xsv.c

   1 /*
   2  *      A Swiss-Army Knife for CSV-like Files
   3  *
   4  *      (c) 2012 Martin Mares <mj@ucw.cz>
   5  */
   6
   7 #include <stdio.h>
   8 #include <stdlib.h>
   9 #include <string.h>
  10 #include <stdarg.h>
  11 #include <getopt.h>
  12
  13 #include <pcre.h>
  14
  15 /*** Memory allocation ***/
  16
  17 static void *xmalloc(size_t bytes)
  18 {
  19         void *p = malloc(bytes);
  20         if (!p) {
  21                 fprintf(stderr, "xsv: Out of memory (cannot allocate %zu bytes)\n", bytes);
  22                 exit(1);
  23         }
  24         return p;
  25 }
  26
  27 static void *xrealloc(void *old, size_t bytes)
  28 {
  29         void *p = realloc(old, bytes);
  30         if (!p) {
  31                 fprintf(stderr, "xsv: Out of memory (cannot allocate %zu bytes)\n", bytes);
  32                 exit(1);
  33         }
  34         return p;
  35 }
  36
  37 #define DECLARE_BUF(name, type) \
  38         typedef struct { type *start; int count; int max; } name##_t;                           \
  39         static inline void name##_init(name##_t *b) { b->start = NULL; b->count = b->max = 0; } \
  40         static inline void name##_reset(name##_t *b) { b->count = 0; }                          \
  41         static inline int name##_count(name##_t *b) { return b->count; }                        \
  42         static void name##_extend(name##_t *b) {                                                \
  43                 b->max = b->max ? 2*b->max : 16;                                                \
  44                 b->start = xrealloc(b->start, b->max * sizeof(type));                           \
  45         }                                                                                       \
  46         static inline type *name##_push(name##_t *b) {                                          \
  47                 if (b->count >= b->max) name##_extend(b);                                       \
  48                 return &b->start[b->count++];                                                   \
  49         }                                                                                       \
  50         static inline type *name##_first(name##_t *b) { return b->start; }                      \
  51         static inline type *name##_nth(name##_t *b, int n) { return &b->start[n]; }             \
  52         // end
  53
  54 /*** Formats and their parameters ***/
  55
  56 enum format_id {
  57         FORM_UNSPEC,
  58         FORM_TSV,
  59         FORM_CSV,
  60         FORM_WS,
  61         FORM_REGEX,
  62 };
  63
  64 struct format {
  65         enum format_id id;
  66         int fs;
  67         int quote;
  68         int quiet;
  69         int (*read_line)(void);
  70         void (*write_line)(void);
  71         // CSV backend:
  72         int always_quote;
  73         // regex backend:
  74         pcre *pcre;
  75         pcre_extra *pcre_extra;
  76 };
  77
  78 static struct format *in_format, *out_format;
  79
  80 struct field {
  81         int start_pos;
  82         int len;
  83 };
  84
  85 DECLARE_BUF(fields, struct field);
  86 DECLARE_BUF(line, unsigned char);
  87
  88 static fields_t in_fields, out_fields;
  89 static struct field *in_field;
  90 static line_t in_line;
  91 static int line_number;
  92
  93 static void new_field(int pos)
  94 {
  95         in_field = fields_push(&in_fields);
  96         in_field->start_pos = pos;
  97         in_field->len = 0;
  98 }
  99
 100 static void ensure_field(int pos)
 101 {
 102         if (!in_field)
 103                 new_field(pos);
 104 }
 105
 106 static void warn(struct format *fmt, char *msg, ...)
 107 {
 108         if (!fmt->quiet) {
 109                 fprintf(stderr, "Warning at line %d: ", line_number);
 110                 va_list args;
 111                 va_start(args, msg);
 112                 vfprintf(stderr, args, msg);
 113                 va_end(args);
 114                 fputc('\n', stderr);
 115         }
 116 }
 117
 118 static int next_line(void)
 119 {
 120         for (;;) {
 121                 int c = getchar();
 122                 if (c == '\r')
 123                         continue;
 124                 if (c < 0)
 125                         return !!line_count(&in_line);
 126                 if (c == '\n')
 127                         return 1;
 128                 *line_push(&in_line) = c;
 129         }
 130 }
 131
 132 static int csv_read(void)
 133 {
 134         int quoted = 0;
 135         for (;;) {
 136                 int c = getchar();
 137                 int i = line_count(&in_line);
 138 restart:
 139                 if (c == '\r')
 140                         continue;
 141                 if (c < 0 || c == '\n') {
 142                         if (quoted)
 143                                 warn(in_format, "Missing closing quote.");
 144                         if (c < 0)
 145                                 return !!fields_count(&in_fields);
 146                         else
 147                                 return 1;
 148                 }
 149                 if (quoted) {
 150                         if (c == in_format->quote) {
 151                                 c = getchar();
 152                                 if (c != in_format->quote) {
 153                                         quoted = 0;
 154                                         goto restart;
 155                                 }
 156                                 // Two quotes assimilate to one
 157                         }
 158                         // Fall through to pushing the character
 159                 } else if (c == in_format->quote) {
 160                         quoted = 1;
 161                         continue;
 162                 } else if (c == in_format->fs && !quoted) {
 163                         ensure_field(i);
 164                         new_field(i);
 165                         continue;
 166                 }
 167                 ensure_field(i);
 168                 *line_push(&in_line) = c;
 169                 in_field->len++;
 170         }
 171 }
 172
 173 static int is_ws(int c)
 174 {
 175         return (c == ' ' || c == '\t' || c == '\f');
 176 }
 177
 178 static void csv_write(void)
 179 {
 180         unsigned char *line = line_first(&in_line);
 181         int n = fields_count(&out_fields);
 182         for (int i=0; i<n; i++) {
 183                 struct field *f = fields_nth(&out_fields, i);
 184                 int need_quotes = 0;
 185                 if (out_format->quote >= 0) {
 186                         need_quotes = out_format->always_quote;
 187                         for (int j=0; !need_quotes && j < f->len; j++) {
 188                                 int c = line[f->start_pos + j];
 189                                 if (c == out_format->fs || c == out_format->quote)
 190                                         need_quotes = 1;
 191                         }
 192                 }
 193                 if (i)
 194                         putchar(out_format->fs);
 195                 if (need_quotes)
 196                         putchar(out_format->quote);
 197                 for (int j=0; j < f->len; j++) {
 198                         int c = line[f->start_pos + j];
 199                         if (c == out_format->fs && !need_quotes)
 200                                 warn(out_format, "Field separator found inside field and quoting is turned off.");
 201                         if (c == out_format->quote)
 202                                 putchar(c);
 203                         putchar(c);
 204                 }
 205                 if (need_quotes)
 206                         putchar(out_format->quote);
 207         }
 208         putchar('\n');
 209 }
 210
 211 static int ws_read(void)
 212 {
 213         if (!next_line())
 214                 return 0;
 215
 216         unsigned char *line = line_first(&in_line);
 217         int n = line_count(&in_line);
 218         if (!n)
 219                 return 1;
 220
 221         int ws = 0;
 222         new_field(0);
 223         for (int i=0; i<n; i++) {
 224                 int c = line[i];
 225                 if (is_ws(c)) {
 226                         ws++;
 227                 } else {
 228                         if (ws) {
 229                                 new_field(i);
 230                                 ws = 0;
 231                         }
 232                         in_field->len++;
 233                 }
 234         }
 235
 236         if (ws)
 237                 new_field(n);
 238         return 1;
 239 }
 240
 241 static const char *regex_set(struct format *f, char *rx)
 242 {
 243         const char *err;
 244         int errpos;
 245         f->pcre = pcre_compile(rx, PCRE_DOLLAR_ENDONLY, &err, &errpos, NULL);
 246         if (!f->pcre)
 247                 return err;
 248
 249         f->pcre_extra = pcre_study(f->pcre, 0, &err);
 250         if (!f->pcre_extra)
 251                 return err;
 252
 253         return NULL;
 254 }
 255
 256 static int regex_read(void)
 257 {
 258         if (!next_line())
 259                 return 0;
 260
 261         unsigned char *c = line_first(&in_line);
 262         int n = line_count(&in_line);
 263         if (!n)
 264                 return 1;
 265
 266         int i = 0;
 267         for (;;) {
 268                 int ovec[3];
 269                 int sep = pcre_exec(in_format->pcre, in_format->pcre_extra, (char *) c, n, i, 0, ovec, 3);
 270                 if (sep < 0) {
 271                         if (sep != PCRE_ERROR_NOMATCH)
 272                                 warn(in_format, "PCRE matching error %d", sep);
 273                         // No further occurrence of the separator: the rest is a single field
 274                         new_field(i);
 275                         in_field->len = n - i;
 276                         return 1;
 277                 }
 278                 new_field(i);
 279                 in_field->len = ovec[0] - i;
 280                 i = ovec[1];
 281         }
 282 }
 283
 284 /*** Transforms ***/
 285
 286 static void trim_fields(void)
 287 {
 288         unsigned char *line = line_first(&in_line);
 289         for (int i = 0; i < fields_count(&in_fields); i++) {
 290                 struct field *f = fields_nth(&in_fields, i);
 291                 while (f->len && is_ws(line[f->start_pos]))
 292                         f->start_pos++, f->len--;
 293                 while (f->len && is_ws(line[f->start_pos + f->len - 1]))
 294                         f->len--;
 295         }
 296 }
 297
 298 /*** Field selection ***/
 299
 300 struct selector {
 301         int first_field, last_field;
 302 };
 303
 304 DECLARE_BUF(selectors, struct selector);
 305 static selectors_t selectors;
 306
 307 static char *parse_selector(char *str)
 308 {
 309         char buf[strlen(str) + 1];
 310         strcpy(buf, str);
 311
 312         struct selector *s = selectors_push(&selectors);
 313         char *sep = strchr(buf, '-');
 314         if (sep) {
 315                 *sep++ = 0;
 316                 s->first_field = atoi(buf);
 317                 s->last_field = atoi(sep);
 318         } else
 319                 s->first_field = s->last_field = atoi(buf);
 320
 321         return NULL;
 322 }
 323
 324 static void finish_parse_selectors(void)
 325 {
 326         if (!selectors_count(&selectors))
 327                 parse_selector("-");
 328 }
 329
 330 static void select_fields(void)
 331 {
 332         for (int i = 0; i < selectors_count(&selectors); i++) {
 333                 struct selector *s = selectors_nth(&selectors, i);
 334                 int first = s->first_field;
 335                 if (first <= 0)
 336                         first = 1;
 337                 int last = s->last_field;
 338                 if (last <= 0)
 339                         last = fields_count(&in_fields);
 340                 for (int j = first; j <= last; j++) {
 341                         struct field *f = fields_push(&out_fields);
 342                         if (j >= 1 && j <= fields_count(&in_fields))
 343                                 *f = *fields_nth(&in_fields, j-1);
 344                         else
 345                                 f->start_pos = f->len = 0;
 346                 }
 347         }
 348 }
 349
 350 /*** Parsing of arguments ***/
 351
 352 static void usage(void)
 353 {
 354         printf("\
 355 Usage: xsv <in-format> [<out-format>] <options> [<fields>]\n\
 356 \n\
 357 Formats:\n\
 358 -t, --tsv               TAB-separated values (default)\n\
 359 -c, --csv               Comma-separated values\n\
 360 -w, --ws                Values separated by arbitrary whitespace\n\
 361 -r, --regex=<rx>        Separator given by Perl regular expression (input only)\n\
 362 \n\
 363 Format parameters:\n\
 364 -d, --fs=<char>         Delimiter of fields\n\
 365 -q, --quiet             Do not show warnings\n\
 366     --always-quote      Put quotes around all fields (CSV output only)\n\
 367 \n\
 368 Other options:\n\
 369     --trim              Trim leading and trailing whitespaces in fields\n\
 370 ");
 371         exit(0);
 372 }
 373
 374 static void bad_args(const char *msg, ...)
 375 {
 376         if (msg) {
 377                 va_list args;
 378                 va_start(args, msg);
 379                 fprintf(stderr, "xsv: ");
 380                 vfprintf(stderr, msg, args);
 381                 fputc('\n', stderr);
 382                 va_end(args);
 383         }
 384         fprintf(stderr, "Try `xsv --help' for more information.\n");
 385         exit(1);
 386 }
 387
 388 static const char short_options[] = "cd:qr:tw";
 389
 390 enum long_options {
 391         OPT_HELP = 256,
 392         OPT_TRIM,
 393         OPT_ALWAYS_QUOTE,
 394 };
 395
 396 static const struct option long_options[] = {
 397         { "always-quote",       0,      NULL,   OPT_ALWAYS_QUOTE },
 398         { "csv",                0,      NULL,   'c' },
 399         { "fs",                 1,      NULL,   'd' },
 400         { "quiet",              0,      NULL,   'q' },
 401         { "regex",              1,      NULL,   'r' },
 402         { "trim",               0,      NULL,   OPT_TRIM },
 403         { "tsv",                0,      NULL,   't' },
 404         { "ws",                 0,      NULL,   'w' },
 405         { "help",               0,      NULL,   OPT_HELP },
 406         { NULL,                 0,      NULL,   0 },
 407 };
 408
 409 static void set_format(int format_id)
 410 {
 411         struct format *f = xmalloc(sizeof(*f));
 412         memset(f, 0, sizeof(*f));
 413         f->id = format_id;
 414
 415         switch (format_id) {
 416                 case FORM_TSV:
 417                         f->fs = '\t';
 418                         f->quote = -1;
 419                         f->read_line = csv_read;
 420                         f->write_line = csv_write;
 421                         break;
 422                 case FORM_CSV:
 423                         f->fs = ',';
 424                         f->quote = '"';
 425                         f->read_line = csv_read;
 426                         f->write_line = csv_write;
 427                         break;
 428                 case FORM_WS:
 429                         f->fs = ' ';
 430                         f->quote = -1;
 431                         f->read_line = ws_read;
 432                         f->write_line = csv_write;
 433                         break;
 434                 case FORM_REGEX:
 435                         f->read_line = regex_read;
 436                         break;
 437         }
 438
 439         if (!in_format)
 440                 in_format = f;
 441         else if (!out_format)
 442                 out_format = f;
 443         else
 444                 bad_args("At most two format may be given.");
 445 }
 446
 447 static struct format *current_format(void)
 448 {
 449         if (out_format)
 450                 return out_format;
 451         if (in_format)
 452                 return in_format;
 453         set_format(FORM_TSV);
 454         return in_format;
 455 }
 456
 457 int main(int argc, char **argv)
 458 {
 459         int opt;
 460         int want_trim = 0;
 461         const char *err;
 462
 463         while ((opt = getopt_long(argc, argv, short_options, long_options, NULL)) >= 0)
 464                 switch (opt) {
 465                         case 'c':
 466                                 set_format(FORM_CSV);
 467                                 break;
 468                         case 'd':
 469                                 if (optarg[0])
 470                                         current_format()->fs = optarg[0];
 471                                 else
 472                                         bad_args("No field delimiter given.");
 473                                 break;
 474                         case 'q':
 475                                 current_format()->quiet = 1;
 476                                 break;
 477                         case 'r':
 478                                 set_format(FORM_REGEX);
 479                                 err = regex_set(current_format(), optarg);
 480                                 if (err)
 481                                         bad_args("Error compiling regex: %s", err);
 482                                 break;
 483                         case 't':
 484                                 set_format(FORM_TSV);
 485                                 break;
 486                         case 'w':
 487                                 set_format(FORM_WS);
 488                                 break;
 489                         case OPT_ALWAYS_QUOTE:
 490                                 if (current_format()->id != FORM_CSV)
 491                                         bad_args("--always-quote makes sense only for CSV.");
 492                                 current_format()->always_quote = 1;
 493                                 break;
 494                         case OPT_HELP:
 495                                 usage();
 496                         case OPT_TRIM:
 497                                 want_trim = 1;
 498                                 break;
 499                         default:
 500                                 bad_args(NULL);
 501                 }
 502
 503         current_format();
 504         if (!out_format)
 505                 out_format = in_format;
 506         if (!in_format->read_line)
 507                 bad_args("Write-only format selected for input.");
 508         if (!out_format->write_line)
 509                 bad_args("Read-only format selected for output.");
 510
 511         for (int i = optind; i < argc; i++) {
 512                 err = parse_selector(argv[i]);
 513                 if (err)
 514                         bad_args(err);
 515         }
 516         finish_parse_selectors();
 517
 518         fields_init(&in_fields);
 519         fields_init(&out_fields);
 520         line_init(&in_line);
 521
 522         for (;;) {
 523                 line_number++;
 524                 fields_reset(&in_fields);
 525                 line_reset(&in_line);
 526                 in_field = NULL;
 527                 if (!in_format->read_line())
 528                         break;
 529
 530                 if (want_trim)
 531                         trim_fields();
 532
 533                 fields_reset(&out_fields);
 534                 select_fields();
 535
 536                 out_format->write_line();
 537         }
 538
 539         return 0;
 540 }