1 #include <u.h>
2 #include <libc.h>
3 #include <bio.h>
4 #include <regexp.h>
5 #include "dfa.h"
6 
7 /***
8  * Regular expression for matching.
9  */
10 
11 char *ignore[] = 
12 {
13 	/* HTML that isn't A, IMG, or FONT */
14 	/* Must have a space somewhere to avoid catching <email@address> */
15 	"<[ 	\n\r]*("
16 		"[^aif]|"
17 		"a[^> \t\r\n]|"
18 		"i[^mM \t\r\n]|"
19 		"im[^gG \t\r\n]|"
20 		"img[^> \t\r\n]|"
21 		"f[^oO \t\r\n]|"
22 		"fo[^Nn \t\r\n]|"
23 		"fon[^tT \t\r\n]|"
24 		"font[^> \r\t\n]"
25 	")[^>]*[ \t\n\r][^>]*>",
26 	"<[ 	\n\r]*("
27 		"i|im|f|fo|fon"
28 	")[ \t\r\n][^>]*>",
29 
30 	/* ignore html comments */
31 	"<!--([^\\-]|-[^\\-]|--[^>]|\n)*-->",
32 
33 	/* random mail strings */
34 	"^message-id:.*\n([ 	].*\n)*",
35 	"^in-reply-to:.*\n([ 	].*\n)*",
36 	"^references:.*\n([ 	].*\n)*",
37 	"^date:.*\n([ 	].*\n)*",
38 	"^delivery-date:.*\n([ 	].*\n)*",
39 	"e?smtp id .*",
40 	"^	id.*",
41 	"boundary=.*",
42 	"name=\"",
43 	"filename=\"",
44 	"news:<[^>]+>",
45 	"^--[^ 	]*$",
46 
47 	/* base64 encoding */
48 	"^[0-9a-zA-Z+\\-=/]+$",
49 
50 	/* uu encoding */
51 	"^[!-Z]+$",
52 
53 	/* little things */
54 	".",
55 	"\n",
56 };
57 
58 char *keywords[] =
59 {
60 	"([a-zA-Z'`$!¡-￿]|[0-9]([.,][0-9])*)+",
61 };
62 
63 int debug;
64 
65 Dreprog*
66 dregcomp(char *buf)
67 {
68 	Reprog *r;
69 	Dreprog *d;
70 
71 	if(debug)
72 		print(">>> '%s'\n", buf);
73 
74 	r = regcomp(buf);
75 	if(r == nil)
76 		sysfatal("regcomp");
77 	d = dregcvt(r);
78 	if(d == nil)
79 		sysfatal("dregcomp");
80 	free(r);
81 	return d;
82 }
83 
84 char*
85 strcpycase(char *d, char *s)
86 {
87 	int cc, esc;
88 
89 	cc = 0;
90 	esc = 0;
91 	while(*s){
92 		if(*s == '[')
93 			cc++;
94 		if(*s == ']')
95 			cc--;
96 		if(!cc && 'a' <= *s && *s <= 'z'){
97 			*d++ = '[';
98 			*d++ = *s;
99 			*d++ = *s+'A'-'a';
100 			*d++ = ']';
101 		}else
102 			*d++ = *s;
103 		if(*s == '\\')
104 			esc++;
105 		else if(esc)
106 			esc--;
107 		s++;
108 	}
109 	return d;
110 }
111 
112 void
113 regerror(char *msg)
114 {
115 	sysfatal("regerror: %s", msg);
116 }
117 
118 void
119 buildre(Dreprog *re[3])
120 {
121 	int i;
122 	static char buf[16384], *s;
123 
124 	re[0] = dregcomp("^From ");
125 	
126 	s = buf;
127 	for(i=0; i<nelem(keywords); i++){
128 		if(i != 0)
129 			*s++ = '|';
130 		s = strcpycase(s, keywords[i]);
131 	}
132 	*s = 0;
133 	re[1] = dregcomp(buf);
134 
135 	s = buf;
136 	for(i=0; i<nelem(ignore); i++){
137 		if(i != 0)
138 			*s++ = '|';
139 		s = strcpycase(s, ignore[i]);
140 	}
141 	*s = 0;
142 	re[2] = dregcomp(buf);
143 }
144 
145 void
146 usage(void)
147 {
148 	fprint(2, "usage: regen [-d]\n");
149 	exits("usage");
150 }
151 
152 void
153 main(int argc, char **argv)
154 {
155 	Dreprog *re[3];
156 	Biobuf b;
157 
158 	ARGBEGIN{
159 	default:
160 		usage();
161 	case 'd':
162 		debug = 1;
163 	}ARGEND
164 
165 	if(argc != 0)
166 		usage();
167 
168 	buildre(re);
169 	Binit(&b, 1, OWRITE);
170 	Bprintdfa(&b, re[0]);
171 	Bprintdfa(&b, re[1]);
172 	Bprintdfa(&b, re[2]);
173 	exits(0);
174 }
175 
176