1 /* sort.c - put input lines into order
2 *
3 * Copyright 2004, 2008 Rob Landley <rob@landley.net>
4 *
5 * See http://opengroup.org/onlinepubs/007904975/utilities/sort.html
6 *
7 * Deviations from POSIX: Lots.
8 * We invented -x
9
10 USE_SORT(NEWTOY(sort, USE_SORT_FLOAT("g")"S:T:m" "o:k*t:" "xVbMCcszdfirun", TOYFLAG_USR|TOYFLAG_BIN|TOYFLAG_ARGFAIL(2)))
11
12 config SORT
13 bool "sort"
14 default y
15 help
16 usage: sort [-runbcdfiMsz] [FILE...] [-k#[,#[x]] [-t X]] [-o FILE]
17
18 Sort all lines of text from input files (or stdin) to stdout.
19
20 -r Reverse
21 -u Unique lines only
22 -n Numeric order (instead of alphabetical)
23 -b Ignore leading blanks (or trailing blanks in second part of key)
24 -C Check whether input is sorted
25 -c Warn if input is unsorted
26 -d Dictionary order (use alphanumeric and whitespace chars only)
27 -f Force uppercase (case insensitive sort)
28 -i Ignore nonprinting characters
29 -M Month sort (jan, feb, etc)
30 -x Hexadecimal numerical sort
31 -s Skip fallback sort (only sort with keys)
32 -z Zero (null) terminated lines
33 -k Sort by "key" (see below)
34 -t Use a key separator other than whitespace
35 -o Output to FILE instead of stdout
36 -V Version numbers (name-1.234-rc6.5b.tgz)
37
38 Sorting by key looks at a subset of the words on each line. -k2 uses the
39 second word to the end of the line, -k2,2 looks at only the second word,
40 -k2,4 looks from the start of the second to the end of the fourth word.
41 -k2.4,5 starts from the fourth character of the second word, to the end
42 of the fifth word. Specifying multiple keys uses the later keys as tie
43 breakers, in order. A type specifier appended to a sort key (such as -2,2n)
44 applies only to sorting that key.
45
46 config SORT_FLOAT
47 bool
48 default y
49 depends on TOYBOX_FLOAT
50 help
51 usage: sort [-g]
52
53 -g General numeric sort (double precision with nan and inf)
54 */
55
56 #define FOR_sort
57 #include "toys.h"
58
59 GLOBALS(
60 char *t;
61 struct arg_list *k;
62 char *o, *T, S;
63
64 void *key_list;
65 unsigned linecount;
66 char **lines, *name;
67 )
68
69 // The sort types are n, g, and M.
70 // u, c, s, and z apply to top level only, not to keys.
71 // b at top level implies bb.
72 // The remaining options can be applied to search keys.
73
74 #define FLAG_bb (1<<31) // Ignore trailing blanks
75
76 struct sort_key {
77 struct sort_key *next_key; // linked list
78 unsigned range[4]; // start word, start char, end word, end char
79 int flags;
80 };
81
82 // Copy of the part of this string corresponding to a key/flags.
83
get_key_data(char * str,struct sort_key * key,int flags)84 static char *get_key_data(char *str, struct sort_key *key, int flags)
85 {
86 int start = 0, end, len, i, j;
87
88 // Special case whole string, so we don't have to make a copy
89
90 if(key->range[0]==1 && !key->range[1] && !key->range[2] && !key->range[3]
91 && !(flags&(FLAG_b|FLAG_d|FLAG_i|FLAG_bb))) return str;
92
93 // Find start of key on first pass, end on second pass
94
95 len = strlen(str);
96 for (j=0; j<2; j++) {
97 if (!key->range[2*j]) end=len;
98
99 // Loop through fields
100 else {
101 end = 0;
102 for (i = 1; i < key->range[2*j]+j; i++) {
103
104 // Skip leading blanks
105 if (str[end] && !TT.t) while (isspace(str[end])) end++;
106
107 // Skip body of key
108 for (; str[end]; end++) {
109 if (TT.t) {
110 if (str[end]==*TT.t) {
111 end++;
112 break;
113 }
114 } else if (isspace(str[end])) break;
115 }
116 }
117 }
118 if (!j) start = end;
119 }
120
121 // Key with explicit separator starts after the separator
122 if (TT.t && str[start]==*TT.t) start++;
123
124 // Strip leading and trailing whitespace if necessary
125 if ((flags&FLAG_b) || (!TT.t && !key->range[3]))
126 while (isspace(str[start])) start++;
127 if (flags&FLAG_bb) while (end>start && isspace(str[end-1])) end--;
128
129 // Handle offsets on start and end
130 if (key->range[3]) {
131 end += key->range[3]-1;
132 if (end>len) end=len;
133 }
134 if (key->range[1]) {
135 start += key->range[1]-1;
136 if (start>len) start=len;
137 }
138
139 // Make the copy
140 if (end<start) end = start;
141 str = xstrndup(str+start, end-start);
142
143 // Handle -d
144 if (flags&FLAG_d) {
145 for (start = end = 0; str[end]; end++)
146 if (isspace(str[end]) || isalnum(str[end])) str[start++] = str[end];
147 str[start] = 0;
148 }
149
150 // Handle -i
151 if (flags&FLAG_i) {
152 for (start = end = 0; str[end]; end++)
153 if (isprint(str[end])) str[start++] = str[end];
154 str[start] = 0;
155 }
156
157 return str;
158 }
159
160 // append a sort_key to key_list.
161
add_key(void)162 static struct sort_key *add_key(void)
163 {
164 void **stupid_compiler = &TT.key_list;
165 struct sort_key **pkey = (struct sort_key **)stupid_compiler;
166
167 while (*pkey) pkey = &((*pkey)->next_key);
168 return *pkey = xzalloc(sizeof(struct sort_key));
169 }
170
171 // Perform actual comparison
compare_values(int flags,char * x,char * y)172 static int compare_values(int flags, char *x, char *y)
173 {
174 if (CFG_SORT_FLOAT && (flags & FLAG_g)) {
175 char *xx,*yy;
176 double dx = strtod(x,&xx), dy = strtod(y,&yy);
177 int xinf, yinf;
178
179 // not numbers < NaN < -infinity < numbers < +infinity
180
181 if (x==xx) return y==yy ? 0 : -1;
182 if (y==yy) return 1;
183
184 // Check for isnan
185 if (dx!=dx) return (dy!=dy) ? 0 : -1;
186 if (dy!=dy) return 1;
187
188 // Check for infinity. (Could underflow, but avoids needing libm.)
189 xinf = (1.0/dx == 0.0);
190 yinf = (1.0/dy == 0.0);
191 if (xinf) {
192 if(dx<0) return (yinf && dy<0) ? 0 : -1;
193 return (yinf && dy>0) ? 0 : 1;
194 }
195 if (yinf) return dy<0 ? 1 : -1;
196
197 return dx<dy ? -1 : dx>dy;
198 } else if (flags & FLAG_M) {
199 struct tm thyme;
200 int dx;
201 char *xx,*yy;
202
203 xx = strptime(x,"%b",&thyme);
204 dx = thyme.tm_mon;
205 yy = strptime(y,"%b",&thyme);
206 if (!xx) return !yy ? 0 : -1;
207 else if (!yy) return 1;
208 else return dx==thyme.tm_mon ? 0 : dx-thyme.tm_mon;
209
210 } else if (flags & FLAG_x) return strtol(x, NULL, 16)-strtol(y, NULL, 16);
211 else if (flags & FLAG_V) {
212 while (*x && *y) {
213 while (*x && *x == *y) x++, y++;
214 if (isdigit(*x) && isdigit(*y)) {
215 long long xx = strtoll(x, &x, 10), yy = strtoll(y, &y, 10);
216
217 if (xx<yy) return -1;
218 if (xx>yy) return 1;
219 } else {
220 char xx = *x ? *x : x[-1], yy = *y ? *y : y[-1];
221
222 // -rc/-pre hack so abc-123 > abc-123-rc1 (other way already - < 0-9)
223 if (xx != yy) {
224 if (xx<yy && !strstart(&y, "-rc") && !strstart(&y, "-pre")) return -1;
225 else return 1;
226 }
227 }
228 }
229 return *x ? !!*y : -1;
230 // This is actually an integer sort with decimals sorted by string fallback.
231 } else if (flags & FLAG_n) {
232 long long dx = atoll(x), dy = atoll(y);
233
234 return dx<dy ? -1 : dx>dy;
235
236 // Ascii sort
237 } else return ((flags&FLAG_f) ? strcasecmp : strcmp)(x, y);
238 }
239
240 // Callback from qsort(): Iterate through key_list and perform comparisons.
compare_keys(const void * xarg,const void * yarg)241 static int compare_keys(const void *xarg, const void *yarg)
242 {
243 int flags = toys.optflags, retval = 0;
244 char *x, *y, *xx = *(char **)xarg, *yy = *(char **)yarg;
245 struct sort_key *key;
246
247 for (key=(void *)TT.key_list; !retval && key; key = key->next_key) {
248 flags = key->flags ? : toys.optflags;
249
250 // Chop out and modify key chunks, handling -dfib
251
252 x = get_key_data(xx, key, flags);
253 y = get_key_data(yy, key, flags);
254
255 retval = compare_values(flags, x, y);
256
257 // Free the copies get_key_data() made.
258
259 if (x != xx) free(x);
260 if (y != yy) free(y);
261
262 if (retval) break;
263 }
264
265 // Perform fallback sort if necessary (always case insensitive, no -f,
266 // the point is to get a stable order even for -f sorts)
267 if (!retval && !FLAG(s)) {
268 flags = toys.optflags;
269 retval = strcmp(xx, yy);
270 }
271
272 return retval * ((flags&FLAG_r) ? -1 : 1);
273 }
274
275 // Read each line from file, appending to a big array.
sort_lines(char ** pline,long len)276 static void sort_lines(char **pline, long len)
277 {
278 char *line;
279
280 if (!pline) return;
281 line = *pline;
282 if (!FLAG(z) && len && line[len-1]=='\n') line[--len] = 0;
283 *pline = 0;
284
285 // handle -c here so we don't allocate more memory than necessary.
286 if (FLAG(C)||FLAG(c)) {
287 if (TT.lines && compare_keys((void *)&TT.lines, &line)>-FLAG(u)) {
288 toys.exitval = 1;
289 if (FLAG(C)) xexit();
290 error_exit("%s: Check line %u", TT.name, TT.linecount+1);
291 }
292 free(TT.lines);
293 TT.lines = (void *)line;
294 } else {
295 if (!(TT.linecount&63))
296 TT.lines = xrealloc(TT.lines, sizeof(char *)*(TT.linecount+64));
297 TT.lines[TT.linecount] = line;
298 }
299 TT.linecount++;
300 }
301
302 // Callback from loopfiles to handle input files.
sort_read(int fd,char * name)303 static void sort_read(int fd, char *name)
304 {
305 TT.name = name;
306 do_lines(fd, '\n'*!FLAG(z), sort_lines);
307 }
308
sort_main(void)309 void sort_main(void)
310 {
311 int idx, jdx, fd = 1;
312
313 if (FLAG(u)) toys.optflags |= FLAG_s;
314
315 // Parse -k sort keys.
316 if (TT.k) {
317 struct arg_list *arg;
318
319 for (arg = TT.k; arg; arg = arg->next) {
320 struct sort_key *key = add_key();
321 char *temp, *temp2, *optlist;
322 int flag;
323
324 idx = 0;
325 temp = arg->arg;
326 while (*temp) {
327 // Start of range
328 key->range[2*idx] = strtol(temp, &temp, 10);
329 if (*temp=='.') key->range[(2*idx)+1] = strtol(temp+1, &temp, 10);
330
331 // Handle flags appended to a key type.
332 for (;*temp;temp++) {
333
334 // Second comma becomes an "Unknown key" error.
335 if (*temp==',' && !idx++) {
336 temp++;
337 break;
338 }
339
340 // Which flag is this?
341 optlist = toys.which->options;
342 temp2 = strchr(optlist, *temp);
343 flag = 1<<(optlist-temp2+strlen(optlist)-1);
344
345 // Was it a flag that can apply to a key?
346 if (!temp2 || flag>FLAG_x || (flag&(FLAG_u|FLAG_c|FLAG_s|FLAG_z)))
347 error_exit("Unknown key option.");
348
349 // b after , means strip _trailing_ space, not leading.
350 if (idx && flag==FLAG_b) flag = FLAG_bb;
351 key->flags |= flag;
352 }
353 }
354 }
355 }
356
357 // global b flag strips both leading and trailing spaces
358 if (FLAG(b)) toys.optflags |= FLAG_bb;
359
360 // If no keys, perform alphabetic sort over the whole line.
361 if (!TT.key_list) add_key()->range[0] = 1;
362
363 // Open input files and read data, populating TT.lines[TT.linecount]
364 loopfiles(toys.optargs, sort_read);
365
366 // The compare (-c) logic was handled in sort_read(),
367 // so if we got here, we're done.
368 if (FLAG(C)||FLAG(c)) goto exit_now;
369
370 // Perform the actual sort
371 qsort(TT.lines, TT.linecount, sizeof(char *), compare_keys);
372
373 // handle unique (-u)
374 if (FLAG(u)) {
375 for (jdx=0, idx=1; idx<TT.linecount; idx++) {
376 if (!compare_keys(&TT.lines[jdx], &TT.lines[idx])) free(TT.lines[idx]);
377 else TT.lines[++jdx] = TT.lines[idx];
378 }
379 if (TT.linecount) TT.linecount = jdx+1;
380 }
381
382 // Open output file if necessary. We can't do this until we've finished
383 // reading in case the output file is one of the input files.
384 if (TT.o) fd = xcreate(TT.o, O_CREAT|O_TRUNC|O_WRONLY, 0666);
385
386 // Output result
387 for (idx = 0; idx<TT.linecount; idx++) {
388 char *s = TT.lines[idx];
389 unsigned i = strlen(s);
390
391 if (!FLAG(z)) s[i] = '\n';
392 xwrite(fd, s, i+1);
393 if (CFG_TOYBOX_FREE) free(s);
394 }
395
396 exit_now:
397 if (CFG_TOYBOX_FREE) {
398 if (fd != 1) close(fd);
399 free(TT.lines);
400 }
401 }
402