blob: e67948a36c88a148f22cd6a5af97e75159cfefa1
1 | /* vi: set sw=4 ts=4: */ |
2 | /* |
3 | * Mini tr implementation for busybox |
4 | * |
5 | ** Copyright (c) 1987,1997, Prentice Hall All rights reserved. |
6 | * |
7 | * The name of Prentice Hall may not be used to endorse or promote |
8 | * products derived from this software without specific prior |
9 | * written permission. |
10 | * |
11 | * Copyright (c) Michiel Huisjes |
12 | * |
13 | * This version of tr is adapted from Minix tr and was modified |
14 | * by Erik Andersen <andersen@codepoet.org> to be used in busybox. |
15 | * |
16 | * Licensed under GPLv2 or later, see file LICENSE in this source tree. |
17 | */ |
18 | /* http://www.opengroup.org/onlinepubs/009695399/utilities/tr.html |
19 | * TODO: graph, print |
20 | */ |
21 | |
22 | //kbuild:lib-$(CONFIG_TR) += tr.o |
23 | |
24 | //config:config TR |
25 | //config: bool "tr" |
26 | //config: default y |
27 | //config: help |
28 | //config: tr is used to squeeze, and/or delete characters from standard |
29 | //config: input, writing to standard output. |
30 | //config: |
31 | //config:config FEATURE_TR_CLASSES |
32 | //config: bool "Enable character classes (such as [:upper:])" |
33 | //config: default y |
34 | //config: depends on TR |
35 | //config: help |
36 | //config: Enable character classes, enabling commands such as: |
37 | //config: tr [:upper:] [:lower:] to convert input into lowercase. |
38 | //config: |
39 | //config:config FEATURE_TR_EQUIV |
40 | //config: bool "Enable equivalence classes" |
41 | //config: default y |
42 | //config: depends on TR |
43 | //config: help |
44 | //config: Enable equivalence classes, which essentially add the enclosed |
45 | //config: character to the current set. For instance, tr [=a=] xyz would |
46 | //config: replace all instances of 'a' with 'xyz'. This option is mainly |
47 | //config: useful for cases when no other way of expressing a character |
48 | //config: is possible. |
49 | |
50 | //usage:#define tr_trivial_usage |
51 | //usage: "[-cds] STRING1 [STRING2]" |
52 | //usage:#define tr_full_usage "\n\n" |
53 | //usage: "Translate, squeeze, or delete characters from stdin, writing to stdout\n" |
54 | //usage: "\n -c Take complement of STRING1" |
55 | //usage: "\n -d Delete input characters coded STRING1" |
56 | //usage: "\n -s Squeeze multiple output characters of STRING2 into one character" |
57 | //usage: |
58 | //usage:#define tr_example_usage |
59 | //usage: "$ echo \"gdkkn vnqkc\" | tr [a-y] [b-z]\n" |
60 | //usage: "hello world\n" |
61 | |
62 | #include "libbb.h" |
63 | |
64 | enum { |
65 | ASCII = 256, |
66 | /* string buffer needs to be at least as big as the whole "alphabet". |
67 | * BUFSIZ == ASCII is ok, but we will realloc in expand |
68 | * even for smallest patterns, let's avoid that by using *2: |
69 | */ |
70 | TR_BUFSIZ = (BUFSIZ > ASCII*2) ? BUFSIZ : ASCII*2, |
71 | }; |
72 | |
73 | static void map(char *pvector, |
74 | char *string1, unsigned string1_len, |
75 | char *string2, unsigned string2_len) |
76 | { |
77 | char last = '0'; |
78 | unsigned i, j; |
79 | |
80 | for (j = 0, i = 0; i < string1_len; i++) { |
81 | if (string2_len <= j) |
82 | pvector[(unsigned char)(string1[i])] = last; |
83 | else |
84 | pvector[(unsigned char)(string1[i])] = last = string2[j++]; |
85 | } |
86 | } |
87 | |
88 | /* supported constructs: |
89 | * Ranges, e.g., 0-9 ==> 0123456789 |
90 | * Escapes, e.g., \a ==> Control-G |
91 | * Character classes, e.g. [:upper:] ==> A...Z |
92 | * Equiv classess, e.g. [=A=] ==> A (hmmmmmmm?) |
93 | * not supported: |
94 | * \ooo-\ooo - octal ranges |
95 | * [x*N] - repeat char x N times |
96 | * [x*] - repeat char x until it fills STRING2: |
97 | * # echo qwe123 | /usr/bin/tr 123456789 '[d]' |
98 | * qwe[d] |
99 | * # echo qwe123 | /usr/bin/tr 123456789 '[d*]' |
100 | * qweddd |
101 | */ |
102 | static unsigned expand(const char *arg, char **buffer_p) |
103 | { |
104 | char *buffer = *buffer_p; |
105 | unsigned pos = 0; |
106 | unsigned size = TR_BUFSIZ; |
107 | unsigned i; /* can't be unsigned char: must be able to hold 256 */ |
108 | unsigned char ac; |
109 | |
110 | while (*arg) { |
111 | if (pos + ASCII > size) { |
112 | size += ASCII; |
113 | *buffer_p = buffer = xrealloc(buffer, size); |
114 | } |
115 | if (*arg == '\\') { |
116 | arg++; |
117 | buffer[pos++] = bb_process_escape_sequence(&arg); |
118 | continue; |
119 | } |
120 | if (arg[1] == '-') { /* "0-9..." */ |
121 | ac = arg[2]; |
122 | if (ac == '\0') { /* "0-": copy verbatim */ |
123 | buffer[pos++] = *arg++; /* copy '0' */ |
124 | continue; /* next iter will copy '-' and stop */ |
125 | } |
126 | i = (unsigned char) *arg; |
127 | while (i <= ac) /* ok: i is unsigned _int_ */ |
128 | buffer[pos++] = i++; |
129 | arg += 3; /* skip 0-9 */ |
130 | continue; |
131 | } |
132 | if ((ENABLE_FEATURE_TR_CLASSES || ENABLE_FEATURE_TR_EQUIV) |
133 | && *arg == '[' |
134 | ) { |
135 | arg++; |
136 | i = (unsigned char) *arg++; |
137 | /* "[xyz...". i=x, arg points to y */ |
138 | if (ENABLE_FEATURE_TR_CLASSES && i == ':') { /* [:class:] */ |
139 | #define CLO ":]\0" |
140 | static const char classes[] ALIGN1 = |
141 | "alpha"CLO "alnum"CLO "digit"CLO |
142 | "lower"CLO "upper"CLO "space"CLO |
143 | "blank"CLO "punct"CLO "cntrl"CLO |
144 | "xdigit"CLO; |
145 | enum { |
146 | CLASS_invalid = 0, /* we increment the retval */ |
147 | CLASS_alpha = 1, |
148 | CLASS_alnum = 2, |
149 | CLASS_digit = 3, |
150 | CLASS_lower = 4, |
151 | CLASS_upper = 5, |
152 | CLASS_space = 6, |
153 | CLASS_blank = 7, |
154 | CLASS_punct = 8, |
155 | CLASS_cntrl = 9, |
156 | CLASS_xdigit = 10, |
157 | //CLASS_graph = 11, |
158 | //CLASS_print = 12, |
159 | }; |
160 | smalluint j; |
161 | char *tmp; |
162 | |
163 | /* xdigit needs 8, not 7 */ |
164 | i = 7 + (arg[0] == 'x'); |
165 | tmp = xstrndup(arg, i); |
166 | j = index_in_strings(classes, tmp) + 1; |
167 | free(tmp); |
168 | |
169 | if (j == CLASS_invalid) |
170 | goto skip_bracket; |
171 | |
172 | arg += i; |
173 | if (j == CLASS_alnum || j == CLASS_digit || j == CLASS_xdigit) { |
174 | for (i = '0'; i <= '9'; i++) |
175 | buffer[pos++] = i; |
176 | } |
177 | if (j == CLASS_alpha || j == CLASS_alnum || j == CLASS_upper) { |
178 | for (i = 'A'; i <= 'Z'; i++) |
179 | buffer[pos++] = i; |
180 | } |
181 | if (j == CLASS_alpha || j == CLASS_alnum || j == CLASS_lower) { |
182 | for (i = 'a'; i <= 'z'; i++) |
183 | buffer[pos++] = i; |
184 | } |
185 | if (j == CLASS_space || j == CLASS_blank) { |
186 | buffer[pos++] = '\t'; |
187 | if (j == CLASS_space) { |
188 | buffer[pos++] = '\n'; |
189 | buffer[pos++] = '\v'; |
190 | buffer[pos++] = '\f'; |
191 | buffer[pos++] = '\r'; |
192 | } |
193 | buffer[pos++] = ' '; |
194 | } |
195 | if (j == CLASS_punct || j == CLASS_cntrl) { |
196 | for (i = '\0'; i < ASCII; i++) { |
197 | if ((j == CLASS_punct && isprint_asciionly(i) && !isalnum(i) && !isspace(i)) |
198 | || (j == CLASS_cntrl && iscntrl(i)) |
199 | ) { |
200 | buffer[pos++] = i; |
201 | } |
202 | } |
203 | } |
204 | if (j == CLASS_xdigit) { |
205 | for (i = 'A'; i <= 'F'; i++) { |
206 | buffer[pos + 6] = i | 0x20; |
207 | buffer[pos++] = i; |
208 | } |
209 | pos += 6; |
210 | } |
211 | continue; |
212 | } |
213 | /* "[xyz...", i=x, arg points to y */ |
214 | if (ENABLE_FEATURE_TR_EQUIV && i == '=') { /* [=CHAR=] */ |
215 | buffer[pos++] = *arg; /* copy CHAR */ |
216 | if (!arg[0] || arg[1] != '=' || arg[2] != ']') |
217 | bb_show_usage(); |
218 | arg += 3; /* skip CHAR=] */ |
219 | continue; |
220 | } |
221 | /* The rest of "[xyz..." cases is treated as normal |
222 | * string, "[" has no special meaning here: |
223 | * tr "[a-z]" "[A-Z]" can be written as tr "a-z" "A-Z", |
224 | * also try tr "[a-z]" "_A-Z+" and you'll see that |
225 | * [] is not special here. |
226 | */ |
227 | skip_bracket: |
228 | arg -= 2; /* points to "[" in "[xyz..." */ |
229 | } |
230 | buffer[pos++] = *arg++; |
231 | } |
232 | return pos; |
233 | } |
234 | |
235 | /* NB: buffer is guaranteed to be at least TR_BUFSIZE |
236 | * (which is >= ASCII) big. |
237 | */ |
238 | static int complement(char *buffer, int buffer_len) |
239 | { |
240 | int len; |
241 | char conv[ASCII]; |
242 | unsigned char ch; |
243 | |
244 | len = 0; |
245 | ch = '\0'; |
246 | while (1) { |
247 | if (memchr(buffer, ch, buffer_len) == NULL) |
248 | conv[len++] = ch; |
249 | if (++ch == '\0') |
250 | break; |
251 | } |
252 | memcpy(buffer, conv, len); |
253 | return len; |
254 | } |
255 | |
256 | int tr_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; |
257 | int tr_main(int argc UNUSED_PARAM, char **argv) |
258 | { |
259 | int i; |
260 | smalluint opts; |
261 | ssize_t read_chars; |
262 | size_t in_index, out_index; |
263 | unsigned last = UCHAR_MAX + 1; /* not equal to any char */ |
264 | unsigned char coded, c; |
265 | char *str1 = xmalloc(TR_BUFSIZ); |
266 | char *str2 = xmalloc(TR_BUFSIZ); |
267 | int str2_length; |
268 | int str1_length; |
269 | char *vector = xzalloc(ASCII * 3); |
270 | char *invec = vector + ASCII; |
271 | char *outvec = vector + ASCII * 2; |
272 | |
273 | #define TR_OPT_complement (3 << 0) |
274 | #define TR_OPT_delete (1 << 2) |
275 | #define TR_OPT_squeeze_reps (1 << 3) |
276 | |
277 | for (i = 0; i < ASCII; i++) { |
278 | vector[i] = i; |
279 | /*invec[i] = outvec[i] = FALSE; - done by xzalloc */ |
280 | } |
281 | |
282 | /* -C/-c difference is that -C complements "characters", |
283 | * and -c complements "values" (binary bytes I guess). |
284 | * In POSIX locale, these are the same. |
285 | */ |
286 | |
287 | opt_complementary = "-1"; |
288 | opts = getopt32(argv, "+Ccds"); /* '+': stop at first non-option */ |
289 | argv += optind; |
290 | |
291 | str1_length = expand(*argv++, &str1); |
292 | str2_length = 0; |
293 | if (opts & TR_OPT_complement) |
294 | str1_length = complement(str1, str1_length); |
295 | if (*argv) { |
296 | if (argv[0][0] == '\0') |
297 | bb_error_msg_and_die("STRING2 cannot be empty"); |
298 | str2_length = expand(*argv, &str2); |
299 | map(vector, str1, str1_length, |
300 | str2, str2_length); |
301 | } |
302 | for (i = 0; i < str1_length; i++) |
303 | invec[(unsigned char)(str1[i])] = TRUE; |
304 | for (i = 0; i < str2_length; i++) |
305 | outvec[(unsigned char)(str2[i])] = TRUE; |
306 | |
307 | goto start_from; |
308 | |
309 | /* In this loop, str1 space is reused as input buffer, |
310 | * str2 - as output one. */ |
311 | for (;;) { |
312 | /* If we're out of input, flush output and read more input. */ |
313 | if ((ssize_t)in_index == read_chars) { |
314 | if (out_index) { |
315 | xwrite(STDOUT_FILENO, str2, out_index); |
316 | start_from: |
317 | out_index = 0; |
318 | } |
319 | read_chars = safe_read(STDIN_FILENO, str1, TR_BUFSIZ); |
320 | if (read_chars <= 0) { |
321 | if (read_chars < 0) |
322 | bb_perror_msg_and_die(bb_msg_read_error); |
323 | break; |
324 | } |
325 | in_index = 0; |
326 | } |
327 | c = str1[in_index++]; |
328 | if ((opts & TR_OPT_delete) && invec[c]) |
329 | continue; |
330 | coded = vector[c]; |
331 | if ((opts & TR_OPT_squeeze_reps) && last == coded |
332 | && (invec[c] || outvec[coded]) |
333 | ) { |
334 | continue; |
335 | } |
336 | str2[out_index++] = last = coded; |
337 | } |
338 | |
339 | if (ENABLE_FEATURE_CLEAN_UP) { |
340 | free(vector); |
341 | free(str2); |
342 | free(str1); |
343 | } |
344 | |
345 | return EXIT_SUCCESS; |
346 | } |
347 |