO_SEARCH_STATE_NOT;
+				break;
+			default:
+				state = TECO_SEARCH_STATE_ESCAPE;
 				continue;
 			}
-			}
 			break;
 
+		case TECO_SEARCH_STATE_ESCAPE: {
+			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);
+			teco_string_append(&re, escaped, strlen(escaped));
+			pattern->data += len;
+			pattern->len -= len;
+			continue;
+		}
+
 		case TECO_SEARCH_STATE_NOT: {
 			state = TECO_SEARCH_STATE_START;
 			g_autofree gchar *temp = teco_class2regexp(&state, pattern, codepage, TRUE, error);
diff --git a/tests/testsuite.at b/tests/testsuite.at
index fc7de4f..33f1bf5 100644
--- a/tests/testsuite.at
+++ b/tests/testsuite.at
@@ -79,6 +79,14 @@ AT_CHECK([$SCITECO -e "[[a 23Ub ]]b Qb\"N(0/0)'"], 0, ignore, ignore)
 AT_CHECK([$SCITECO -e "[[\$ @FG'..' ]]\$ :Q\$-1Q\$-^^r\"=(0/0)'"], 0, ignore, ignore)
 AT_CLEANUP
 
+AT_SETUP([Searches])
+# FIXME: We cannot currently easily insert a single ASCII 5 (^E), as it must be followed
+# by a 2nd character. It can be quoted, but cannot be written as Caret+E.
+# You also cannot search for a single ASCII 5 using Caret+E.
+# 2 additional ^Q are translated to a single ^Q and interpreted at the search-pattern layer.
+AT_CHECK([$SCITECO -e "@I/^Q\05/ J @:S/^Q^Q^Q\05/\"F(0/0)'"], 0, ignore, ignore)
+AT_CLEANUP
+
 AT_SETUP([Editing local registers in macro calls])
 AT_CHECK([$SCITECO -e '@^Ua{@EQ.x//} :Ma @^U.x/FOO/'], 0, ignore, ignore)
 AT_CHECK([$SCITECO -e '@^Ua{@EQ.x//}  Ma @^U.x/FOO/'], 1, ignore, ignore)