blob: f552096f334d7134ff0d62a975cd4057d0b4eefb
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 | //config:config TR |
22 | //config: bool "tr" |
23 | //config: default y |
24 | //config: help |
25 | //config: tr is used to squeeze, and/or delete characters from standard |
26 | //config: input, writing to standard output. |
27 | //config: |
28 | //config:config FEATURE_TR_CLASSES |
29 | //config: bool "Enable character classes (such as [:upper:])" |
30 | //config: default y |
31 | //config: depends on TR |
32 | //config: help |
33 | //config: Enable character classes, enabling commands such as: |
34 | //config: tr [:upper:] [:lower:] to convert input into lowercase. |
35 | //config: |
36 | //config:config FEATURE_TR_EQUIV |
37 | //config: bool "Enable equivalence classes" |
38 | //config: default y |
39 | //config: depends on TR |
40 | //config: help |
41 | //config: Enable equivalence classes, which essentially add the enclosed |
42 | //config: character to the current set. For instance, tr [=a=] xyz would |
43 | //config: replace all instances of 'a' with 'xyz'. This option is mainly |
44 | //config: useful for cases when no other way of expressing a character |
45 | //config: is possible. |
46 | |
47 | //applet:IF_TR(APPLET(tr, BB_DIR_USR_BIN, BB_SUID_DROP)) |
48 | |
49 | //kbuild:lib-$(CONFIG_TR) += tr.o |
50 | |
51 | //usage:#define tr_trivial_usage |
52 | //usage: "[-cds] STRING1 [STRING2]" |
53 | //usage:#define tr_full_usage "\n\n" |
54 | //usage: "Translate, squeeze, or delete characters from stdin, writing to stdout\n" |
55 | //usage: "\n -c Take complement of STRING1" |
56 | //usage: "\n -d Delete input characters coded STRING1" |
57 | //usage: "\n -s Squeeze multiple output characters of STRING2 into one character" |
58 | //usage: |
59 | //usage:#define tr_example_usage |
60 | //usage: "$ echo \"gdkkn vnqkc\" | tr [a-y] [b-z]\n" |
61 | //usage: "hello world\n" |
62 | |
63 | #include "libbb.h" |
64 | |
65 | enum { |
66 | ASCII = 256, |
67 | /* string buffer needs to be at least as big as the whole "alphabet". |
68 | * BUFSIZ == ASCII is ok, but we will realloc in expand |
69 | * even for smallest patterns, let's avoid that by using *2: |
70 | */ |
71 | TR_BUFSIZ = (BUFSIZ > ASCII*2) ? BUFSIZ : ASCII*2, |
72 | }; |
73 | |
74 | static void map(char *pvector, |
75 | char *string1, unsigned string1_len, |
76 | char *string2, unsigned string2_len) |
77 | { |
78 | char last = '0'; |
79 | unsigned i, j; |
80 | |
81 | for (j = 0, i = 0; i < string1_len; i++) { |
82 | if (string2_len <= j) |
83 | pvector[(unsigned char)(string1[i])] = last; |
84 | else |
85 | pvector[(unsigned char)(string1[i])] = last = string2[j++]; |
86 | } |
87 | } |
88 | |
89 | /* supported constructs: |
90 | * Ranges, e.g., 0-9 ==> 0123456789 |
91 | * Escapes, e.g., \a ==> Control-G |
92 | * Character classes, e.g. [:upper:] ==> A...Z |
93 | * Equiv classess, e.g. [=A=] ==> A (hmmmmmmm?) |
94 | * not supported: |
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(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 | const char *z; |
117 | arg++; |
118 | z = arg; |
119 | ac = bb_process_escape_sequence(&z); |
120 | arg = (char *)z; |
121 | arg--; |
122 | *arg = ac; |
123 | /* |
124 | * fall through, there may be a range. |
125 | * If not, current char will be treated anyway. |
126 | */ |
127 | } |
128 | if (arg[1] == '-') { /* "0-9..." */ |
129 | ac = arg[2]; |
130 | if (ac == '\0') { /* "0-": copy verbatim */ |
131 | buffer[pos++] = *arg++; /* copy '0' */ |
132 | continue; /* next iter will copy '-' and stop */ |
133 | } |
134 | i = (unsigned char) *arg; |
135 | arg += 3; /* skip 0-9 or 0-\ */ |
136 | if (ac == '\\') { |
137 | const char *z; |
138 | z = arg; |
139 | ac = bb_process_escape_sequence(&z); |
140 | arg = (char *)z; |
141 | } |
142 | while (i <= ac) /* ok: i is unsigned _int_ */ |
143 | buffer[pos++] = i++; |
144 | continue; |
145 | } |
146 | if ((ENABLE_FEATURE_TR_CLASSES || ENABLE_FEATURE_TR_EQUIV) |
147 | && *arg == '[' |
148 | ) { |
149 | arg++; |
150 | i = (unsigned char) *arg++; |
151 | /* "[xyz...". i=x, arg points to y */ |
152 | if (ENABLE_FEATURE_TR_CLASSES && i == ':') { /* [:class:] */ |
153 | #define CLO ":]\0" |
154 | static const char classes[] ALIGN1 = |
155 | "alpha"CLO "alnum"CLO "digit"CLO |
156 | "lower"CLO "upper"CLO "space"CLO |
157 | "blank"CLO "punct"CLO "cntrl"CLO |
158 | "xdigit"CLO; |
159 | enum { |
160 | CLASS_invalid = 0, /* we increment the retval */ |
161 | CLASS_alpha = 1, |
162 | CLASS_alnum = 2, |
163 | CLASS_digit = 3, |
164 | CLASS_lower = 4, |
165 | CLASS_upper = 5, |
166 | CLASS_space = 6, |
167 | CLASS_blank = 7, |
168 | CLASS_punct = 8, |
169 | CLASS_cntrl = 9, |
170 | CLASS_xdigit = 10, |
171 | //CLASS_graph = 11, |
172 | //CLASS_print = 12, |
173 | }; |
174 | smalluint j; |
175 | char *tmp; |
176 | |
177 | /* xdigit needs 8, not 7 */ |
178 | i = 7 + (arg[0] == 'x'); |
179 | tmp = xstrndup(arg, i); |
180 | j = index_in_strings(classes, tmp) + 1; |
181 | free(tmp); |
182 | |
183 | if (j == CLASS_invalid) |
184 | goto skip_bracket; |
185 | |
186 | arg += i; |
187 | if (j == CLASS_alnum || j == CLASS_digit || j == CLASS_xdigit) { |
188 | for (i = '0'; i <= '9'; i++) |
189 | buffer[pos++] = i; |
190 | } |
191 | if (j == CLASS_alpha || j == CLASS_alnum || j == CLASS_upper) { |
192 | for (i = 'A'; i <= 'Z'; i++) |
193 | buffer[pos++] = i; |
194 | } |
195 | if (j == CLASS_alpha || j == CLASS_alnum || j == CLASS_lower) { |
196 | for (i = 'a'; i <= 'z'; i++) |
197 | buffer[pos++] = i; |
198 | } |
199 | if (j == CLASS_space || j == CLASS_blank) { |
200 | buffer[pos++] = '\t'; |
201 | if (j == CLASS_space) { |
202 | buffer[pos++] = '\n'; |
203 | buffer[pos++] = '\v'; |
204 | buffer[pos++] = '\f'; |
205 | buffer[pos++] = '\r'; |
206 | } |
207 | buffer[pos++] = ' '; |
208 | } |
209 | if (j == CLASS_punct || j == CLASS_cntrl) { |
210 | for (i = '\0'; i < ASCII; i++) { |
211 | if ((j == CLASS_punct && isprint_asciionly(i) && !isalnum(i) && !isspace(i)) |
212 | || (j == CLASS_cntrl && iscntrl(i)) |
213 | ) { |
214 | buffer[pos++] = i; |
215 | } |
216 | } |
217 | } |
218 | if (j == CLASS_xdigit) { |
219 | for (i = 'A'; i <= 'F'; i++) { |
220 | buffer[pos + 6] = i | 0x20; |
221 | buffer[pos++] = i; |
222 | } |
223 | pos += 6; |
224 | } |
225 | continue; |
226 | } |
227 | /* "[xyz...", i=x, arg points to y */ |
228 | if (ENABLE_FEATURE_TR_EQUIV && i == '=') { /* [=CHAR=] */ |
229 | buffer[pos++] = *arg; /* copy CHAR */ |
230 | if (!arg[0] || arg[1] != '=' || arg[2] != ']') |
231 | bb_show_usage(); |
232 | arg += 3; /* skip CHAR=] */ |
233 | continue; |
234 | } |
235 | /* The rest of "[xyz..." cases is treated as normal |
236 | * string, "[" has no special meaning here: |
237 | * tr "[a-z]" "[A-Z]" can be written as tr "a-z" "A-Z", |
238 | * also try tr "[a-z]" "_A-Z+" and you'll see that |
239 | * [] is not special here. |
240 | */ |
241 | skip_bracket: |
242 | arg -= 2; /* points to "[" in "[xyz..." */ |
243 | } |
244 | buffer[pos++] = *arg++; |
245 | } |
246 | return pos; |
247 | } |
248 | |
249 | /* NB: buffer is guaranteed to be at least TR_BUFSIZE |
250 | * (which is >= ASCII) big. |
251 | */ |
252 | static int complement(char *buffer, int buffer_len) |
253 | { |
254 | int len; |
255 | char conv[ASCII]; |
256 | unsigned char ch; |
257 | |
258 | len = 0; |
259 | ch = '\0'; |
260 | while (1) { |
261 | if (memchr(buffer, ch, buffer_len) == NULL) |
262 | conv[len++] = ch; |
263 | if (++ch == '\0') |
264 | break; |
265 | } |
266 | memcpy(buffer, conv, len); |
267 | return len; |
268 | } |
269 | |
270 | int tr_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE; |
271 | int tr_main(int argc UNUSED_PARAM, char **argv) |
272 | { |
273 | int i; |
274 | smalluint opts; |
275 | ssize_t read_chars; |
276 | size_t in_index, out_index; |
277 | unsigned last = UCHAR_MAX + 1; /* not equal to any char */ |
278 | unsigned char coded, c; |
279 | char *str1 = xmalloc(TR_BUFSIZ); |
280 | char *str2 = xmalloc(TR_BUFSIZ); |
281 | int str2_length; |
282 | int str1_length; |
283 | char *vector = xzalloc(ASCII * 3); |
284 | char *invec = vector + ASCII; |
285 | char *outvec = vector + ASCII * 2; |
286 | |
287 | #define TR_OPT_complement (3 << 0) |
288 | #define TR_OPT_delete (1 << 2) |
289 | #define TR_OPT_squeeze_reps (1 << 3) |
290 | |
291 | for (i = 0; i < ASCII; i++) { |
292 | vector[i] = i; |
293 | /*invec[i] = outvec[i] = FALSE; - done by xzalloc */ |
294 | } |
295 | |
296 | /* -C/-c difference is that -C complements "characters", |
297 | * and -c complements "values" (binary bytes I guess). |
298 | * In POSIX locale, these are the same. |
299 | */ |
300 | |
301 | opt_complementary = "-1"; |
302 | opts = getopt32(argv, "+Ccds"); /* '+': stop at first non-option */ |
303 | argv += optind; |
304 | |
305 | str1_length = expand(*argv++, &str1); |
306 | str2_length = 0; |
307 | if (opts & TR_OPT_complement) |
308 | str1_length = complement(str1, str1_length); |
309 | if (*argv) { |
310 | if (argv[0][0] == '\0') |
311 | bb_error_msg_and_die("STRING2 cannot be empty"); |
312 | str2_length = expand(*argv, &str2); |
313 | map(vector, str1, str1_length, |
314 | str2, str2_length); |
315 | } |
316 | for (i = 0; i < str1_length; i++) |
317 | invec[(unsigned char)(str1[i])] = TRUE; |
318 | for (i = 0; i < str2_length; i++) |
319 | outvec[(unsigned char)(str2[i])] = TRUE; |
320 | |
321 | goto start_from; |
322 | |
323 | /* In this loop, str1 space is reused as input buffer, |
324 | * str2 - as output one. */ |
325 | for (;;) { |
326 | /* If we're out of input, flush output and read more input. */ |
327 | if ((ssize_t)in_index == read_chars) { |
328 | if (out_index) { |
329 | xwrite(STDOUT_FILENO, str2, out_index); |
330 | start_from: |
331 | out_index = 0; |
332 | } |
333 | read_chars = safe_read(STDIN_FILENO, str1, TR_BUFSIZ); |
334 | if (read_chars <= 0) { |
335 | if (read_chars < 0) |
336 | bb_perror_msg_and_die(bb_msg_read_error); |
337 | break; |
338 | } |
339 | in_index = 0; |
340 | } |
341 | c = str1[in_index++]; |
342 | if ((opts & TR_OPT_delete) && invec[c]) |
343 | continue; |
344 | coded = vector[c]; |
345 | if ((opts & TR_OPT_squeeze_reps) && last == coded |
346 | && (invec[c] || outvec[coded]) |
347 | ) { |
348 | continue; |
349 | } |
350 | str2[out_index++] = last = coded; |
351 | } |
352 | |
353 | if (ENABLE_FEATURE_CLEAN_UP) { |
354 | free(vector); |
355 | free(str2); |
356 | free(str1); |
357 | } |
358 | |
359 | return EXIT_SUCCESS; |
360 | } |
361 |