Make Union and Blocks, avoid postfix notation

src/lexer.c

@@ -1,5 +1,6 @@
 #include <stdlib.h>
 #include <stdio.h>
+#include <ctype.h>
 #include "main.h"
 
 struct patch_list
@@ -11,10 +12,10 @@ struct patch_list
 struct frag 
 {
 	struct rgx_nfa *start;
-	struct patch_list *out; 
+	struct patch_list *pl; 
 };
 
-struct patch_list * append(struct patch_list * pl, struct rgx_nfa ** nfa)
+struct patch_list * pl_append(struct patch_list * pl, struct rgx_nfa ** nfa)
 {
 	struct patch_list * res, * root = pl;
 	res = malloc(sizeof(struct patch_list)); 
@@ -33,7 +34,25 @@ struct patch_list * append(struct patch_list * pl, struct rgx_nfa ** nfa)
 	return root;
 }
 
-struct rgx_nfa * patch(struct patch_list *pl, struct rgx_nfa *to) {
+void pl_merge(struct patch_list ** pl1, struct patch_list * pl2)
+{
+	struct patch_list * pl = *pl1;
+
+	if (!pl ) {
+		*pl1 = pl2;
+		return;
+	}
+
+	while (pl->next) {
+		pl = pl->next;
+	}
+
+	pl->next = pl2;
+
+}
+
+struct rgx_nfa * patch(struct patch_list *pl, struct rgx_nfa *to) 
+{
 	while (pl) {
 		*pl->ptr = to;
 		pl = pl->next;
@@ -41,6 +60,29 @@ struct rgx_nfa * patch(struct patch_list *pl, struct rgx_nfa *to) {
 	return to;
 }
 
+void collapse(struct frag * stack, unsigned char * i) 
+{
+	if (stack[*i].start->op == BLOCK ) {
+		stack[(*i)--] ;
+		return;
+	}
+
+	while (*i > 0 )
+	{
+		struct frag to;
+
+		to = stack[(*i)--];
+
+		if (stack[*i].start->op == BLOCK ) {
+			stack[*i]= to;
+			return;
+		}
+
+		patch(stack[*i].pl, to.start);
+		stack[*i].pl = to.pl;
+	}
+}
+
 struct rgx_nfa * mknfa(struct rgx_nfa nfa)
 {
 	struct rgx_nfa * res;
@@ -50,91 +92,115 @@ struct rgx_nfa * mknfa(struct rgx_nfa nfa)
 }
 
 struct rgx_nfa * rgx_compile(struct rgx_nfa * l, char * s, int v) {
-	struct rgx_nfa *r;
 	unsigned char i=-1;
-	struct frag stack[0xff], a, b;
+	struct frag stack[0xff];
+	struct patch_list  * union_pl = NULL;
+
+	if (l) {
+		struct rgx_nfa *r;
+		r = mknfa((struct rgx_nfa){UNION, {l, NULL}, 1});
+		stack[++i] = (struct frag){r, pl_append(NULL, &r->node[1])};
+	}
 
 	for(;*s; ++s) 
 	{
+		struct rgx_nfa *r;
+		struct frag a, b;
 
 		switch(*s)
 		{
+		case '(':
+			r = mknfa((struct rgx_nfa){BLOCK, {NULL}, 0});
+			stack[++i] = (struct frag){r, NULL };
+			break;
+		case ')':
+			collapse(stack, &i);
+			pl_merge(&stack[i].pl, union_pl);
+			union_pl=NULL;
+			break;
 		case '|':
+			collapse(stack, &i);
+			pl_merge(&union_pl, stack[i].pl);
+
+			if (i!= 0) {
+				a = stack[i--];
+				r = mknfa((struct rgx_nfa){BLOCK, {NULL}, 0});
+				stack[++i] = (struct frag){r, NULL };
+				stack[++i]= a;
+			}
+
+			r = mknfa((struct rgx_nfa){UNION, {stack[i].start, NULL}, 0});
+			stack[i] = (struct frag){r, pl_append(NULL, &r->node[1])};
 			break;
 
+		case '\\':
+			switch(*(++s))
+			{
+			case 's':
+			case 'd':
+			case 'w':
+				a.start = mknfa((struct rgx_nfa){CLASS, {NULL}, *s});
+				a.pl =  pl_append(NULL, &a.start->node[0]);
+				stack[++i] = a;
+				break;
+			case '\\':
+			case '|':
+			case '(':
+			case ')':
+			case '*':
+			case '+':
+			case '?':
+				a.start = mknfa((struct rgx_nfa){RUNE, {NULL}, *s});
+				a.pl =  pl_append(NULL, &a.start->node[0]);
+				stack[++i] = a;
+				break;
+			default:
+				a.start = mknfa((struct rgx_nfa){RUNE, {NULL}, '\\'});
+				a.pl =  pl_append(NULL, &a.start->node[0]);
+				stack[++i] = a;
+
+				a.start = mknfa((struct rgx_nfa){RUNE, {NULL}, *s});
+				a.pl =  pl_append(NULL, &a.start->node[0]);
+				stack[++i] = a;
+				break;
+			}
+			break;
 		case '*':
-			b = stack[i--];
+			r = mknfa((struct rgx_nfa){UNION, {stack[i].start, NULL}, 0});
-
+			patch(stack[i].pl, r);
-			if (i==0xff) 
+			stack[i] = (struct frag){r, pl_append(NULL, &r->node[1])};
-			{
+			break;
-				r = mknfa((struct rgx_nfa){UNION, {b.start, NULL}});
-				patch(b.out, r);
-				stack[++i] = (struct frag){r, append(NULL, &r->node[1])};
-				break;
-			}
-			else
-			{
-				a = stack[i--];
-				r = mknfa((struct rgx_nfa){UNION, {b.start, NULL}});
-				patch(a.out, r);
-				patch(b.out, r);
-				a.out = append(NULL, &r->node[1]);
-				stack[++i] = a;
-				break;
-			}
 		case '?': 
-			b = stack[i--];
+			r = mknfa((struct rgx_nfa){UNION, {stack[i].start, NULL}, 0});
-
+			stack[i] = (struct frag){r, pl_append(stack[i].pl, &r->node[1]) };
-			if (i==0xff) 
+			break; 
-			{
-				r = mknfa((struct rgx_nfa){UNION, {b.start, NULL}});
-				b.out = append(b.out, &r->node[1]);
-				stack[++i] = (struct frag){r, b.out };
-				break; 
-			}
-			else
-			{
-				a = stack[i--];
-				r = mknfa((struct rgx_nfa){UNION, {b.start, NULL}});
-				a.out = append(b.out, &patch(a.out, r)->node[1]);
-				stack[++i] = a;
-				break; 
-			}
-
 		case '+':
-			// This implemention may cause a infinite loop that crashes.
+			r = mknfa((struct rgx_nfa){UNION, {stack[i].start, NULL}, 0});
-			a = stack[i--];
+			stack[i].pl = pl_append(NULL, &patch(stack[i].pl, r)->node[1]);
-			r = mknfa((struct rgx_nfa){UNION, {NULL, a.start}});
-			a.out = append(NULL, &patch(a.out, r)->node[0]);
-			stack[++i] = a;
 			break;
 		case '.':
-			a.start = mknfa((struct rgx_nfa){C_ANY, {NULL}});
+			a.start = mknfa((struct rgx_nfa){ANY, {NULL}});
-			a.out =  append(NULL, &a.start->node[0]);
+			a.pl =  pl_append(NULL, &a.start->node[0]);
 			stack[++i] = a;
 			break;
 		default:
 			a.start = mknfa((struct rgx_nfa){RUNE, {NULL}, *s});
-			a.out =  append(NULL, &a.start->node[0]);
+			a.pl =  pl_append(NULL, &a.start->node[0]);
 			stack[++i] = a;
 			break;
 		}
 	}
 
-	// Append Final
+	// Collapse and add FINAL state
-	{
+	while ( i!=0 ) {
-		stack[++i].start = mknfa((struct rgx_nfa){FINAL, {NULL}, v});
+		if (stack[i].start->op == BLOCK )
-		stack[i].out =  append(NULL, &stack[i].start->node[0]);
+			stack[i--];
-	}
+		collapse(stack, &i);
-
-	// Collapse Stack
-	while (i > 0)
-	{
-		struct frag to;
-		to = stack[i--];
-		patch(stack[i].out, to.start);
 	}
+	pl_merge(&stack[i].pl, union_pl);
+	union_pl = NULL;
 
+	patch(stack[i].pl, mknfa((struct rgx_nfa){FINAL, {NULL}, v}));
 	return stack[i].start;
 }
 
@@ -144,19 +210,47 @@ int rgx_run(struct rgx_nfa *l, char *s) {
 		switch (l->op)
 		{
 		case FINAL:
+			//printf("FINAL: %c\n", *s? *s:'0');
 			return *s ? 0: l->c;
-		case C_ANY:
+		case ANY:
+			//printf("ANY: %c\n", *s? *s:'0');
 			if (!*(s++)) 
 				return 0;
 			l = l->node[0];
 			break;
+		case CLASS:
+			//printf("CLASS(%c): %c\n", l->c, *s? *s: '0');
+			switch(l->c)
+			{
+			case 's':
+				if(!isspace(*(s++)))
+					return 0;
+				l = l->node[0];
+				break;
+			case 'd':
+				if(!isdigit(*(s++)))
+					return 0;
+				l = l->node[0];
+				break;
+			case 'w':
+				if(!isalpha(*(s++)))
+					return 0;
+				l = l->node[0];
+				break;
+			default:
+				return 0;
+			}
+			break;
 		case RUNE:
+			//printf("RUNE(%c): %c\n", l->c, *s?*s:'0');
 			if (*(s++) != l->c)
 				return 0;
 			l = l->node[0];
 			break;
 		case UNION:
+			//printf("{\n");
 			int res = rgx_run(l->node[0], s);
+			//printf("}\n");
 			if (res)
 				return res;
 			l = l->node[1];

src/main.c

@@ -0,0 +1,34 @@
+#include <stdio.h>
+#include "main.h"
+
+char * tokens[] = 
+{
+	"\"(\\\"|\"!)*\"",
+	"\\d+",
+	"L|R|N",
+	"\\w+",
+	"\\s+",
+	"[\\s*",
+	"\\s*]",
+	"\\s*:\\s*",
+	"\\s*\\|\\s*",
+	"\\s*<\\s*",
+	"\\s*;",
+	NULL,
+};
+
+int main(void)
+{
+	int i;
+	struct rgx_nfa * l = NULL;
+
+	//for (i=0; tokens[i]; ++i) {
+		l=rgx_compile(l, tokens[0], 1);
+	//}
+
+	printf("\n%d\n", rgx_run(l, "\"\"")) ;
+	printf("\n%d\n", rgx_run(l, "\"b\"")) ;
+	printf("\n%d\n", rgx_run(l, "\"bb\"")) ;
+	printf("\n%d\n", rgx_run(l, "\"\"\"")) ;
+	printf("\n%d\n", rgx_run(l, "\"\\\"\"")) ;
+}

src/main.h

@@ -1,6 +1,6 @@
 struct rgx_nfa
 {
-	enum {RUNE, C_ANY, UNION, FINAL} op;
+	enum {RUNE, ANY, UNION, FINAL, CLASS, BLOCK} op;
 	struct rgx_nfa *node[2];
 	unsigned char c;
 };

src/test.c

@@ -24,27 +24,46 @@ struct test
 	{"a.*c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"acc", 1}, {"ac", 1}, {"abbc", 1}, {NULL}}},
 	{"a.?c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"acc", 1}, {"ac", 1}, {NULL}}},
 
+	// TODO: Implement NFA to DFA to avoid loops
+
 	// Equal to *
 	{"ab+?c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 1}, {"ac", 1}, {NULL}}},
 	{"ab*?c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 1}, {"ac", 1}, {NULL}}},
-	{"ab?+c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"ababc", 1}, {"ac", 1}, {NULL}}},
+	//{"ab?+c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"ababc", 1}, {"ac", 1}, {NULL}}},
-	{"ab?*c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"ababc", 1}, {"ac", 1}, {NULL}}},
+	//{"ab?*c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"ababc", 1}, {"ac", 1}, {NULL}}},
-
-	// . Alt
-	//{"a|b", {{"a", 1},{"abc", 0}, {"b", 1 }, {"ba", 0}, {NULL}}},
-
-	{"ab??c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 0}, {"ac", 1}, {"c", 1}, {NULL}}},
-	{"ab++c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 1}, {"ac", 0}, {"c", 0}, {"ababc",0}, {NULL}}},
-	{"ab**c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 1}, {"ac", 1}, {"c", 1}, {"ababc",1}, {NULL}}},
-
-	// Xd
-	{"a?", {{"aaaa", 0}, {"a", 1}, {NULL}}},
-	{"a*", {{"aaaa", 1}, {"a", 1}, {NULL}}},
-	// REDOS
 	{"a+*", {{"aaaa", 1}, {"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa", 1}, {NULL}}},
+	//{"a**", {{"aaaa", 1}, {"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa", 1}, {NULL}}},
+
+	// Nilpotent
+	{"ab??c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 0}, {"ac", 1}, {"c", 0}, {NULL}}},
+	{"ab++c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 1}, {"ac", 0}, {"c", 0}, {"ababc",0}, {NULL}}},
+	//{"ab**c", {{"a", 0},{"abc", 1}, {"abcd",0 }, {"abbc", 1}, {"ac", 1}, {"c", 1}, {"ababc",1}, {NULL}}},
+
+	// UNION
+	{"a()", {{"a", 1},{"abc", 0}, {"b", 0 }, {"ba", 0}, {NULL}}},
+	{"a(b)", {{"ab", 1},{"abc", 0}, {"b", 0 }, {"ba", 0}, {NULL}}},
+	{"a(b)*", {{"ab", 1},{"abb", 1}, {"a", 1 }, {"ba", 0}, {NULL}}},
+	{"a|b", {{"a", 1},{"abc", 0}, {"b", 1 }, {"ba", 0}, {NULL}}},
+	{"(a|.)*", {{"a", 1},{"abc", 1}, {"b", 1 }, {"ba", 1}, {NULL}}},
+	{"(.|a)*", {{"a", 1},{"abc", 1}, {"b", 1 }, {"ba", 1}, {NULL}}},
+	{"(b|a)*", {{"a", 1},{"cab", 0}, {"b", 1 }, {"ab", 1}, {"cacb", 0}, {NULL}}},
+	{"c(b|a)*", {{"ca", 1},{"cab", 1}, {"b", 0 }, {"bb", 0}, {"cacb", 0}, {NULL}}},
+	{"ab|b", {{"ab", 1}, {"b", 1}, {"abb", 0}, {NULL}}},
+	{"a|b|c", {{"a", 1}, {"b", 1}, {"c", 1},{"d",0},  {NULL}}},
+	{"((a|b)|c)d", {{"ad", 1}, {"bd", 1}, {"cd", 1},{"a",0},  {NULL}}},
+
+	// BLOCK
+	{"(abb", {{"abb", 1}, {"ab", 0}, {"b", 0}, {NULL}}},
+	{"abb)", {{"abb", 1}, {"ab", 0}, {"b", 0}, {NULL}}},
+	{"a(bb", {{"abb", 1}, {"ab", 0}, {"b", 0}, {NULL}}},
+	{"ab)b", {{"abb", 1}, {"ab", 0}, {"b", 0}, {NULL}}},
+	{"(ab)b", {{"abb", 1}, {"ab", 0}, {"b", 0}, {NULL}}},
+	{"(ab)+", {{"ab", 1}, {"abab", 1}, {"b", 0}, {NULL}}},
+	{"(a(bc)*)", {{"a", 1}, {"abc", 1}, {"abb", 0}, {NULL}}},
+	{"(ab|cd)", {{"ab", 1}, {"cd", 1}, {"abcd", 0}, {NULL}}},
+
+	// REDOS
 	{"a*b?a*", {{"aaa", 1}, {"aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaba", 1}, {NULL}}},
-	{"aa*+", {{"aaaa", 1}, {NULL}}},
-	{"a*+", {{"a", 1}, {NULL}}},
 } ;
 
 
@@ -60,6 +79,11 @@ int main(void)
 		printf("\n\t%s\n", t.regex);
 		nfa = rgx_compile(NULL, t.regex, 1);
 
+		if (!nfa ){
+			printf("Malformed\n");
+			continue;
+		}
+
 		for( struct match * m = t.matches; m->s; ++m)
 		{
 			int res;