date.con commit Added Packing Heursitics IRC writeup. (b116b29)
   1/*
   2 * GIT - The information manager from hell
   3 *
   4 * Copyright (C) Linus Torvalds, 2005
   5 */
   6
   7#include <time.h>
   8#include <sys/time.h>
   9
  10#include "cache.h"
  11
  12static time_t my_mktime(struct tm *tm)
  13{
  14        static const int mdays[] = {
  15            0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334
  16        };
  17        int year = tm->tm_year - 70;
  18        int month = tm->tm_mon;
  19        int day = tm->tm_mday;
  20
  21        if (year < 0 || year > 129) /* algo only works for 1970-2099 */
  22                return -1;
  23        if (month < 0 || month > 11) /* array bounds */
  24                return -1;
  25        if (month < 2 || (year + 2) % 4)
  26                day--;
  27        return (year * 365 + (year + 1) / 4 + mdays[month] + day) * 24*60*60UL +
  28                tm->tm_hour * 60*60 + tm->tm_min * 60 + tm->tm_sec;
  29}
  30
  31static const char *month_names[] = {
  32        "January", "February", "March", "April", "May", "June",
  33        "July", "August", "September", "October", "November", "December"
  34};
  35
  36static const char *weekday_names[] = {
  37        "Sundays", "Mondays", "Tuesdays", "Wednesdays", "Thursdays", "Fridays", "Saturdays"
  38};
  39
  40/*
  41 * The "tz" thing is passed in as this strange "decimal parse of tz"
  42 * thing, which means that tz -0100 is passed in as the integer -100,
  43 * even though it means "sixty minutes off"
  44 */
  45const char *show_date(unsigned long time, int tz)
  46{
  47        struct tm *tm;
  48        time_t t;
  49        static char timebuf[200];
  50        int minutes;
  51
  52        minutes = tz < 0 ? -tz : tz;
  53        minutes = (minutes / 100)*60 + (minutes % 100);
  54        minutes = tz < 0 ? -minutes : minutes;
  55        t = time + minutes * 60;
  56        tm = gmtime(&t);
  57        if (!tm)
  58                return NULL;
  59        sprintf(timebuf, "%.3s %.3s %d %02d:%02d:%02d %d %+05d",
  60                weekday_names[tm->tm_wday],
  61                month_names[tm->tm_mon],
  62                tm->tm_mday,
  63                tm->tm_hour, tm->tm_min, tm->tm_sec,
  64                tm->tm_year + 1900, tz);
  65        return timebuf;
  66}
  67
  68/*
  69 * Check these. And note how it doesn't do the summer-time conversion.
  70 *
  71 * In my world, it's always summer, and things are probably a bit off
  72 * in other ways too.
  73 */
  74static const struct {
  75        const char *name;
  76        int offset;
  77        int dst;
  78} timezone_names[] = {
  79        { "IDLW", -12, 0, },    /* International Date Line West */
  80        { "NT",   -11, 0, },    /* Nome */
  81        { "CAT",  -10, 0, },    /* Central Alaska */
  82        { "HST",  -10, 0, },    /* Hawaii Standard */
  83        { "HDT",  -10, 1, },    /* Hawaii Daylight */
  84        { "YST",   -9, 0, },    /* Yukon Standard */
  85        { "YDT",   -9, 1, },    /* Yukon Daylight */
  86        { "PST",   -8, 0, },    /* Pacific Standard */
  87        { "PDT",   -8, 1, },    /* Pacific Daylight */
  88        { "MST",   -7, 0, },    /* Mountain Standard */
  89        { "MDT",   -7, 1, },    /* Mountain Daylight */
  90        { "CST",   -6, 0, },    /* Central Standard */
  91        { "CDT",   -6, 1, },    /* Central Daylight */
  92        { "EST",   -5, 0, },    /* Eastern Standard */
  93        { "EDT",   -5, 1, },    /* Eastern Daylight */
  94        { "AST",   -3, 0, },    /* Atlantic Standard */
  95        { "ADT",   -3, 1, },    /* Atlantic Daylight */
  96        { "WAT",   -1, 0, },    /* West Africa */
  97
  98        { "GMT",    0, 0, },    /* Greenwich Mean */
  99        { "UTC",    0, 0, },    /* Universal (Coordinated) */
 100
 101        { "WET",    0, 0, },    /* Western European */
 102        { "BST",    0, 1, },    /* British Summer */
 103        { "CET",   +1, 0, },    /* Central European */
 104        { "MET",   +1, 0, },    /* Middle European */
 105        { "MEWT",  +1, 0, },    /* Middle European Winter */
 106        { "MEST",  +1, 1, },    /* Middle European Summer */
 107        { "CEST",  +1, 1, },    /* Central European Summer */
 108        { "MESZ",  +1, 1, },    /* Middle European Summer */
 109        { "FWT",   +1, 0, },    /* French Winter */
 110        { "FST",   +1, 1, },    /* French Summer */
 111        { "EET",   +2, 0, },    /* Eastern Europe, USSR Zone 1 */
 112        { "EEST",  +2, 1, },    /* Eastern European Daylight */
 113        { "WAST",  +7, 0, },    /* West Australian Standard */
 114        { "WADT",  +7, 1, },    /* West Australian Daylight */
 115        { "CCT",   +8, 0, },    /* China Coast, USSR Zone 7 */
 116        { "JST",   +9, 0, },    /* Japan Standard, USSR Zone 8 */
 117        { "EAST", +10, 0, },    /* Eastern Australian Standard */
 118        { "EADT", +10, 1, },    /* Eastern Australian Daylight */
 119        { "GST",  +10, 0, },    /* Guam Standard, USSR Zone 9 */
 120        { "NZT",  +11, 0, },    /* New Zealand */
 121        { "NZST", +11, 0, },    /* New Zealand Standard */
 122        { "NZDT", +11, 1, },    /* New Zealand Daylight */
 123        { "IDLE", +12, 0, },    /* International Date Line East */
 124};
 125
 126static int match_string(const char *date, const char *str)
 127{
 128        int i = 0;
 129
 130        for (i = 0; *date; date++, str++, i++) {
 131                if (*date == *str)
 132                        continue;
 133                if (toupper(*date) == toupper(*str))
 134                        continue;
 135                if (!isalnum(*date))
 136                        break;
 137                return 0;
 138        }
 139        return i;
 140}
 141
 142static int skip_alpha(const char *date)
 143{
 144        int i = 0;
 145        do {
 146                i++;
 147        } while (isalpha(date[i]));
 148        return i;
 149}
 150
 151/*
 152* Parse month, weekday, or timezone name
 153*/
 154static int match_alpha(const char *date, struct tm *tm, int *offset)
 155{
 156        int i;
 157
 158        for (i = 0; i < 12; i++) {
 159                int match = match_string(date, month_names[i]);
 160                if (match >= 3) {
 161                        tm->tm_mon = i;
 162                        return match;
 163                }
 164        }
 165
 166        for (i = 0; i < 7; i++) {
 167                int match = match_string(date, weekday_names[i]);
 168                if (match >= 3) {
 169                        tm->tm_wday = i;
 170                        return match;
 171                }
 172        }
 173
 174        for (i = 0; i < ARRAY_SIZE(timezone_names); i++) {
 175                int match = match_string(date, timezone_names[i].name);
 176                if (match >= 3) {
 177                        int off = timezone_names[i].offset;
 178
 179                        /* This is bogus, but we like summer */
 180                        off += timezone_names[i].dst;
 181
 182                        /* Only use the tz name offset if we don't have anything better */
 183                        if (*offset == -1)
 184                                *offset = 60*off;
 185
 186                        return match;
 187                }
 188        }
 189
 190        if (match_string(date, "PM") == 2) {
 191                if (tm->tm_hour > 0 && tm->tm_hour < 12)
 192                        tm->tm_hour += 12;
 193                return 2;
 194        }
 195
 196        /* BAD CRAP */
 197        return skip_alpha(date);
 198}
 199
 200static int is_date(int year, int month, int day, struct tm *tm)
 201{
 202        if (month > 0 && month < 13 && day > 0 && day < 32) {
 203                if (year == -1) {
 204                        tm->tm_mon = month-1;
 205                        tm->tm_mday = day;
 206                        return 1;
 207                }
 208                if (year >= 1970 && year < 2100) {
 209                        year -= 1900;
 210                } else if (year > 70 && year < 100) {
 211                        /* ok */
 212                } else if (year < 38) {
 213                        year += 100;
 214                } else
 215                        return 0;
 216
 217                tm->tm_mon = month-1;
 218                tm->tm_mday = day;
 219                tm->tm_year = year;
 220                return 1;
 221        }
 222        return 0;
 223}
 224
 225static int match_multi_number(unsigned long num, char c, const char *date, char *end, struct tm *tm)
 226{
 227        long num2, num3;
 228
 229        num2 = strtol(end+1, &end, 10);
 230        num3 = -1;
 231        if (*end == c && isdigit(end[1]))
 232                num3 = strtol(end+1, &end, 10);
 233
 234        /* Time? Date? */
 235        switch (c) {
 236        case ':':
 237                if (num3 < 0)
 238                        num3 = 0;
 239                if (num < 25 && num2 >= 0 && num2 < 60 && num3 >= 0 && num3 <= 60) {
 240                        tm->tm_hour = num;
 241                        tm->tm_min = num2;
 242                        tm->tm_sec = num3;
 243                        break;
 244                }
 245                return 0;
 246
 247        case '-':
 248        case '/':
 249                if (num > 70) {
 250                        /* yyyy-mm-dd? */
 251                        if (is_date(num, num2, num3, tm))
 252                                break;
 253                        /* yyyy-dd-mm? */
 254                        if (is_date(num, num3, num2, tm))
 255                                break;
 256                }
 257                /* mm/dd/yy ? */
 258                if (is_date(num3, num, num2, tm))
 259                        break;
 260                /* dd/mm/yy ? */
 261                if (is_date(num3, num2, num, tm))
 262                        break;
 263                return 0;
 264        }
 265        return end - date;
 266}
 267
 268/*
 269 * We've seen a digit. Time? Year? Date? 
 270 */
 271static int match_digit(const char *date, struct tm *tm, int *offset, int *tm_gmt)
 272{
 273        int n;
 274        char *end;
 275        unsigned long num;
 276
 277        num = strtoul(date, &end, 10);
 278
 279        /*
 280         * Seconds since 1970? We trigger on that for anything after Jan 1, 2000
 281         */
 282        if (num > 946684800) {
 283                time_t time = num;
 284                if (gmtime_r(&time, tm)) {
 285                        *tm_gmt = 1;
 286                        return end - date;
 287                }
 288        }
 289
 290        /*
 291         * Check for special formats: num[:-/]num[same]num
 292         */
 293        switch (*end) {
 294        case ':':
 295        case '/':
 296        case '-':
 297                if (isdigit(end[1])) {
 298                        int match = match_multi_number(num, *end, date, end, tm);
 299                        if (match)
 300                                return match;
 301                }
 302        }
 303
 304        /*
 305         * None of the special formats? Try to guess what
 306         * the number meant. We use the number of digits
 307         * to make a more educated guess..
 308         */
 309        n = 0;
 310        do {
 311                n++;
 312        } while (isdigit(date[n]));
 313
 314        /* Four-digit year or a timezone? */
 315        if (n == 4) {
 316                if (num <= 1200 && *offset == -1) {
 317                        unsigned int minutes = num % 100;
 318                        unsigned int hours = num / 100;
 319                        *offset = hours*60 + minutes;
 320                } else if (num > 1900 && num < 2100)
 321                        tm->tm_year = num - 1900;
 322                return n;
 323        }
 324
 325        /*
 326         * NOTE! We will give precedence to day-of-month over month or
 327         * year numbers in the 1-12 range. So 05 is always "mday 5",
 328         * unless we already have a mday..
 329         *
 330         * IOW, 01 Apr 05 parses as "April 1st, 2005".
 331         */
 332        if (num > 0 && num < 32 && tm->tm_mday < 0) {
 333                tm->tm_mday = num;
 334                return n;
 335        }
 336
 337        /* Two-digit year? */
 338        if (n == 2 && tm->tm_year < 0) {
 339                if (num < 10 && tm->tm_mday >= 0) {
 340                        tm->tm_year = num + 100;
 341                        return n;
 342                }
 343                if (num >= 70) {
 344                        tm->tm_year = num;
 345                        return n;
 346                }
 347        }
 348
 349        if (num > 0 && num < 32) {
 350                tm->tm_mday = num;
 351        } else if (num > 1900) {
 352                tm->tm_year = num - 1900;
 353        } else if (num > 70) {
 354                tm->tm_year = num;
 355        } else if (num > 0 && num < 13) {
 356                tm->tm_mon = num-1;
 357        }
 358                
 359        return n;
 360}
 361
 362static int match_tz(const char *date, int *offp)
 363{
 364        char *end;
 365        int offset = strtoul(date+1, &end, 10);
 366        int min, hour;
 367        int n = end - date - 1;
 368
 369        min = offset % 100;
 370        hour = offset / 100;
 371
 372        /*
 373         * Don't accept any random crap.. At least 3 digits, and
 374         * a valid minute. We might want to check that the minutes
 375         * are divisible by 30 or something too.
 376         */
 377        if (min < 60 && n > 2) {
 378                offset = hour*60+min;
 379                if (*date == '-')
 380                        offset = -offset;
 381
 382                *offp = offset;
 383        }
 384        return end - date;
 385}
 386
 387static int date_string(unsigned long date, int offset, char *buf, int len)
 388{
 389        int sign = '+';
 390
 391        if (offset < 0) {
 392                offset = -offset;
 393                sign = '-';
 394        }
 395        return snprintf(buf, len, "%lu %c%02d%02d", date, sign, offset / 60, offset % 60);
 396}
 397
 398/* Gr. strptime is crap for this; it doesn't have a way to require RFC2822
 399   (i.e. English) day/month names, and it doesn't work correctly with %z. */
 400int parse_date(const char *date, char *result, int maxlen)
 401{
 402        struct tm tm;
 403        int offset, tm_gmt;
 404        time_t then;
 405
 406        memset(&tm, 0, sizeof(tm));
 407        tm.tm_year = -1;
 408        tm.tm_mon = -1;
 409        tm.tm_mday = -1;
 410        tm.tm_isdst = -1;
 411        offset = -1;
 412        tm_gmt = 0;
 413
 414        for (;;) {
 415                int match = 0;
 416                unsigned char c = *date;
 417
 418                /* Stop at end of string or newline */
 419                if (!c || c == '\n')
 420                        break;
 421
 422                if (isalpha(c))
 423                        match = match_alpha(date, &tm, &offset);
 424                else if (isdigit(c))
 425                        match = match_digit(date, &tm, &offset, &tm_gmt);
 426                else if ((c == '-' || c == '+') && isdigit(date[1]))
 427                        match = match_tz(date, &offset);
 428
 429                if (!match) {
 430                        /* BAD CRAP */
 431                        match = 1;
 432                }       
 433
 434                date += match;
 435        }
 436
 437        /* mktime uses local timezone */
 438        then = my_mktime(&tm); 
 439        if (offset == -1)
 440                offset = (then - mktime(&tm)) / 60;
 441
 442        if (then == -1)
 443                return -1;
 444
 445        if (!tm_gmt)
 446                then -= offset * 60;
 447        return date_string(then, offset, result, maxlen);
 448}
 449
 450void datestamp(char *buf, int bufsize)
 451{
 452        time_t now;
 453        int offset;
 454
 455        time(&now);
 456
 457        offset = my_mktime(localtime(&now)) - now;
 458        offset /= 60;
 459
 460        date_string(now, offset, buf, bufsize);
 461}
 462
 463static void update_tm(struct tm *tm, unsigned long sec)
 464{
 465        time_t n = mktime(tm) - sec;
 466        localtime_r(&n, tm);
 467}
 468
 469static void date_yesterday(struct tm *tm, int *num)
 470{
 471        update_tm(tm, 24*60*60);
 472}
 473
 474static void date_time(struct tm *tm, int hour)
 475{
 476        if (tm->tm_hour < hour)
 477                date_yesterday(tm, NULL);
 478        tm->tm_hour = hour;
 479        tm->tm_min = 0;
 480        tm->tm_sec = 0;
 481}
 482
 483static void date_midnight(struct tm *tm, int *num)
 484{
 485        date_time(tm, 0);
 486}
 487
 488static void date_noon(struct tm *tm, int *num)
 489{
 490        date_time(tm, 12);
 491}
 492
 493static void date_tea(struct tm *tm, int *num)
 494{
 495        date_time(tm, 17);
 496}
 497
 498static const struct special {
 499        const char *name;
 500        void (*fn)(struct tm *, int *);
 501} special[] = {
 502        { "yesterday", date_yesterday },
 503        { "noon", date_noon },
 504        { "midnight", date_midnight },
 505        { "tea", date_tea },
 506        { NULL }
 507};
 508
 509static const char *number_name[] = {
 510        "zero", "one", "two", "three", "four",
 511        "five", "six", "seven", "eight", "nine", "ten",
 512};
 513
 514static const struct typelen {
 515        const char *type;
 516        int length;
 517} typelen[] = {
 518        { "seconds", 1 },
 519        { "minutes", 60 },
 520        { "hours", 60*60 },
 521        { "days", 24*60*60 },
 522        { "weeks", 7*24*60*60 },
 523        { NULL }
 524};      
 525
 526static const char *approxidate_alpha(const char *date, struct tm *tm, int *num)
 527{
 528        const struct typelen *tl;
 529        const struct special *s;
 530        const char *end = date;
 531        int n = 1, i;
 532
 533        while (isalpha(*++end))
 534                n++;
 535
 536        for (i = 0; i < 12; i++) {
 537                int match = match_string(date, month_names[i]);
 538                if (match >= 3) {
 539                        tm->tm_mon = i;
 540                        return end;
 541                }
 542        }
 543
 544        for (s = special; s->name; s++) {
 545                int len = strlen(s->name);
 546                if (match_string(date, s->name) == len) {
 547                        s->fn(tm, num);
 548                        return end;
 549                }
 550        }
 551
 552        if (!*num) {
 553                for (i = 1; i < 11; i++) {
 554                        int len = strlen(number_name[i]);
 555                        if (match_string(date, number_name[i]) == len) {
 556                                *num = i;
 557                                return end;
 558                        }
 559                }
 560                if (match_string(date, "last") == 4)
 561                        *num = 1;
 562                return end;
 563        }
 564
 565        tl = typelen;
 566        while (tl->type) {
 567                int len = strlen(tl->type);
 568                if (match_string(date, tl->type) >= len-1) {
 569                        update_tm(tm, tl->length * *num);
 570                        *num = 0;
 571                        return end;
 572                }
 573                tl++;
 574        }
 575
 576        for (i = 0; i < 7; i++) {
 577                int match = match_string(date, weekday_names[i]);
 578                if (match >= 3) {
 579                        int diff, n = *num -1;
 580                        *num = 0;
 581
 582                        diff = tm->tm_wday - i;
 583                        if (diff <= 0)
 584                                n++;
 585                        diff += 7*n;
 586
 587                        update_tm(tm, diff * 24 * 60 * 60);
 588                        return end;
 589                }
 590        }
 591
 592        if (match_string(date, "months") >= 5) {
 593                int n = tm->tm_mon - *num;
 594                *num = 0;
 595                while (n < 0) {
 596                        n += 12;
 597                        tm->tm_year--;
 598                }
 599                tm->tm_mon = n;
 600                return end;
 601        }
 602
 603        if (match_string(date, "years") >= 4) {
 604                tm->tm_year -= *num;
 605                *num = 0;
 606                return end;
 607        }
 608
 609        return end;
 610}
 611
 612unsigned long approxidate(const char *date)
 613{
 614        int number = 0;
 615        struct tm tm, now;
 616        struct timeval tv;
 617        char buffer[50];
 618
 619        if (parse_date(date, buffer, sizeof(buffer)) > 0)
 620                return strtoul(buffer, NULL, 10);
 621
 622        gettimeofday(&tv, NULL);
 623        localtime_r(&tv.tv_sec, &tm);
 624        now = tm;
 625        for (;;) {
 626                unsigned char c = *date;
 627                if (!c)
 628                        break;
 629                date++;
 630                if (isdigit(c)) {
 631                        char *end;
 632                        number = strtoul(date-1, &end, 10);
 633                        date = end;
 634                        continue;
 635                }
 636                if (isalpha(c))
 637                        date = approxidate_alpha(date-1, &tm, &number);
 638        }
 639        if (number > 0 && number < 32)
 640                tm.tm_mday = number;
 641        if (tm.tm_mon > now.tm_mon && tm.tm_year == now.tm_year)
 642                tm.tm_year--;
 643        return mktime(&tm);
 644}