/*
* Mini tr implementation for busybox
*
- ** Copyright (c) 1987,1997, Prentice Hall All rights reserved.
+ * Copyright (c) 1987,1997, Prentice Hall All rights reserved.
*
* The name of Prentice Hall may not be used to endorse or promote
* products derived from this software without specific prior
* This version of tr is adapted from Minix tr and was modified
* by Erik Andersen <andersen@codepoet.org> to be used in busybox.
*
- * Licensed under GPLv2 or later, see file LICENSE in this tarball for details.
+ * Licensed under GPLv2 or later, see file LICENSE in this source tree.
*/
/* http://www.opengroup.org/onlinepubs/009695399/utilities/tr.html
- * TODO: xdigit, graph, print
+ * TODO: graph, print
*/
-#include "busybox.h"
+//config:config TR
+//config: bool "tr (5.5 kb)"
+//config: default y
+//config: help
+//config: tr is used to squeeze, and/or delete characters from standard
+//config: input, writing to standard output.
+//config:
+//config:config FEATURE_TR_CLASSES
+//config: bool "Enable character classes (such as [:upper:])"
+//config: default y
+//config: depends on TR
+//config: help
+//config: Enable character classes, enabling commands such as:
+//config: tr [:upper:] [:lower:] to convert input into lowercase.
+//config:
+//config:config FEATURE_TR_EQUIV
+//config: bool "Enable equivalence classes"
+//config: default y
+//config: depends on TR
+//config: help
+//config: Enable equivalence classes, which essentially add the enclosed
+//config: character to the current set. For instance, tr [=a=] xyz would
+//config: replace all instances of 'a' with 'xyz'. This option is mainly
+//config: useful for cases when no other way of expressing a character
+//config: is possible.
-#define ASCII 0377
+//applet:IF_TR(APPLET(tr, BB_DIR_USR_BIN, BB_SUID_DROP))
-#define TR_OPT_complement (1<<0)
-#define TR_OPT_delete (1<<1)
-#define TR_OPT_squeeze_reps (1<<2)
-/* some "globals" shared across this file */
-/* these last are pointers to static buffers declared in tr_main */
-static char *poutput, *pvector, *pinvec, *poutvec;
+//kbuild:lib-$(CONFIG_TR) += tr.o
-static void ATTRIBUTE_NORETURN convert(const smalluint flags)
-{
- size_t read_chars = 0, in_index = 0, out_index = 0, c, coded, last = -1;
+//usage:#define tr_trivial_usage
+//usage: "[-cds] STRING1 [STRING2]"
+//usage:#define tr_full_usage "\n\n"
+//usage: "Translate, squeeze, or delete characters from stdin, writing to stdout\n"
+//usage: "\n -c Take complement of STRING1"
+//usage: "\n -d Delete input characters coded STRING1"
+//usage: "\n -s Squeeze multiple output characters of STRING2 into one character"
+//usage:
+//usage:#define tr_example_usage
+//usage: "$ echo \"gdkkn vnqkc\" | tr [a-y] [b-z]\n"
+//usage: "hello world\n"
- for (;;) {
- /* If we're out of input, flush output and read more input. */
- if (in_index == read_chars) {
- if (out_index) {
- xwrite(STDOUT_FILENO, (char *)poutput, out_index);
- out_index = 0;
- }
- if ((read_chars = read(STDIN_FILENO, bb_common_bufsiz1, BUFSIZ)) <= 0) {
- if (write(STDOUT_FILENO, (char *)poutput, out_index) != out_index)
- bb_perror_msg(bb_msg_write_error);
- exit(EXIT_SUCCESS);
- }
- in_index = 0;
- }
- c = bb_common_bufsiz1[in_index++];
- coded = pvector[c];
- if ((flags & TR_OPT_delete) && pinvec[c])
- continue;
- if ((flags & TR_OPT_squeeze_reps) && last == coded &&
- (pinvec[c] || poutvec[coded]))
- continue;
- poutput[out_index++] = last = coded;
- }
- /* NOTREACHED */
-}
+#include "libbb.h"
+
+enum {
+ ASCII = 256,
+ /* string buffer needs to be at least as big as the whole "alphabet".
+ * BUFSIZ == ASCII is ok, but we will realloc in expand
+ * even for smallest patterns, let's avoid that by using *2:
+ */
+ TR_BUFSIZ = (BUFSIZ > ASCII*2) ? BUFSIZ : ASCII*2,
+};
-static void map(unsigned char *string1, unsigned int string1_len,
- unsigned char *string2, unsigned int string2_len)
+static void map(char *pvector,
+ char *string1, unsigned string1_len,
+ char *string2, unsigned string2_len)
{
char last = '0';
- unsigned int i, j;
+ unsigned i, j;
for (j = 0, i = 0; i < string1_len; i++) {
if (string2_len <= j)
- pvector[string1[i]] = last;
+ pvector[(unsigned char)(string1[i])] = last;
else
- pvector[string1[i]] = last = string2[j++];
+ pvector[(unsigned char)(string1[i])] = last = string2[j++];
}
}
/* supported constructs:
- * Ranges, e.g., [0-9] ==> 0123456789
- * Escapes, e.g., \a ==> Control-G
- * Character classes, e.g. [:upper:] ==> A ... Z
+ * Ranges, e.g., 0-9 ==> 0123456789
+ * Escapes, e.g., \a ==> Control-G
+ * Character classes, e.g. [:upper:] ==> A...Z
+ * Equiv classess, e.g. [=A=] ==> A (hmmmmmmm?)
+ * not supported:
+ * [x*N] - repeat char x N times
+ * [x*] - repeat char x until it fills STRING2:
+ * # echo qwe123 | /usr/bin/tr 123456789 '[d]'
+ * qwe[d]
+ * # echo qwe123 | /usr/bin/tr 123456789 '[d*]'
+ * qweddd
*/
-static unsigned int expand(const char *arg, char *buffer)
+static unsigned expand(char *arg, char **buffer_p)
{
- char *buffer_start = buffer;
- unsigned i; /* XXX: FIXME: use unsigned char? */
+ char *buffer = *buffer_p;
+ unsigned pos = 0;
+ unsigned size = TR_BUFSIZ;
+ unsigned i; /* can't be unsigned char: must be able to hold 256 */
unsigned char ac;
-#if ENABLE_FEATURE_TR_CLASSES
-#define CLO ":]"
- const char * const classes[] = {
- "alpha"CLO, "alnum"CLO, "digit"CLO, "lower"CLO, "upper"CLO, "space"CLO,
- "blank"CLO, "punct"CLO, "cntrl"CLO, NULL
- };
-#define CLASS_invalid 0 /* we increment the retval */
-#define CLASS_alpha 1
-#define CLASS_alnum 2
-#define CLASS_digit 3
-#define CLASS_lower 4
-#define CLASS_upper 5
-#define CLASS_space 6
-#define CLASS_blank 7
-#define CLASS_punct 8
-#define CLASS_cntrl 9
-//#define CLASS_xdigit 10
-//#define CLASS_graph 11
-//#define CLASS_print 12
-#endif
+
while (*arg) {
+ if (pos + ASCII > size) {
+ size += ASCII;
+ *buffer_p = buffer = xrealloc(buffer, size);
+ }
if (*arg == '\\') {
+ const char *z;
arg++;
- *buffer++ = bb_process_escape_sequence(&arg);
- } else if (*(arg+1) == '-') {
- ac = *(arg+2);
- if (ac == 0) {
- *buffer++ = *arg++;
- continue;
+ z = arg;
+ ac = bb_process_escape_sequence(&z);
+ arg = (char *)z;
+ arg--;
+ *arg = ac;
+ /*
+ * fall through, there may be a range.
+ * If not, current char will be treated anyway.
+ */
+ }
+ if (arg[1] == '-') { /* "0-9..." */
+ ac = arg[2];
+ if (ac == '\0') { /* "0-": copy verbatim */
+ buffer[pos++] = *arg++; /* copy '0' */
+ continue; /* next iter will copy '-' and stop */
}
- i = *arg;
- while (i <= ac)
- *buffer++ = i++;
- arg += 3; /* Skip the assumed a-z */
- } else if (*arg == '[') {
+ i = (unsigned char) *arg;
+ arg += 3; /* skip 0-9 or 0-\ */
+ if (ac == '\\') {
+ const char *z;
+ z = arg;
+ ac = bb_process_escape_sequence(&z);
+ arg = (char *)z;
+ }
+ while (i <= ac) /* ok: i is unsigned _int_ */
+ buffer[pos++] = i++;
+ continue;
+ }
+ if ((ENABLE_FEATURE_TR_CLASSES || ENABLE_FEATURE_TR_EQUIV)
+ && *arg == '['
+ ) {
arg++;
- i = *arg++;
- if (ENABLE_FEATURE_TR_CLASSES && i == ':') {
+ i = (unsigned char) *arg++;
+ /* "[xyz...". i=x, arg points to y */
+ if (ENABLE_FEATURE_TR_CLASSES && i == ':') { /* [:class:] */
+#define CLO ":]\0"
+ static const char classes[] ALIGN1 =
+ "alpha"CLO "alnum"CLO "digit"CLO
+ "lower"CLO "upper"CLO "space"CLO
+ "blank"CLO "punct"CLO "cntrl"CLO
+ "xdigit"CLO;
+ enum {
+ CLASS_invalid = 0, /* we increment the retval */
+ CLASS_alpha = 1,
+ CLASS_alnum = 2,
+ CLASS_digit = 3,
+ CLASS_lower = 4,
+ CLASS_upper = 5,
+ CLASS_space = 6,
+ CLASS_blank = 7,
+ CLASS_punct = 8,
+ CLASS_cntrl = 9,
+ CLASS_xdigit = 10,
+ //CLASS_graph = 11,
+ //CLASS_print = 12,
+ };
smalluint j;
- { /* not really pretty.. */
- char *tmp = xstrndup(arg, 7); // warning: xdigit needs 8, not 7
- j = index_in_str_array(classes, tmp) + 1;
+ char *tmp;
+
+ /* xdigit needs 8, not 7 */
+ i = 7 + (arg[0] == 'x');
+ tmp = xstrndup(arg, i);
+ j = index_in_strings(classes, tmp) + 1;
free(tmp);
- }
- if (j == CLASS_alnum || j == CLASS_digit) {
+
+ if (j == CLASS_invalid)
+ goto skip_bracket;
+
+ arg += i;
+ if (j == CLASS_alnum || j == CLASS_digit || j == CLASS_xdigit) {
for (i = '0'; i <= '9'; i++)
- *buffer++ = i;
+ buffer[pos++] = i;
}
if (j == CLASS_alpha || j == CLASS_alnum || j == CLASS_upper) {
for (i = 'A'; i <= 'Z'; i++)
- *buffer++ = i;
+ buffer[pos++] = i;
}
if (j == CLASS_alpha || j == CLASS_alnum || j == CLASS_lower) {
for (i = 'a'; i <= 'z'; i++)
- *buffer++ = i;
+ buffer[pos++] = i;
}
if (j == CLASS_space || j == CLASS_blank) {
- *buffer++ = '\t';
+ buffer[pos++] = '\t';
if (j == CLASS_space) {
- *buffer++ = '\n';
- *buffer++ = '\v';
- *buffer++ = '\f';
- *buffer++ = '\r';
+ buffer[pos++] = '\n';
+ buffer[pos++] = '\v';
+ buffer[pos++] = '\f';
+ buffer[pos++] = '\r';
}
- *buffer++ = ' ';
+ buffer[pos++] = ' ';
}
if (j == CLASS_punct || j == CLASS_cntrl) {
- for (i = 0; i <= ASCII; i++)
- if ((j == CLASS_punct &&
- isprint(i) && (!isalnum(i)) && (!isspace(i))) ||
- (j == CLASS_cntrl && iscntrl(i)))
- *buffer++ = i;
+ for (i = '\0'; i < ASCII; i++) {
+ if ((j == CLASS_punct && isprint_asciionly(i) && !isalnum(i) && !isspace(i))
+ || (j == CLASS_cntrl && iscntrl(i))
+ ) {
+ buffer[pos++] = i;
+ }
+ }
}
- if (j == CLASS_invalid) {
- *buffer++ = '[';
- *buffer++ = ':';
- continue;
+ if (j == CLASS_xdigit) {
+ for (i = 'A'; i <= 'F'; i++) {
+ buffer[pos + 6] = i | 0x20;
+ buffer[pos++] = i;
+ }
+ pos += 6;
}
- break;
- }
- if (ENABLE_FEATURE_TR_EQUIV && i == '=') {
- *buffer++ = *arg;
- arg += 3; /* Skip the closing =] */
continue;
}
- if (*arg++ != '-') {
- *buffer++ = '[';
- arg -= 2;
+ /* "[xyz...", i=x, arg points to y */
+ if (ENABLE_FEATURE_TR_EQUIV && i == '=') { /* [=CHAR=] */
+ buffer[pos++] = *arg; /* copy CHAR */
+ if (!arg[0] || arg[1] != '=' || arg[2] != ']')
+ bb_show_usage();
+ arg += 3; /* skip CHAR=] */
continue;
}
- ac = *arg++;
- while (i <= ac)
- *buffer++ = i++;
- arg++; /* Skip the assumed ']' */
- } else
- *buffer++ = *arg++;
+ /* The rest of "[xyz..." cases is treated as normal
+ * string, "[" has no special meaning here:
+ * tr "[a-z]" "[A-Z]" can be written as tr "a-z" "A-Z",
+ * also try tr "[a-z]" "_A-Z+" and you'll see that
+ * [] is not special here.
+ */
+ skip_bracket:
+ arg -= 2; /* points to "[" in "[xyz..." */
+ }
+ buffer[pos++] = *arg++;
}
- return (buffer - buffer_start);
+ return pos;
}
+/* NB: buffer is guaranteed to be at least TR_BUFSIZE
+ * (which is >= ASCII) big.
+ */
static int complement(char *buffer, int buffer_len)
{
- short i, j, ix;
- char conv[ASCII + 2];
+ int len;
+ char conv[ASCII];
+ unsigned char ch;
- ix = 0;
- for (i = 0; i <= ASCII; i++) {
- for (j = 0; j < buffer_len; j++)
- if (buffer[j] == i)
- break;
- if (j == buffer_len)
- conv[ix++] = i & ASCII;
+ len = 0;
+ ch = '\0';
+ while (1) {
+ if (memchr(buffer, ch, buffer_len) == NULL)
+ conv[len++] = ch;
+ if (++ch == '\0')
+ break;
}
- memcpy(buffer, conv, ix);
- return ix;
+ memcpy(buffer, conv, len);
+ return len;
}
-int tr_main(int argc, char **argv);
-int tr_main(int argc, char **argv)
+int tr_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
+int tr_main(int argc UNUSED_PARAM, char **argv)
{
- unsigned char *ptr;
- int output_length = 0, input_length;
- int idx = 1;
int i;
- smalluint flags = 0;
- RESERVE_CONFIG_UBUFFER(output, BUFSIZ);
- RESERVE_CONFIG_BUFFER(vector, ASCII+1);
- RESERVE_CONFIG_BUFFER(invec, ASCII+1);
- RESERVE_CONFIG_BUFFER(outvec, ASCII+1);
+ smalluint opts;
+ ssize_t read_chars;
+ size_t in_index, out_index;
+ unsigned last = UCHAR_MAX + 1; /* not equal to any char */
+ unsigned char coded, c;
+ char *str1 = xmalloc(TR_BUFSIZ);
+ char *str2 = xmalloc(TR_BUFSIZ);
+ int str2_length;
+ int str1_length;
+ char *vector = xzalloc(ASCII * 3);
+ char *invec = vector + ASCII;
+ char *outvec = vector + ASCII * 2;
- /* ... but make them available globally */
- poutput = output;
- pvector = vector;
- pinvec = invec;
- poutvec = outvec;
+#define TR_OPT_complement (3 << 0)
+#define TR_OPT_delete (1 << 2)
+#define TR_OPT_squeeze_reps (1 << 3)
- if (argc > 1 && argv[idx][0] == '-') {
- for (ptr = (unsigned char *) &argv[idx][1]; *ptr; ptr++) {
- if (*ptr == 'c')
- flags |= TR_OPT_complement;
- else if (*ptr == 'd')
- flags |= TR_OPT_delete;
- else if (*ptr == 's')
- flags |= TR_OPT_squeeze_reps;
- else
- bb_show_usage();
- }
- idx++;
- }
- for (i = 0; i <= ASCII; i++) {
+ for (i = 0; i < ASCII; i++) {
vector[i] = i;
- invec[i] = outvec[i] = FALSE;
+ /*invec[i] = outvec[i] = FALSE; - done by xzalloc */
}
- if (argv[idx] != NULL) {
- input_length = expand(argv[idx++], bb_common_bufsiz1);
- if (flags & TR_OPT_complement)
- input_length = complement(bb_common_bufsiz1, input_length);
- if (argv[idx] != NULL) {
- if (*argv[idx] == '\0')
- bb_error_msg_and_die("STRING2 cannot be empty");
- output_length = expand(argv[idx], output);
- map(bb_common_bufsiz1, input_length, output, output_length);
+ /* -C/-c difference is that -C complements "characters",
+ * and -c complements "values" (binary bytes I guess).
+ * In POSIX locale, these are the same.
+ */
+
+ /* '+': stop at first non-option */
+ opts = getopt32(argv, "^+" "Ccds" "\0" "-1");
+ argv += optind;
+
+ str1_length = expand(*argv++, &str1);
+ str2_length = 0;
+ if (opts & TR_OPT_complement)
+ str1_length = complement(str1, str1_length);
+ if (*argv) {
+ if (argv[0][0] == '\0')
+ bb_error_msg_and_die("STRING2 cannot be empty");
+ str2_length = expand(*argv, &str2);
+ map(vector, str1, str1_length,
+ str2, str2_length);
+ }
+ for (i = 0; i < str1_length; i++)
+ invec[(unsigned char)(str1[i])] = TRUE;
+ for (i = 0; i < str2_length; i++)
+ outvec[(unsigned char)(str2[i])] = TRUE;
+
+ goto start_from;
+
+ /* In this loop, str1 space is reused as input buffer,
+ * str2 - as output one. */
+ for (;;) {
+ /* If we're out of input, flush output and read more input. */
+ if ((ssize_t)in_index == read_chars) {
+ if (out_index) {
+ xwrite(STDOUT_FILENO, str2, out_index);
+ start_from:
+ out_index = 0;
+ }
+ read_chars = safe_read(STDIN_FILENO, str1, TR_BUFSIZ);
+ if (read_chars <= 0) {
+ if (read_chars < 0)
+ bb_perror_msg_and_die(bb_msg_read_error);
+ break;
+ }
+ in_index = 0;
}
- for (i = 0; i < input_length; i++)
- invec[(unsigned char)bb_common_bufsiz1[i]] = TRUE;
- for (i = 0; i < output_length; i++)
- outvec[output[i]] = TRUE;
+ c = str1[in_index++];
+ if ((opts & TR_OPT_delete) && invec[c])
+ continue;
+ coded = vector[c];
+ if ((opts & TR_OPT_squeeze_reps) && last == coded
+ && (invec[c] || outvec[coded])
+ ) {
+ continue;
+ }
+ str2[out_index++] = last = coded;
}
- convert(flags);
+
+ if (ENABLE_FEATURE_CLEAN_UP) {
+ free(vector);
+ free(str2);
+ free(str1);
+ }
+
return EXIT_SUCCESS;
}
projects / oweals / busybox.git / blobdiff