date.con commit date.c: allow even more varied time formats (198b0fb)
   1/*
   2 * GIT - The information manager from hell
   3 *
   4 * Copyright (C) Linus Torvalds, 2005
   5 */
   6
   7#include <stdio.h>
   8#include <stdlib.h>
   9#include <string.h>
  10#include <ctype.h>
  11#include <time.h>
  12
  13static time_t my_mktime(struct tm *tm)
  14{
  15        static const int mdays[] = {
  16            0, 31, 59, 90, 120, 151, 181, 212, 243, 273, 304, 334
  17        };
  18        int year = tm->tm_year - 70;
  19        int month = tm->tm_mon;
  20        int day = tm->tm_mday;
  21
  22        if (year < 0 || year > 129) /* algo only works for 1970-2099 */
  23                return -1;
  24        if (month < 0 || month > 11) /* array bounds */
  25                return -1;
  26        if (month < 2 || (year + 2) % 4)
  27                day--;
  28        return (year * 365 + (year + 1) / 4 + mdays[month] + day) * 24*60*60UL +
  29                tm->tm_hour * 60*60 + tm->tm_min * 60 + tm->tm_sec;
  30}
  31
  32static const char *month_names[] = {
  33        "January", "February", "March", "April", "May", "June",
  34        "July", "August", "September", "October", "November", "December"
  35};
  36
  37static const char *weekday_names[] = {
  38        "Sunday", "Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday"
  39};
  40
  41/*
  42 * Check these. And note how it doesn't do the summer-time conversion.
  43 *
  44 * In my world, it's always summer, and things are probably a bit off
  45 * in other ways too.
  46 */
  47static const struct {
  48        const char *name;
  49        int offset;
  50        int dst;
  51} timezone_names[] = {
  52        { "IDLW", -12, 0, },    /* International Date Line West */
  53        { "NT",   -11, 0, },    /* Nome */
  54        { "CAT",  -10, 0, },    /* Central Alaska */
  55        { "HST",  -10, 0, },    /* Hawaii Standard */
  56        { "HDT",  -10, 1, },    /* Hawaii Daylight */
  57        { "YST",   -9, 0, },    /* Yukon Standard */
  58        { "YDT",   -9, 1, },    /* Yukon Daylight */
  59        { "PST",   -8, 0, },    /* Pacific Standard */
  60        { "PDT",   -8, 1, },    /* Pacific Daylight */
  61        { "MST",   -7, 0, },    /* Mountain Standard */
  62        { "MDT",   -7, 1, },    /* Mountain Daylight */
  63        { "CST",   -6, 0, },    /* Central Standard */
  64        { "CDT",   -6, 1, },    /* Central Daylight */
  65        { "EST",   -5, 0, },    /* Eastern Standard */
  66        { "EDT",   -5, 1, },    /* Eastern Daylight */
  67        { "AST",   -3, 0, },    /* Atlantic Standard */
  68        { "ADT",   -3, 1, },    /* Atlantic Daylight */
  69        { "WAT",   -1, 0, },    /* West Africa */
  70
  71        { "GMT",    0, 0, },    /* Greenwich Mean */
  72        { "UTC",    0, 0, },    /* Universal (Coordinated) */
  73
  74        { "WET",    0, 0, },    /* Western European */
  75        { "BST",    0, 1, },    /* British Summer */
  76        { "CET",   +1, 0, },    /* Central European */
  77        { "MET",   +1, 0, },    /* Middle European */
  78        { "MEWT",  +1, 0, },    /* Middle European Winter */
  79        { "MEST",  +1, 1, },    /* Middle European Summer */
  80        { "CEST",  +1, 1, },    /* Central European Summer */
  81        { "MESZ",  +1, 1, },    /* Middle European Summer */
  82        { "FWT",   +1, 0, },    /* French Winter */
  83        { "FST",   +1, 1, },    /* French Summer */
  84        { "EET",   +2, 0, },    /* Eastern Europe, USSR Zone 1 */
  85        { "EEST",  +2, 1, },    /* Eastern European Daylight */
  86        { "WAST",  +7, 0, },    /* West Australian Standard */
  87        { "WADT",  +7, 1, },    /* West Australian Daylight */
  88        { "CCT",   +8, 0, },    /* China Coast, USSR Zone 7 */
  89        { "JST",   +9, 0, },    /* Japan Standard, USSR Zone 8 */
  90        { "EAST", +10, 0, },    /* Eastern Australian Standard */
  91        { "EADT", +10, 1, },    /* Eastern Australian Daylight */
  92        { "GST",  +10, 0, },    /* Guam Standard, USSR Zone 9 */
  93        { "NZT",  +11, 0, },    /* New Zealand */
  94        { "NZST", +11, 0, },    /* New Zealand Standard */
  95        { "NZDT", +11, 1, },    /* New Zealand Daylight */
  96        { "IDLE", +12, 0, },    /* International Date Line East */
  97};
  98
  99#define NR_TZ (sizeof(timezone_names) / sizeof(timezone_names[0]))
 100        
 101static int match_string(const char *date, const char *str)
 102{
 103        int i = 0;
 104
 105        for (i = 0; *date; date++, str++, i++) {
 106                if (*date == *str)
 107                        continue;
 108                if (toupper(*date) == toupper(*str))
 109                        continue;
 110                if (!isalnum(*date))
 111                        break;
 112                return 0;
 113        }
 114        return i;
 115}
 116
 117/*
 118* Parse month, weekday, or timezone name
 119*/
 120static int match_alpha(const char *date, struct tm *tm, int *offset)
 121{
 122        int i;
 123
 124        for (i = 0; i < 12; i++) {
 125                int match = match_string(date, month_names[i]);
 126                if (match >= 3) {
 127                        tm->tm_mon = i;
 128                        return match;
 129                }
 130        }
 131
 132        for (i = 0; i < 7; i++) {
 133                int match = match_string(date, weekday_names[i]);
 134                if (match >= 3) {
 135                        tm->tm_wday = i;
 136                        return match;
 137                }
 138        }
 139
 140        for (i = 0; i < NR_TZ; i++) {
 141                int match = match_string(date, timezone_names[i].name);
 142                if (match >= 3) {
 143                        int off = timezone_names[i].offset;
 144
 145                        /* This is bogus, but we like summer */
 146                        off += timezone_names[i].dst;
 147
 148                        /* Only use the tz name offset if we don't have anything better */
 149                        if (*offset == -1)
 150                                *offset = 60*off;
 151
 152                        return match;
 153                }
 154        }
 155
 156        /* BAD CRAP */
 157        return 0;
 158}
 159
 160static int is_date(int year, int month, int day, struct tm *tm)
 161{
 162        if (month > 0 && month < 13 && day > 0 && day < 32) {
 163                if (year == -1) {
 164                        tm->tm_mon = month-1;
 165                        tm->tm_mday = day;
 166                        return 1;
 167                }
 168                if (year >= 1970 && year < 2100) {
 169                        year -= 1900;
 170                } else if (year > 70 && year < 100) {
 171                        /* ok */
 172                } else if (year < 38) {
 173                        year += 100;
 174                } else
 175                        return 0;
 176
 177                tm->tm_mon = month-1;
 178                tm->tm_mday = day;
 179                tm->tm_year = year;
 180                return 1;
 181        }
 182        return 0;
 183}
 184
 185static int match_multi_number(unsigned long num, char c, char *date, char *end, struct tm *tm)
 186{
 187        long num2, num3;
 188
 189        num2 = strtol(end+1, &end, 10);
 190        num3 = -1;
 191        if (*end == c && isdigit(end[1]))
 192                num3 = strtol(end+1, &end, 10);
 193
 194        /* Time? Date? */
 195        switch (c) {
 196        case ':':
 197                if (num3 < 0)
 198                        num3 = 0;
 199                if (num < 25 && num2 >= 0 && num2 < 60 && num3 >= 0 && num3 <= 60) {
 200                        tm->tm_hour = num;
 201                        tm->tm_min = num2;
 202                        tm->tm_sec = num3;
 203                        break;
 204                }
 205                return 0;
 206
 207        case '-':
 208        case '/':
 209                if (num > 70) {
 210                        /* yyyy-mm-dd? */
 211                        if (is_date(num, num2, num3, tm))
 212                                break;
 213                        /* yyyy-dd-mm? */
 214                        if (is_date(num, num3, num2, tm))
 215                                break;
 216                }
 217                /* mm/dd/yy ? */
 218                if (is_date(num3, num2, num, tm))
 219                        break;
 220                /* dd/mm/yy ? */
 221                if (is_date(num3, num, num2, tm))
 222                        break;
 223                return 0;
 224        }
 225        return end - date;
 226}
 227
 228/*
 229 * We've seen a digit. Time? Year? Date? 
 230 */
 231static int match_digit(char *date, struct tm *tm, int *offset)
 232{
 233        int n;
 234        char *end;
 235        unsigned long num;
 236
 237        num = strtoul(date, &end, 10);
 238
 239        /*
 240         * Seconds since 1970? We trigger on that for anything after Jan 1, 2000
 241         */
 242        if (num > 946684800) {
 243                time_t time = num;
 244                if (gmtime_r(&time, tm))
 245                        return end - date;
 246        }
 247
 248        /*
 249         * Check for special formats: num[:-/]num[same]num
 250         */
 251        switch (*end) {
 252        case ':':
 253        case '/':
 254        case '-':
 255                if (isdigit(end[1])) {
 256                        int match = match_multi_number(num, *end, date, end, tm);
 257                        if (match)
 258                                return match;
 259                }
 260        }
 261
 262        /*
 263         * None of the special formats? Try to guess what
 264         * the number meant. We use the number of digits
 265         * to make a more educated guess..
 266         */
 267        n = 0;
 268        do {
 269                n++;
 270        } while (isdigit(date[n]));
 271
 272        /* Four-digit year or a timezone? */
 273        if (n == 4) {
 274                if (num <= 1200 && *offset == -1) {
 275                        unsigned int minutes = num % 100;
 276                        unsigned int hours = num / 100;
 277                        *offset = hours*60 + minutes;
 278                } else if (num > 1900 && num < 2100)
 279                        tm->tm_year = num - 1900;
 280                return n;
 281        }
 282
 283        /*
 284         * NOTE! We will give precedence to day-of-month over month or
 285         * year numebers in the 1-12 range. So 05 is always "mday 5",
 286         * unless we already have a mday..
 287         *
 288         * IOW, 01 Apr 05 parses as "April 1st, 2005".
 289         */
 290        if (num > 0 && num < 32 && tm->tm_mday < 0) {
 291                tm->tm_mday = num;
 292                return n;
 293        }
 294
 295        /* Two-digit year? */
 296        if (n == 2 && tm->tm_year < 0) {
 297                if (num < 10 && tm->tm_mday >= 0) {
 298                        tm->tm_year = num + 100;
 299                        return n;
 300                }
 301                if (num >= 70) {
 302                        tm->tm_year = num;
 303                        return n;
 304                }
 305        }
 306
 307        if (num > 0 && num < 32) {
 308                tm->tm_mday = num;
 309        } else if (num > 1900) {
 310                tm->tm_year = num - 1900;
 311        } else if (num > 70) {
 312                tm->tm_year = num;
 313        } else if (num > 0 && num < 13) {
 314                tm->tm_mon = num-1;
 315        }
 316                
 317        return n;
 318}
 319
 320static int match_tz(char *date, int *offp)
 321{
 322        char *end;
 323        int offset = strtoul(date+1, &end, 10);
 324        int min, hour;
 325        int n = end - date - 1;
 326
 327        min = offset % 100;
 328        hour = offset / 100;
 329
 330        /*
 331         * Don't accept any random crap.. At least 3 digits, and
 332         * a valid minute. We might want to check that the minutes
 333         * are divisible by 30 or something too.
 334         */
 335        if (min >= 60 || n < 3)
 336                return 0;
 337
 338        offset = hour*60+min;
 339        if (*date == '-')
 340                offset = -offset;
 341
 342        *offp = offset;
 343        return end - date;
 344}
 345
 346/* Gr. strptime is crap for this; it doesn't have a way to require RFC2822
 347   (i.e. English) day/month names, and it doesn't work correctly with %z. */
 348void parse_date(char *date, char *result, int maxlen)
 349{
 350        struct tm tm;
 351        int offset, sign;
 352        time_t then;
 353
 354        memset(&tm, 0, sizeof(tm));
 355        tm.tm_year = -1;
 356        tm.tm_mon = -1;
 357        tm.tm_mday = -1;
 358        tm.tm_isdst = -1;
 359        offset = -1;
 360
 361        for (;;) {
 362                int match = 0;
 363                unsigned char c = *date;
 364
 365                /* Stop at end of string or newline */
 366                if (!c || c == '\n')
 367                        break;
 368
 369                if (isalpha(c))
 370                        match = match_alpha(date, &tm, &offset);
 371                else if (isdigit(c))
 372                        match = match_digit(date, &tm, &offset);
 373                else if ((c == '-' || c == '+') && isdigit(date[1]))
 374                        match = match_tz(date, &offset);
 375
 376                if (!match) {
 377                        /* BAD CRAP */
 378                        match = 1;
 379                }       
 380
 381                date += match;
 382        }
 383
 384        /* mktime uses local timezone */
 385        then = my_mktime(&tm); 
 386        if (offset == -1)
 387                offset = (then - mktime(&tm)) / 60;
 388
 389        if (then == -1)
 390                return;
 391
 392        then -= offset * 60;
 393
 394        sign = '+';
 395        if (offset < 0) {
 396                offset = -offset;
 397                sign = '-';
 398        }
 399
 400        snprintf(result, maxlen, "%lu %c%02d%02d", then, sign, offset/60, offset % 60);
 401}
 402
 403void datestamp(char *buf, int bufsize)
 404{
 405        time_t now;
 406        int offset;
 407
 408        time(&now);
 409
 410        offset = my_mktime(localtime(&now)) - now;
 411        offset /= 60;
 412
 413        snprintf(buf, bufsize, "%lu %+05d", now, offset/60*100 + offset%60);
 414}