improvements

regex.c

@@ -3,8 +3,6 @@
 #include <stdint.h>
 #include <stdio.h>
 
-// TODO POSIX ERE use "[]]" for plain ']'
-
 // Checks if the first char in needle matches on c
 // needle may begin with a plain character or a character class
 bool char_match(char c, char *needle, char **nneedl)
@@ -16,27 +14,26 @@ bool char_match(char c, char *needle, char **nneedl)
 	m = false;
 	if (*needle == '[')
 	{
+		// match against character class
 		needle++;
 		neg = *needle == '^';
 		if (neg)
 			needle++;
 		for (db = 1; *needle && (*needle != ']' || db); needle++)
 		{
-			if (db > 1 && ((*needle == c) != neg))
+			if ((*needle == c) != neg)
 				m = true;
-			else if (db > 1)
+			if (db)
 				db--;
-			else if (*needle == '\\')
-				db++;
-			else if (*needle == ']')
-				db--;
-			else if ((*needle == c) != neg)
-				m = true;
 		}
 	}
 	else if (*needle == '$')
+	{
 		m = (c == 0);
-	else if (*needle == '.')
+		while (*needle) needle++;
+		if (nneedl) *nneedl = needle;
+	}
+	else if (*needle == '.' || *needle == 0)
 		m = true;
 	else
 		m = (*needle == c);
@@ -50,15 +47,14 @@ char *next_var(char *needle)
 {
 	for (int dp = 0, db = 0;
 			*needle && (*needle != '|' || dp > 0 || db > 0);
-			needle++) {
+			needle++)
-		if (db && *needle == ']')
+	{
+		if (db > 1)
 			db--;
-		else if (db == 1 && *needle == '\\')
+		else if (db && *needle == ']')
-			db++;
-		else if (db == 2)
 			db--;
 		else if (!db && *needle == '[')
-			db++;
+			db = 2;
 		else if (!db && *needle == '(')
 			dp++;
 		else if (!db && *needle == ')')
@@ -77,18 +73,12 @@ char *next_char(char *needle)
 	int db;
 	if (*needle && *needle != '[')
 		return needle + 1;
-	for (db = (*needle == '[' ? 1 : -1);
+	for (db = (*needle == '[' ? 2 : 0);
-			*needle && db;
+			*needle && (*needle != ']' || db);
 			needle++)
 	{
-		if (*needle == ']')
+		if (db > 1)
 			db--;
-		else if (db == 1 && *needle == '\\')
-			db++;
-		else if (db > 1)
-			db--;
-		else if (db < 0)
-			db++;
 	}
 	return needle;
 }
@@ -116,25 +106,17 @@ char *start_of_group(char *needle0, char *needle)
 		else if (!db && *needle == '(')
 			dp--;
 		else if (!db && *needle == ']')
-			db++;
+			db = 1;
 		else if (db && *needle == '[')
-			db--;
+			db = 0;
-		else if (db && *needle == '\\')
-			db--;
 	}
 	return needle;
 }
 
 char quant(char c)
 {
-	switch (c)
+	char *r = strchr("*+?", c);
-	{
+	return r ? *r : 0;
-		case '*':
-		case '+':
-		case '?':
-			return c;
-	}
-	return 0;
 }
 
 // Tries to prefix-match the regex needle on haystack
@@ -152,24 +134,21 @@ bool regex_match(char *haystack, char *needle)
 	states = strlen(needle) + 1;
 	// active states
 	bool bv[states] = { };
-	// next iteration active states
-	bool bvtmp[states] = { };
 	// acceptance state
 	accept = states - 1;
 	na = false;
 
 	// first, decativate all states
-	for (size_t st = 0; st < states; st++)
+	for (st = 0; st < states; st++)
 		bv[st] = false;
 	// then, find all top-level variants, and activate them
-	for (char *var = needle; *var; var = next_var(var)) {
+	for (var = needle; *var; var = next_var(var))
 		bv[var - needle] = true;
-	}
 
 	// iterate over the haystack until accepting
 	do {
 		// handle groups, variants, and quantifier
-		for (size_t st = 0; st < states - 1; st++)
+		for (st = 0; st < states - 1; st++)
 		{
 			if (!bv[st])
 				continue;
@@ -193,19 +172,20 @@ bool regex_match(char *haystack, char *needle)
 				// then, handle quantifiers
 				fprintf(stderr, "quant %c\n", q);
 				fprintf(stderr, "suffix \"%s\"\n", needle + (q ? st + 2 : st + 1));
-				if (q == '+' || q == '*')
+				nneedl = start_of_group(needle, needle + st);
+				if ((q == '+' || q == '*') && !bv[nneedl - needle])
 				{
 					// spicy: if the quantifier allows multiple occurences, we have to activate a previous state
-					nneedl = start_of_group(needle, needle + st);
 					fprintf(stderr, "start of group: \"%s\"\n", nneedl);
 					bv[nneedl - needle] = true;
 					st = nneedl - needle - 1;
 				}
+				else
+					bv[nneedl - needle] = false;
 			}
 			else if (needle[st] == '(')
 			{
 				// at start of group, look at variants and quantifier
-				bv[st] = false;
 				for (var = needle + st + 1; *var && *var != ')'; var = next_var(var))
 						bv[var - needle] = true;
 				if (*var == ')')
@@ -223,31 +203,30 @@ bool regex_match(char *haystack, char *needle)
 					bv[nneedl + 1 - needle] = true;
 			}
 		}
-		memcpy(bvtmp, bv, sizeof(bv));
 		// actual match, backwards
 		na = true;
-		for (st = states; st > 0; st--)
+		for (st = accept; st > 0; st--)
 		{
-			if (!bv[st-1])
+			// skip beginnings of groups b/c they have been handled before
+			if (!bv[st-1] || needle[st-1] == '(')
 				continue;
-			na = false;
 			// perform match for active state
-			fprintf(stderr, "state %lu: \"%s\" ~ \"%s\"\n", st-1, haystack, needle + st - 1);
+			bv[st-1] = false;
-			bvtmp[st-1] = false;
 			m = char_match(*haystack, needle + st-1, &nneedl);
 			q = quant(*nneedl);
-			if (m || q == '*' || q == '?')
+			fprintf(stderr, "state %lu: \"%s\" ~ \"%s\" => %b\n", st-1, haystack, needle + st - 1, m);
+			if (m)
 			{
+				na = false;
 				if (q)
 					nneedl++;
 				if (q == '*' || q == '+')
-					bvtmp[st-1] = true;
+					bv[st-1] = true;
-				bvtmp[nneedl - needle] = true;
+				bv[nneedl - needle] = true;
 			}
 		}
-		memcpy(bv, bvtmp, sizeof(bv));
 		putc('\n', stderr);
-		// check if we have reached the end of haystack, and advance
+		// check if we have reached the end of haystack, and advance if not
 		if (!*(haystack++))
 			break;
 	} while (!bv[accept] && !na);