Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * datetime.c
4 : * Support functions for date/time types.
5 : *
6 : * Portions Copyright (c) 1996-2024, PostgreSQL Global Development Group
7 : * Portions Copyright (c) 1994, Regents of the University of California
8 : *
9 : *
10 : * IDENTIFICATION
11 : * src/backend/utils/adt/datetime.c
12 : *
13 : *-------------------------------------------------------------------------
14 : */
15 : #include "postgres.h"
16 :
17 : #include <ctype.h>
18 : #include <limits.h>
19 : #include <math.h>
20 :
21 : #include "access/htup_details.h"
22 : #include "access/xact.h"
23 : #include "common/int.h"
24 : #include "common/string.h"
25 : #include "funcapi.h"
26 : #include "miscadmin.h"
27 : #include "nodes/nodeFuncs.h"
28 : #include "parser/scansup.h"
29 : #include "utils/builtins.h"
30 : #include "utils/date.h"
31 : #include "utils/datetime.h"
32 : #include "utils/guc.h"
33 : #include "utils/tzparser.h"
34 :
35 : static int DecodeNumber(int flen, char *str, bool haveTextMonth,
36 : int fmask, int *tmask,
37 : struct pg_tm *tm, fsec_t *fsec, bool *is2digits);
38 : static int DecodeNumberField(int len, char *str,
39 : int fmask, int *tmask,
40 : struct pg_tm *tm, fsec_t *fsec, bool *is2digits);
41 : static int DecodeTimeCommon(char *str, int fmask, int range,
42 : int *tmask, struct pg_itm *itm);
43 : static int DecodeTime(char *str, int fmask, int range,
44 : int *tmask, struct pg_tm *tm, fsec_t *fsec);
45 : static int DecodeTimeForInterval(char *str, int fmask, int range,
46 : int *tmask, struct pg_itm_in *itm_in);
47 : static const datetkn *datebsearch(const char *key, const datetkn *base, int nel);
48 : static int DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
49 : struct pg_tm *tm);
50 : static char *AppendSeconds(char *cp, int sec, fsec_t fsec,
51 : int precision, bool fillzeros);
52 : static bool int64_multiply_add(int64 val, int64 multiplier, int64 *sum);
53 : static bool AdjustFractMicroseconds(double frac, int64 scale,
54 : struct pg_itm_in *itm_in);
55 : static bool AdjustFractDays(double frac, int scale,
56 : struct pg_itm_in *itm_in);
57 : static bool AdjustFractYears(double frac, int scale,
58 : struct pg_itm_in *itm_in);
59 : static bool AdjustMicroseconds(int64 val, double fval, int64 scale,
60 : struct pg_itm_in *itm_in);
61 : static bool AdjustDays(int64 val, int scale,
62 : struct pg_itm_in *itm_in);
63 : static bool AdjustMonths(int64 val, struct pg_itm_in *itm_in);
64 : static bool AdjustYears(int64 val, int scale,
65 : struct pg_itm_in *itm_in);
66 : static int DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp,
67 : pg_time_t *tp);
68 : static bool DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t,
69 : const char *abbr, pg_tz *tzp,
70 : int *offset, int *isdst);
71 : static pg_tz *FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp,
72 : DateTimeErrorExtra *extra);
73 :
74 :
75 : const int day_tab[2][13] =
76 : {
77 : {31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0},
78 : {31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0}
79 : };
80 :
81 : const char *const months[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun",
82 : "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", NULL};
83 :
84 : const char *const days[] = {"Sunday", "Monday", "Tuesday", "Wednesday",
85 : "Thursday", "Friday", "Saturday", NULL};
86 :
87 :
88 : /*****************************************************************************
89 : * PRIVATE ROUTINES *
90 : *****************************************************************************/
91 :
92 : /*
93 : * datetktbl holds date/time keywords.
94 : *
95 : * Note that this table must be strictly alphabetically ordered to allow an
96 : * O(ln(N)) search algorithm to be used.
97 : *
98 : * The token field must be NUL-terminated; we truncate entries to TOKMAXLEN
99 : * characters to fit.
100 : *
101 : * The static table contains no TZ, DTZ, or DYNTZ entries; rather those
102 : * are loaded from configuration files and stored in zoneabbrevtbl, whose
103 : * abbrevs[] field has the same format as the static datetktbl.
104 : */
105 : static const datetkn datetktbl[] = {
106 : /* token, type, value */
107 : {"+infinity", RESERV, DTK_LATE}, /* same as "infinity" */
108 : {EARLY, RESERV, DTK_EARLY}, /* "-infinity" reserved for "early time" */
109 : {DA_D, ADBC, AD}, /* "ad" for years > 0 */
110 : {"allballs", RESERV, DTK_ZULU}, /* 00:00:00 */
111 : {"am", AMPM, AM},
112 : {"apr", MONTH, 4},
113 : {"april", MONTH, 4},
114 : {"at", IGNORE_DTF, 0}, /* "at" (throwaway) */
115 : {"aug", MONTH, 8},
116 : {"august", MONTH, 8},
117 : {DB_C, ADBC, BC}, /* "bc" for years <= 0 */
118 : {"d", UNITS, DTK_DAY}, /* "day of month" for ISO input */
119 : {"dec", MONTH, 12},
120 : {"december", MONTH, 12},
121 : {"dow", UNITS, DTK_DOW}, /* day of week */
122 : {"doy", UNITS, DTK_DOY}, /* day of year */
123 : {"dst", DTZMOD, SECS_PER_HOUR},
124 : {EPOCH, RESERV, DTK_EPOCH}, /* "epoch" reserved for system epoch time */
125 : {"feb", MONTH, 2},
126 : {"february", MONTH, 2},
127 : {"fri", DOW, 5},
128 : {"friday", DOW, 5},
129 : {"h", UNITS, DTK_HOUR}, /* "hour" */
130 : {LATE, RESERV, DTK_LATE}, /* "infinity" reserved for "late time" */
131 : {"isodow", UNITS, DTK_ISODOW}, /* ISO day of week, Sunday == 7 */
132 : {"isoyear", UNITS, DTK_ISOYEAR}, /* year in terms of the ISO week date */
133 : {"j", UNITS, DTK_JULIAN},
134 : {"jan", MONTH, 1},
135 : {"january", MONTH, 1},
136 : {"jd", UNITS, DTK_JULIAN},
137 : {"jul", MONTH, 7},
138 : {"julian", UNITS, DTK_JULIAN},
139 : {"july", MONTH, 7},
140 : {"jun", MONTH, 6},
141 : {"june", MONTH, 6},
142 : {"m", UNITS, DTK_MONTH}, /* "month" for ISO input */
143 : {"mar", MONTH, 3},
144 : {"march", MONTH, 3},
145 : {"may", MONTH, 5},
146 : {"mm", UNITS, DTK_MINUTE}, /* "minute" for ISO input */
147 : {"mon", DOW, 1},
148 : {"monday", DOW, 1},
149 : {"nov", MONTH, 11},
150 : {"november", MONTH, 11},
151 : {NOW, RESERV, DTK_NOW}, /* current transaction time */
152 : {"oct", MONTH, 10},
153 : {"october", MONTH, 10},
154 : {"on", IGNORE_DTF, 0}, /* "on" (throwaway) */
155 : {"pm", AMPM, PM},
156 : {"s", UNITS, DTK_SECOND}, /* "seconds" for ISO input */
157 : {"sat", DOW, 6},
158 : {"saturday", DOW, 6},
159 : {"sep", MONTH, 9},
160 : {"sept", MONTH, 9},
161 : {"september", MONTH, 9},
162 : {"sun", DOW, 0},
163 : {"sunday", DOW, 0},
164 : {"t", ISOTIME, DTK_TIME}, /* Filler for ISO time fields */
165 : {"thu", DOW, 4},
166 : {"thur", DOW, 4},
167 : {"thurs", DOW, 4},
168 : {"thursday", DOW, 4},
169 : {TODAY, RESERV, DTK_TODAY}, /* midnight */
170 : {TOMORROW, RESERV, DTK_TOMORROW}, /* tomorrow midnight */
171 : {"tue", DOW, 2},
172 : {"tues", DOW, 2},
173 : {"tuesday", DOW, 2},
174 : {"wed", DOW, 3},
175 : {"wednesday", DOW, 3},
176 : {"weds", DOW, 3},
177 : {"y", UNITS, DTK_YEAR}, /* "year" for ISO input */
178 : {YESTERDAY, RESERV, DTK_YESTERDAY} /* yesterday midnight */
179 : };
180 :
181 : static const int szdatetktbl = sizeof datetktbl / sizeof datetktbl[0];
182 :
183 : /*
184 : * deltatktbl: same format as datetktbl, but holds keywords used to represent
185 : * time units (eg, for intervals, and for EXTRACT).
186 : */
187 : static const datetkn deltatktbl[] = {
188 : /* token, type, value */
189 : {"@", IGNORE_DTF, 0}, /* postgres relative prefix */
190 : {DAGO, AGO, 0}, /* "ago" indicates negative time offset */
191 : {"c", UNITS, DTK_CENTURY}, /* "century" relative */
192 : {"cent", UNITS, DTK_CENTURY}, /* "century" relative */
193 : {"centuries", UNITS, DTK_CENTURY}, /* "centuries" relative */
194 : {DCENTURY, UNITS, DTK_CENTURY}, /* "century" relative */
195 : {"d", UNITS, DTK_DAY}, /* "day" relative */
196 : {DDAY, UNITS, DTK_DAY}, /* "day" relative */
197 : {"days", UNITS, DTK_DAY}, /* "days" relative */
198 : {"dec", UNITS, DTK_DECADE}, /* "decade" relative */
199 : {DDECADE, UNITS, DTK_DECADE}, /* "decade" relative */
200 : {"decades", UNITS, DTK_DECADE}, /* "decades" relative */
201 : {"decs", UNITS, DTK_DECADE}, /* "decades" relative */
202 : {"h", UNITS, DTK_HOUR}, /* "hour" relative */
203 : {DHOUR, UNITS, DTK_HOUR}, /* "hour" relative */
204 : {"hours", UNITS, DTK_HOUR}, /* "hours" relative */
205 : {"hr", UNITS, DTK_HOUR}, /* "hour" relative */
206 : {"hrs", UNITS, DTK_HOUR}, /* "hours" relative */
207 : {"m", UNITS, DTK_MINUTE}, /* "minute" relative */
208 : {"microsecon", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
209 : {"mil", UNITS, DTK_MILLENNIUM}, /* "millennium" relative */
210 : {"millennia", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */
211 : {DMILLENNIUM, UNITS, DTK_MILLENNIUM}, /* "millennium" relative */
212 : {"millisecon", UNITS, DTK_MILLISEC}, /* relative */
213 : {"mils", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */
214 : {"min", UNITS, DTK_MINUTE}, /* "minute" relative */
215 : {"mins", UNITS, DTK_MINUTE}, /* "minutes" relative */
216 : {DMINUTE, UNITS, DTK_MINUTE}, /* "minute" relative */
217 : {"minutes", UNITS, DTK_MINUTE}, /* "minutes" relative */
218 : {"mon", UNITS, DTK_MONTH}, /* "months" relative */
219 : {"mons", UNITS, DTK_MONTH}, /* "months" relative */
220 : {DMONTH, UNITS, DTK_MONTH}, /* "month" relative */
221 : {"months", UNITS, DTK_MONTH},
222 : {"ms", UNITS, DTK_MILLISEC},
223 : {"msec", UNITS, DTK_MILLISEC},
224 : {DMILLISEC, UNITS, DTK_MILLISEC},
225 : {"mseconds", UNITS, DTK_MILLISEC},
226 : {"msecs", UNITS, DTK_MILLISEC},
227 : {"qtr", UNITS, DTK_QUARTER}, /* "quarter" relative */
228 : {DQUARTER, UNITS, DTK_QUARTER}, /* "quarter" relative */
229 : {"s", UNITS, DTK_SECOND},
230 : {"sec", UNITS, DTK_SECOND},
231 : {DSECOND, UNITS, DTK_SECOND},
232 : {"seconds", UNITS, DTK_SECOND},
233 : {"secs", UNITS, DTK_SECOND},
234 : {DTIMEZONE, UNITS, DTK_TZ}, /* "timezone" time offset */
235 : {"timezone_h", UNITS, DTK_TZ_HOUR}, /* timezone hour units */
236 : {"timezone_m", UNITS, DTK_TZ_MINUTE}, /* timezone minutes units */
237 : {"us", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
238 : {"usec", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
239 : {DMICROSEC, UNITS, DTK_MICROSEC}, /* "microsecond" relative */
240 : {"useconds", UNITS, DTK_MICROSEC}, /* "microseconds" relative */
241 : {"usecs", UNITS, DTK_MICROSEC}, /* "microseconds" relative */
242 : {"w", UNITS, DTK_WEEK}, /* "week" relative */
243 : {DWEEK, UNITS, DTK_WEEK}, /* "week" relative */
244 : {"weeks", UNITS, DTK_WEEK}, /* "weeks" relative */
245 : {"y", UNITS, DTK_YEAR}, /* "year" relative */
246 : {DYEAR, UNITS, DTK_YEAR}, /* "year" relative */
247 : {"years", UNITS, DTK_YEAR}, /* "years" relative */
248 : {"yr", UNITS, DTK_YEAR}, /* "year" relative */
249 : {"yrs", UNITS, DTK_YEAR} /* "years" relative */
250 : };
251 :
252 : static const int szdeltatktbl = sizeof deltatktbl / sizeof deltatktbl[0];
253 :
254 : static TimeZoneAbbrevTable *zoneabbrevtbl = NULL;
255 :
256 : /* Caches of recent lookup results in the above tables */
257 :
258 : static const datetkn *datecache[MAXDATEFIELDS] = {NULL};
259 :
260 : static const datetkn *deltacache[MAXDATEFIELDS] = {NULL};
261 :
262 : static const datetkn *abbrevcache[MAXDATEFIELDS] = {NULL};
263 :
264 :
265 : /*
266 : * Calendar time to Julian date conversions.
267 : * Julian date is commonly used in astronomical applications,
268 : * since it is numerically accurate and computationally simple.
269 : * The algorithms here will accurately convert between Julian day
270 : * and calendar date for all non-negative Julian days
271 : * (i.e. from Nov 24, -4713 on).
272 : *
273 : * Rewritten to eliminate overflow problems. This now allows the
274 : * routines to work correctly for all Julian day counts from
275 : * 0 to 2147483647 (Nov 24, -4713 to Jun 3, 5874898) assuming
276 : * a 32-bit integer. Longer types should also work to the limits
277 : * of their precision.
278 : *
279 : * Actually, date2j() will work sanely, in the sense of producing
280 : * valid negative Julian dates, significantly before Nov 24, -4713.
281 : * We rely on it to do so back to Nov 1, -4713; see IS_VALID_JULIAN()
282 : * and associated commentary in timestamp.h.
283 : */
284 :
285 : int
286 344882 : date2j(int year, int month, int day)
287 : {
288 : int julian;
289 : int century;
290 :
291 344882 : if (month > 2)
292 : {
293 185788 : month += 1;
294 185788 : year += 4800;
295 : }
296 : else
297 : {
298 159094 : month += 13;
299 159094 : year += 4799;
300 : }
301 :
302 344882 : century = year / 100;
303 344882 : julian = year * 365 - 32167;
304 344882 : julian += year / 4 - century + century / 4;
305 344882 : julian += 7834 * month / 256 + day;
306 :
307 344882 : return julian;
308 : } /* date2j() */
309 :
310 : void
311 274854 : j2date(int jd, int *year, int *month, int *day)
312 : {
313 : unsigned int julian;
314 : unsigned int quad;
315 : unsigned int extra;
316 : int y;
317 :
318 274854 : julian = jd;
319 274854 : julian += 32044;
320 274854 : quad = julian / 146097;
321 274854 : extra = (julian - quad * 146097) * 4 + 3;
322 274854 : julian += 60 + quad * 3 + extra / 146097;
323 274854 : quad = julian / 1461;
324 274854 : julian -= quad * 1461;
325 274854 : y = julian * 4 / 1461;
326 549708 : julian = ((y != 0) ? ((julian + 305) % 365) : ((julian + 306) % 366))
327 274854 : + 123;
328 274854 : y += quad * 4;
329 274854 : *year = y - 4800;
330 274854 : quad = julian * 2141 / 65536;
331 274854 : *day = julian - 7834 * quad / 256;
332 274854 : *month = (quad + 10) % MONTHS_PER_YEAR + 1;
333 274854 : } /* j2date() */
334 :
335 :
336 : /*
337 : * j2day - convert Julian date to day-of-week (0..6 == Sun..Sat)
338 : *
339 : * Note: various places use the locution j2day(date - 1) to produce a
340 : * result according to the convention 0..6 = Mon..Sun. This is a bit of
341 : * a crock, but will work as long as the computation here is just a modulo.
342 : */
343 : int
344 51522 : j2day(int date)
345 : {
346 51522 : date += 1;
347 51522 : date %= 7;
348 : /* Cope if division truncates towards zero, as it probably does */
349 51522 : if (date < 0)
350 0 : date += 7;
351 :
352 51522 : return date;
353 : } /* j2day() */
354 :
355 :
356 : /*
357 : * GetCurrentDateTime()
358 : *
359 : * Get the transaction start time ("now()") broken down as a struct pg_tm,
360 : * converted according to the session timezone setting.
361 : *
362 : * This is just a convenience wrapper for GetCurrentTimeUsec, to cover the
363 : * case where caller doesn't need either fractional seconds or tz offset.
364 : */
365 : void
366 2810 : GetCurrentDateTime(struct pg_tm *tm)
367 : {
368 : fsec_t fsec;
369 :
370 2810 : GetCurrentTimeUsec(tm, &fsec, NULL);
371 2810 : }
372 :
373 : /*
374 : * GetCurrentTimeUsec()
375 : *
376 : * Get the transaction start time ("now()") broken down as a struct pg_tm,
377 : * including fractional seconds and timezone offset. The time is converted
378 : * according to the session timezone setting.
379 : *
380 : * Callers may pass tzp = NULL if they don't need the offset, but this does
381 : * not affect the conversion behavior (unlike timestamp2tm()).
382 : *
383 : * Internally, we cache the result, since this could be called many times
384 : * in a transaction, within which now() doesn't change.
385 : */
386 : void
387 2984 : GetCurrentTimeUsec(struct pg_tm *tm, fsec_t *fsec, int *tzp)
388 : {
389 2984 : TimestampTz cur_ts = GetCurrentTransactionStartTimestamp();
390 :
391 : /*
392 : * The cache key must include both current time and current timezone. By
393 : * representing the timezone by just a pointer, we're assuming that
394 : * distinct timezone settings could never have the same pointer value.
395 : * This is true by virtue of the hashtable used inside pg_tzset();
396 : * however, it might need another look if we ever allow entries in that
397 : * hash to be recycled.
398 : */
399 : static TimestampTz cache_ts = 0;
400 : static pg_tz *cache_timezone = NULL;
401 : static struct pg_tm cache_tm;
402 : static fsec_t cache_fsec;
403 : static int cache_tz;
404 :
405 2984 : if (cur_ts != cache_ts || session_timezone != cache_timezone)
406 : {
407 : /*
408 : * Make sure cache is marked invalid in case of error after partial
409 : * update within timestamp2tm.
410 : */
411 964 : cache_timezone = NULL;
412 :
413 : /*
414 : * Perform the computation, storing results into cache. We do not
415 : * really expect any error here, since current time surely ought to be
416 : * within range, but check just for sanity's sake.
417 : */
418 964 : if (timestamp2tm(cur_ts, &cache_tz, &cache_tm, &cache_fsec,
419 : NULL, session_timezone) != 0)
420 0 : ereport(ERROR,
421 : (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
422 : errmsg("timestamp out of range")));
423 :
424 : /* OK, so mark the cache valid. */
425 964 : cache_ts = cur_ts;
426 964 : cache_timezone = session_timezone;
427 : }
428 :
429 2984 : *tm = cache_tm;
430 2984 : *fsec = cache_fsec;
431 2984 : if (tzp != NULL)
432 162 : *tzp = cache_tz;
433 2984 : }
434 :
435 :
436 : /*
437 : * Append seconds and fractional seconds (if any) at *cp.
438 : *
439 : * precision is the max number of fraction digits, fillzeros says to
440 : * pad to two integral-seconds digits.
441 : *
442 : * Returns a pointer to the new end of string. No NUL terminator is put
443 : * there; callers are responsible for NUL terminating str themselves.
444 : *
445 : * Note that any sign is stripped from the input sec and fsec values.
446 : */
447 : static char *
448 133626 : AppendSeconds(char *cp, int sec, fsec_t fsec, int precision, bool fillzeros)
449 : {
450 : Assert(precision >= 0);
451 :
452 133626 : if (fillzeros)
453 130314 : cp = pg_ultostr_zeropad(cp, abs(sec), 2);
454 : else
455 3312 : cp = pg_ultostr(cp, abs(sec));
456 :
457 : /* fsec_t is just an int32 */
458 133626 : if (fsec != 0)
459 : {
460 21966 : int32 value = abs(fsec);
461 21966 : char *end = &cp[precision + 1];
462 21966 : bool gotnonzero = false;
463 :
464 21966 : *cp++ = '.';
465 :
466 : /*
467 : * Append the fractional seconds part. Note that we don't want any
468 : * trailing zeros here, so since we're building the number in reverse
469 : * we'll skip appending zeros until we've output a non-zero digit.
470 : */
471 153762 : while (precision--)
472 : {
473 131796 : int32 oldval = value;
474 : int32 remainder;
475 :
476 131796 : value /= 10;
477 131796 : remainder = oldval - value * 10;
478 :
479 : /* check if we got a non-zero */
480 131796 : if (remainder)
481 82632 : gotnonzero = true;
482 :
483 131796 : if (gotnonzero)
484 99774 : cp[precision] = '0' + remainder;
485 : else
486 32022 : end = &cp[precision];
487 : }
488 :
489 : /*
490 : * If we still have a non-zero value then precision must have not been
491 : * enough to print the number. We punt the problem to pg_ultostr(),
492 : * which will generate a correct answer in the minimum valid width.
493 : */
494 21966 : if (value)
495 0 : return pg_ultostr(cp, abs(fsec));
496 :
497 21966 : return end;
498 : }
499 : else
500 111660 : return cp;
501 : }
502 :
503 :
504 : /*
505 : * Variant of above that's specialized to timestamp case.
506 : *
507 : * Returns a pointer to the new end of string. No NUL terminator is put
508 : * there; callers are responsible for NUL terminating str themselves.
509 : */
510 : static char *
511 113750 : AppendTimestampSeconds(char *cp, struct pg_tm *tm, fsec_t fsec)
512 : {
513 113750 : return AppendSeconds(cp, tm->tm_sec, fsec, MAX_TIMESTAMP_PRECISION, true);
514 : }
515 :
516 :
517 : /*
518 : * Add val * multiplier to *sum.
519 : * Returns true if successful, false on overflow.
520 : */
521 : static bool
522 8776 : int64_multiply_add(int64 val, int64 multiplier, int64 *sum)
523 : {
524 : int64 product;
525 :
526 17492 : if (pg_mul_s64_overflow(val, multiplier, &product) ||
527 8716 : pg_add_s64_overflow(*sum, product, sum))
528 156 : return false;
529 8620 : return true;
530 : }
531 :
532 : /*
533 : * Multiply frac by scale (to produce microseconds) and add to itm_in->tm_usec.
534 : * Returns true if successful, false if itm_in overflows.
535 : */
536 : static bool
537 9918 : AdjustFractMicroseconds(double frac, int64 scale,
538 : struct pg_itm_in *itm_in)
539 : {
540 : int64 usec;
541 :
542 : /* Fast path for common case */
543 9918 : if (frac == 0)
544 9456 : return true;
545 :
546 : /*
547 : * We assume the input frac has abs value less than 1, so overflow of frac
548 : * or usec is not an issue for interesting values of scale.
549 : */
550 462 : frac *= scale;
551 462 : usec = (int64) frac;
552 :
553 : /* Round off any fractional microsecond */
554 462 : frac -= usec;
555 462 : if (frac > 0.5)
556 24 : usec++;
557 438 : else if (frac < -0.5)
558 30 : usec--;
559 :
560 462 : return !pg_add_s64_overflow(itm_in->tm_usec, usec, &itm_in->tm_usec);
561 : }
562 :
563 : /*
564 : * Multiply frac by scale (to produce days). Add the integral part of the
565 : * result to itm_in->tm_mday, the fractional part to itm_in->tm_usec.
566 : * Returns true if successful, false if itm_in overflows.
567 : */
568 : static bool
569 1104 : AdjustFractDays(double frac, int scale,
570 : struct pg_itm_in *itm_in)
571 : {
572 : int extra_days;
573 :
574 : /* Fast path for common case */
575 1104 : if (frac == 0)
576 918 : return true;
577 :
578 : /*
579 : * We assume the input frac has abs value less than 1, so overflow of frac
580 : * or extra_days is not an issue.
581 : */
582 186 : frac *= scale;
583 186 : extra_days = (int) frac;
584 :
585 : /* ... but this could overflow, if tm_mday is already nonzero */
586 186 : if (pg_add_s32_overflow(itm_in->tm_mday, extra_days, &itm_in->tm_mday))
587 48 : return false;
588 :
589 : /* Handle any fractional day */
590 138 : frac -= extra_days;
591 138 : return AdjustFractMicroseconds(frac, USECS_PER_DAY, itm_in);
592 : }
593 :
594 : /*
595 : * Multiply frac by scale (to produce years), then further scale up to months.
596 : * Add the integral part of the result to itm_in->tm_mon, discarding any
597 : * fractional part.
598 : * Returns true if successful, false if itm_in overflows.
599 : */
600 : static bool
601 1220 : AdjustFractYears(double frac, int scale,
602 : struct pg_itm_in *itm_in)
603 : {
604 : /*
605 : * As above, we assume abs(frac) < 1, so this can't overflow for any
606 : * interesting value of scale.
607 : */
608 1220 : int extra_months = (int) rint(frac * scale * MONTHS_PER_YEAR);
609 :
610 1220 : return !pg_add_s32_overflow(itm_in->tm_mon, extra_months, &itm_in->tm_mon);
611 : }
612 :
613 : /*
614 : * Add (val + fval) * scale to itm_in->tm_usec.
615 : * Returns true if successful, false if itm_in overflows.
616 : */
617 : static bool
618 2806 : AdjustMicroseconds(int64 val, double fval, int64 scale,
619 : struct pg_itm_in *itm_in)
620 : {
621 : /* Handle the integer part */
622 2806 : if (!int64_multiply_add(val, scale, &itm_in->tm_usec))
623 150 : return false;
624 : /* Handle the float part */
625 2656 : return AdjustFractMicroseconds(fval, scale, itm_in);
626 : }
627 :
628 : /*
629 : * Multiply val by scale (to produce days) and add to itm_in->tm_mday.
630 : * Returns true if successful, false if itm_in overflows.
631 : */
632 : static bool
633 7382 : AdjustDays(int64 val, int scale, struct pg_itm_in *itm_in)
634 : {
635 : int days;
636 :
637 7382 : if (val < INT_MIN || val > INT_MAX)
638 36 : return false;
639 14680 : return !pg_mul_s32_overflow((int32) val, scale, &days) &&
640 7334 : !pg_add_s32_overflow(itm_in->tm_mday, days, &itm_in->tm_mday);
641 : }
642 :
643 : /*
644 : * Add val to itm_in->tm_mon (no need for scale here, as val is always
645 : * in months already).
646 : * Returns true if successful, false if itm_in overflows.
647 : */
648 : static bool
649 1074 : AdjustMonths(int64 val, struct pg_itm_in *itm_in)
650 : {
651 1074 : if (val < INT_MIN || val > INT_MAX)
652 12 : return false;
653 1062 : return !pg_add_s32_overflow(itm_in->tm_mon, (int32) val, &itm_in->tm_mon);
654 : }
655 :
656 : /*
657 : * Multiply val by scale (to produce years) and add to itm_in->tm_year.
658 : * Returns true if successful, false if itm_in overflows.
659 : */
660 : static bool
661 1358 : AdjustYears(int64 val, int scale,
662 : struct pg_itm_in *itm_in)
663 : {
664 : int years;
665 :
666 1358 : if (val < INT_MIN || val > INT_MAX)
667 24 : return false;
668 2632 : return !pg_mul_s32_overflow((int32) val, scale, &years) &&
669 1298 : !pg_add_s32_overflow(itm_in->tm_year, years, &itm_in->tm_year);
670 : }
671 :
672 :
673 : /*
674 : * Parse the fractional part of a number (decimal point and optional digits,
675 : * followed by end of string). Returns the fractional value into *frac.
676 : *
677 : * Returns 0 if successful, DTERR code if bogus input detected.
678 : */
679 : static int
680 23202 : ParseFraction(char *cp, double *frac)
681 : {
682 : /* Caller should always pass the start of the fraction part */
683 : Assert(*cp == '.');
684 :
685 : /*
686 : * We want to allow just "." with no digits, but some versions of strtod
687 : * will report EINVAL for that, so special-case it.
688 : */
689 23202 : if (cp[1] == '\0')
690 : {
691 0 : *frac = 0;
692 : }
693 : else
694 : {
695 23202 : errno = 0;
696 23202 : *frac = strtod(cp, &cp);
697 : /* check for parse failure */
698 23202 : if (*cp != '\0' || errno != 0)
699 12 : return DTERR_BAD_FORMAT;
700 : }
701 23190 : return 0;
702 : }
703 :
704 : /*
705 : * Fetch a fractional-second value with suitable error checking.
706 : * Same as ParseFraction except we convert the result to integer microseconds.
707 : */
708 : static int
709 22704 : ParseFractionalSecond(char *cp, fsec_t *fsec)
710 : {
711 : double frac;
712 : int dterr;
713 :
714 22704 : dterr = ParseFraction(cp, &frac);
715 22704 : if (dterr)
716 12 : return dterr;
717 22692 : *fsec = rint(frac * 1000000);
718 22692 : return 0;
719 : }
720 :
721 :
722 : /* ParseDateTime()
723 : * Break string into tokens based on a date/time context.
724 : * Returns 0 if successful, DTERR code if bogus input detected.
725 : *
726 : * timestr - the input string
727 : * workbuf - workspace for field string storage. This must be
728 : * larger than the largest legal input for this datetime type --
729 : * some additional space will be needed to NUL terminate fields.
730 : * buflen - the size of workbuf
731 : * field[] - pointers to field strings are returned in this array
732 : * ftype[] - field type indicators are returned in this array
733 : * maxfields - dimensions of the above two arrays
734 : * *numfields - set to the actual number of fields detected
735 : *
736 : * The fields extracted from the input are stored as separate,
737 : * null-terminated strings in the workspace at workbuf. Any text is
738 : * converted to lower case.
739 : *
740 : * Several field types are assigned:
741 : * DTK_NUMBER - digits and (possibly) a decimal point
742 : * DTK_DATE - digits and two delimiters, or digits and text
743 : * DTK_TIME - digits, colon delimiters, and possibly a decimal point
744 : * DTK_STRING - text (no digits or punctuation)
745 : * DTK_SPECIAL - leading "+" or "-" followed by text
746 : * DTK_TZ - leading "+" or "-" followed by digits (also eats ':', '.', '-')
747 : *
748 : * Note that some field types can hold unexpected items:
749 : * DTK_NUMBER can hold date fields (yy.ddd)
750 : * DTK_STRING can hold months (January) and time zones (PST)
751 : * DTK_DATE can hold time zone names (America/New_York, GMT-8)
752 : */
753 : int
754 85868 : ParseDateTime(const char *timestr, char *workbuf, size_t buflen,
755 : char **field, int *ftype, int maxfields, int *numfields)
756 : {
757 85868 : int nf = 0;
758 85868 : const char *cp = timestr;
759 85868 : char *bufp = workbuf;
760 85868 : const char *bufend = workbuf + buflen;
761 :
762 : /*
763 : * Set the character pointed-to by "bufptr" to "newchar", and increment
764 : * "bufptr". "end" gives the end of the buffer -- we return an error if
765 : * there is no space left to append a character to the buffer. Note that
766 : * "bufptr" is evaluated twice.
767 : */
768 : #define APPEND_CHAR(bufptr, end, newchar) \
769 : do \
770 : { \
771 : if (((bufptr) + 1) >= (end)) \
772 : return DTERR_BAD_FORMAT; \
773 : *(bufptr)++ = newchar; \
774 : } while (0)
775 :
776 : /* outer loop through fields */
777 376532 : while (*cp != '\0')
778 : {
779 : /* Ignore spaces between fields */
780 290664 : if (isspace((unsigned char) *cp))
781 : {
782 83408 : cp++;
783 83408 : continue;
784 : }
785 :
786 : /* Record start of current field */
787 207256 : if (nf >= maxfields)
788 0 : return DTERR_BAD_FORMAT;
789 207256 : field[nf] = bufp;
790 :
791 : /* leading digit? then date or time */
792 207256 : if (isdigit((unsigned char) *cp))
793 : {
794 141452 : APPEND_CHAR(bufp, bufend, *cp++);
795 422992 : while (isdigit((unsigned char) *cp))
796 281540 : APPEND_CHAR(bufp, bufend, *cp++);
797 :
798 : /* time field? */
799 141452 : if (*cp == ':')
800 : {
801 60016 : ftype[nf] = DTK_TIME;
802 60016 : APPEND_CHAR(bufp, bufend, *cp++);
803 492228 : while (isdigit((unsigned char) *cp) ||
804 141056 : (*cp == ':') || (*cp == '.'))
805 432212 : APPEND_CHAR(bufp, bufend, *cp++);
806 : }
807 : /* date field? allow embedded text month */
808 81436 : else if (*cp == '-' || *cp == '/' || *cp == '.')
809 65732 : {
810 : /* save delimiting character to use later */
811 65732 : char delim = *cp;
812 :
813 65732 : APPEND_CHAR(bufp, bufend, *cp++);
814 : /* second field is all digits? then no embedded text month */
815 65732 : if (isdigit((unsigned char) *cp))
816 : {
817 65636 : ftype[nf] = ((delim == '.') ? DTK_NUMBER : DTK_DATE);
818 196944 : while (isdigit((unsigned char) *cp))
819 131308 : APPEND_CHAR(bufp, bufend, *cp++);
820 :
821 : /*
822 : * insist that the delimiters match to get a three-field
823 : * date.
824 : */
825 65636 : if (*cp == delim)
826 : {
827 65030 : ftype[nf] = DTK_DATE;
828 65030 : APPEND_CHAR(bufp, bufend, *cp++);
829 197944 : while (isdigit((unsigned char) *cp) || *cp == delim)
830 132914 : APPEND_CHAR(bufp, bufend, *cp++);
831 : }
832 : }
833 : else
834 : {
835 96 : ftype[nf] = DTK_DATE;
836 756 : while (isalnum((unsigned char) *cp) || *cp == delim)
837 660 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
838 : }
839 : }
840 :
841 : /*
842 : * otherwise, number only and will determine year, month, day, or
843 : * concatenated fields later...
844 : */
845 : else
846 15704 : ftype[nf] = DTK_NUMBER;
847 : }
848 : /* Leading decimal point? Then fractional seconds... */
849 65804 : else if (*cp == '.')
850 : {
851 0 : APPEND_CHAR(bufp, bufend, *cp++);
852 0 : while (isdigit((unsigned char) *cp))
853 0 : APPEND_CHAR(bufp, bufend, *cp++);
854 :
855 0 : ftype[nf] = DTK_NUMBER;
856 : }
857 :
858 : /*
859 : * text? then date string, month, day of week, special, or timezone
860 : */
861 65804 : else if (isalpha((unsigned char) *cp))
862 : {
863 : bool is_date;
864 :
865 24316 : ftype[nf] = DTK_STRING;
866 24316 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
867 97304 : while (isalpha((unsigned char) *cp))
868 72988 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
869 :
870 : /*
871 : * Dates can have embedded '-', '/', or '.' separators. It could
872 : * also be a timezone name containing embedded '/', '+', '-', '_',
873 : * or ':' (but '_' or ':' can't be the first punctuation). If the
874 : * next character is a digit or '+', we need to check whether what
875 : * we have so far is a recognized non-timezone keyword --- if so,
876 : * don't believe that this is the start of a timezone.
877 : */
878 24316 : is_date = false;
879 24316 : if (*cp == '-' || *cp == '/' || *cp == '.')
880 1510 : is_date = true;
881 22806 : else if (*cp == '+' || isdigit((unsigned char) *cp))
882 : {
883 1902 : *bufp = '\0'; /* null-terminate current field value */
884 : /* we need search only the core token table, not TZ names */
885 1902 : if (datebsearch(field[nf], datetktbl, szdatetktbl) == NULL)
886 1512 : is_date = true;
887 : }
888 24316 : if (is_date)
889 : {
890 3022 : ftype[nf] = DTK_DATE;
891 : do
892 : {
893 14158 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
894 14158 : } while (*cp == '+' || *cp == '-' ||
895 13876 : *cp == '/' || *cp == '_' ||
896 13720 : *cp == '.' || *cp == ':' ||
897 27470 : isalnum((unsigned char) *cp));
898 : }
899 : }
900 : /* sign? then special or numeric timezone */
901 41488 : else if (*cp == '+' || *cp == '-')
902 : {
903 40924 : APPEND_CHAR(bufp, bufend, *cp++);
904 : /* soak up leading whitespace */
905 40948 : while (isspace((unsigned char) *cp))
906 24 : cp++;
907 : /* numeric timezone? */
908 : /* note that "DTK_TZ" could also be a signed float or yyyy-mm */
909 40924 : if (isdigit((unsigned char) *cp))
910 : {
911 40058 : ftype[nf] = DTK_TZ;
912 40058 : APPEND_CHAR(bufp, bufend, *cp++);
913 93898 : while (isdigit((unsigned char) *cp) ||
914 41756 : *cp == ':' || *cp == '.' || *cp == '-')
915 53840 : APPEND_CHAR(bufp, bufend, *cp++);
916 : }
917 : /* special? */
918 866 : else if (isalpha((unsigned char) *cp))
919 : {
920 866 : ftype[nf] = DTK_SPECIAL;
921 866 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
922 6928 : while (isalpha((unsigned char) *cp))
923 6062 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
924 : }
925 : /* otherwise something wrong... */
926 : else
927 0 : return DTERR_BAD_FORMAT;
928 : }
929 : /* ignore other punctuation but use as delimiter */
930 564 : else if (ispunct((unsigned char) *cp))
931 : {
932 564 : cp++;
933 564 : continue;
934 : }
935 : /* otherwise, something is not right... */
936 : else
937 0 : return DTERR_BAD_FORMAT;
938 :
939 : /* force in a delimiter after each field */
940 206692 : *bufp++ = '\0';
941 206692 : nf++;
942 : }
943 :
944 85868 : *numfields = nf;
945 :
946 85868 : return 0;
947 : }
948 :
949 :
950 : /* DecodeDateTime()
951 : * Interpret previously parsed fields for general date and time.
952 : * Return 0 if full date, 1 if only time, and negative DTERR code if problems.
953 : * (Currently, all callers treat 1 as an error return too.)
954 : *
955 : * Inputs are field[] and ftype[] arrays, of length nf.
956 : * Other arguments are outputs.
957 : *
958 : * External format(s):
959 : * "<weekday> <month>-<day>-<year> <hour>:<minute>:<second>"
960 : * "Fri Feb-7-1997 15:23:27"
961 : * "Feb-7-1997 15:23:27"
962 : * "2-7-1997 15:23:27"
963 : * "1997-2-7 15:23:27"
964 : * "1997.038 15:23:27" (day of year 1-366)
965 : * Also supports input in compact time:
966 : * "970207 152327"
967 : * "97038 152327"
968 : * "20011225T040506.789-07"
969 : *
970 : * Use the system-provided functions to get the current time zone
971 : * if not specified in the input string.
972 : *
973 : * If the date is outside the range of pg_time_t (in practice that could only
974 : * happen if pg_time_t is just 32 bits), then assume UTC time zone - thomas
975 : * 1997-05-27
976 : */
977 : int
978 69776 : DecodeDateTime(char **field, int *ftype, int nf,
979 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
980 : DateTimeErrorExtra *extra)
981 : {
982 69776 : int fmask = 0,
983 : tmask,
984 : type;
985 69776 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
986 : int i;
987 : int val;
988 : int dterr;
989 69776 : int mer = HR24;
990 69776 : bool haveTextMonth = false;
991 69776 : bool isjulian = false;
992 69776 : bool is2digits = false;
993 69776 : bool bc = false;
994 69776 : pg_tz *namedTz = NULL;
995 69776 : pg_tz *abbrevTz = NULL;
996 : pg_tz *valtz;
997 69776 : char *abbrev = NULL;
998 : struct pg_tm cur_tm;
999 :
1000 : /*
1001 : * We'll insist on at least all of the date fields, but initialize the
1002 : * remaining fields in case they are not set later...
1003 : */
1004 69776 : *dtype = DTK_DATE;
1005 69776 : tm->tm_hour = 0;
1006 69776 : tm->tm_min = 0;
1007 69776 : tm->tm_sec = 0;
1008 69776 : *fsec = 0;
1009 : /* don't know daylight savings time status apriori */
1010 69776 : tm->tm_isdst = -1;
1011 69776 : if (tzp != NULL)
1012 69776 : *tzp = 0;
1013 :
1014 242746 : for (i = 0; i < nf; i++)
1015 : {
1016 173282 : switch (ftype[i])
1017 : {
1018 66386 : case DTK_DATE:
1019 :
1020 : /*
1021 : * Integral julian day with attached time zone? All other
1022 : * forms with JD will be separated into distinct fields, so we
1023 : * handle just this case here.
1024 : */
1025 66386 : if (ptype == DTK_JULIAN)
1026 : {
1027 : char *cp;
1028 : int jday;
1029 :
1030 6 : if (tzp == NULL)
1031 0 : return DTERR_BAD_FORMAT;
1032 :
1033 6 : errno = 0;
1034 6 : jday = strtoint(field[i], &cp, 10);
1035 6 : if (errno == ERANGE || jday < 0)
1036 0 : return DTERR_FIELD_OVERFLOW;
1037 :
1038 6 : j2date(jday, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1039 6 : isjulian = true;
1040 :
1041 : /* Get the time zone from the end of the string */
1042 6 : dterr = DecodeTimezone(cp, tzp);
1043 6 : if (dterr)
1044 0 : return dterr;
1045 :
1046 6 : tmask = DTK_DATE_M | DTK_TIME_M | DTK_M(TZ);
1047 6 : ptype = 0;
1048 6 : break;
1049 : }
1050 :
1051 : /*
1052 : * Already have a date? Then this might be a time zone name
1053 : * with embedded punctuation (e.g. "America/New_York") or a
1054 : * run-together time with trailing time zone (e.g. hhmmss-zz).
1055 : * - thomas 2001-12-25
1056 : *
1057 : * We consider it a time zone if we already have month & day.
1058 : * This is to allow the form "mmm dd hhmmss tz year", which
1059 : * we've historically accepted.
1060 : */
1061 66380 : else if (ptype != 0 ||
1062 66368 : ((fmask & (DTK_M(MONTH) | DTK_M(DAY))) ==
1063 : (DTK_M(MONTH) | DTK_M(DAY))))
1064 : {
1065 : /* No time zone accepted? Then quit... */
1066 1416 : if (tzp == NULL)
1067 0 : return DTERR_BAD_FORMAT;
1068 :
1069 1416 : if (isdigit((unsigned char) *field[i]) || ptype != 0)
1070 18 : {
1071 : char *cp;
1072 :
1073 : /*
1074 : * Allow a preceding "t" field, but no other units.
1075 : */
1076 18 : if (ptype != 0)
1077 : {
1078 : /* Sanity check; should not fail this test */
1079 12 : if (ptype != DTK_TIME)
1080 0 : return DTERR_BAD_FORMAT;
1081 12 : ptype = 0;
1082 : }
1083 :
1084 : /*
1085 : * Starts with a digit but we already have a time
1086 : * field? Then we are in trouble with a date and time
1087 : * already...
1088 : */
1089 18 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1090 0 : return DTERR_BAD_FORMAT;
1091 :
1092 18 : if ((cp = strchr(field[i], '-')) == NULL)
1093 0 : return DTERR_BAD_FORMAT;
1094 :
1095 : /* Get the time zone from the end of the string */
1096 18 : dterr = DecodeTimezone(cp, tzp);
1097 18 : if (dterr)
1098 0 : return dterr;
1099 18 : *cp = '\0';
1100 :
1101 : /*
1102 : * Then read the rest of the field as a concatenated
1103 : * time
1104 : */
1105 18 : dterr = DecodeNumberField(strlen(field[i]), field[i],
1106 : fmask,
1107 : &tmask, tm,
1108 : fsec, &is2digits);
1109 18 : if (dterr < 0)
1110 0 : return dterr;
1111 :
1112 : /*
1113 : * modify tmask after returning from
1114 : * DecodeNumberField()
1115 : */
1116 18 : tmask |= DTK_M(TZ);
1117 : }
1118 : else
1119 : {
1120 1398 : namedTz = pg_tzset(field[i]);
1121 1398 : if (!namedTz)
1122 : {
1123 36 : extra->dtee_timezone = field[i];
1124 36 : return DTERR_BAD_TIMEZONE;
1125 : }
1126 : /* we'll apply the zone setting below */
1127 1362 : tmask = DTK_M(TZ);
1128 : }
1129 : }
1130 : else
1131 : {
1132 64964 : dterr = DecodeDate(field[i], fmask,
1133 : &tmask, &is2digits, tm);
1134 64964 : if (dterr)
1135 36 : return dterr;
1136 : }
1137 66308 : break;
1138 :
1139 54686 : case DTK_TIME:
1140 :
1141 : /*
1142 : * This might be an ISO time following a "t" field.
1143 : */
1144 54686 : if (ptype != 0)
1145 : {
1146 : /* Sanity check; should not fail this test */
1147 12 : if (ptype != DTK_TIME)
1148 0 : return DTERR_BAD_FORMAT;
1149 12 : ptype = 0;
1150 : }
1151 54686 : dterr = DecodeTime(field[i], fmask, INTERVAL_FULL_RANGE,
1152 : &tmask, tm, fsec);
1153 54686 : if (dterr)
1154 0 : return dterr;
1155 :
1156 : /* check for time overflow */
1157 54686 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec,
1158 : *fsec))
1159 0 : return DTERR_FIELD_OVERFLOW;
1160 54686 : break;
1161 :
1162 36682 : case DTK_TZ:
1163 : {
1164 : int tz;
1165 :
1166 36682 : if (tzp == NULL)
1167 12 : return DTERR_BAD_FORMAT;
1168 :
1169 36682 : dterr = DecodeTimezone(field[i], &tz);
1170 36682 : if (dterr)
1171 12 : return dterr;
1172 36670 : *tzp = tz;
1173 36670 : tmask = DTK_M(TZ);
1174 : }
1175 36670 : break;
1176 :
1177 6280 : case DTK_NUMBER:
1178 :
1179 : /*
1180 : * Deal with cases where previous field labeled this one
1181 : */
1182 6280 : if (ptype != 0)
1183 : {
1184 : char *cp;
1185 : int value;
1186 :
1187 132 : errno = 0;
1188 132 : value = strtoint(field[i], &cp, 10);
1189 132 : if (errno == ERANGE)
1190 12 : return DTERR_FIELD_OVERFLOW;
1191 132 : if (*cp != '.' && *cp != '\0')
1192 0 : return DTERR_BAD_FORMAT;
1193 :
1194 : switch (ptype)
1195 : {
1196 84 : case DTK_JULIAN:
1197 : /* previous field was a label for "julian date" */
1198 84 : if (value < 0)
1199 0 : return DTERR_FIELD_OVERFLOW;
1200 84 : tmask = DTK_DATE_M;
1201 84 : j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1202 84 : isjulian = true;
1203 :
1204 : /* fractional Julian Day? */
1205 84 : if (*cp == '.')
1206 : {
1207 : double time;
1208 :
1209 12 : dterr = ParseFraction(cp, &time);
1210 12 : if (dterr)
1211 0 : return dterr;
1212 12 : time *= USECS_PER_DAY;
1213 12 : dt2time(time,
1214 : &tm->tm_hour, &tm->tm_min,
1215 : &tm->tm_sec, fsec);
1216 12 : tmask |= DTK_TIME_M;
1217 : }
1218 84 : break;
1219 :
1220 36 : case DTK_TIME:
1221 : /* previous field was "t" for ISO time */
1222 36 : dterr = DecodeNumberField(strlen(field[i]), field[i],
1223 : (fmask | DTK_DATE_M),
1224 : &tmask, tm,
1225 : fsec, &is2digits);
1226 36 : if (dterr < 0)
1227 0 : return dterr;
1228 36 : if (tmask != DTK_TIME_M)
1229 0 : return DTERR_BAD_FORMAT;
1230 36 : break;
1231 :
1232 12 : default:
1233 12 : return DTERR_BAD_FORMAT;
1234 : break;
1235 : }
1236 :
1237 120 : ptype = 0;
1238 120 : *dtype = DTK_DATE;
1239 : }
1240 : else
1241 : {
1242 : char *cp;
1243 : int flen;
1244 :
1245 6148 : flen = strlen(field[i]);
1246 6148 : cp = strchr(field[i], '.');
1247 :
1248 : /* Embedded decimal and no date yet? */
1249 6148 : if (cp != NULL && !(fmask & DTK_DATE_M))
1250 : {
1251 30 : dterr = DecodeDate(field[i], fmask,
1252 : &tmask, &is2digits, tm);
1253 30 : if (dterr)
1254 0 : return dterr;
1255 : }
1256 : /* embedded decimal and several digits before? */
1257 6118 : else if (cp != NULL && flen - strlen(cp) > 2)
1258 : {
1259 : /*
1260 : * Interpret as a concatenated date or time Set the
1261 : * type field to allow decoding other fields later.
1262 : * Example: 20011223 or 040506
1263 : */
1264 12 : dterr = DecodeNumberField(flen, field[i], fmask,
1265 : &tmask, tm,
1266 : fsec, &is2digits);
1267 12 : if (dterr < 0)
1268 0 : return dterr;
1269 : }
1270 :
1271 : /*
1272 : * Is this a YMD or HMS specification, or a year number?
1273 : * YMD and HMS are required to be six digits or more, so
1274 : * if it is 5 digits, it is a year. If it is six or more
1275 : * digits, we assume it is YMD or HMS unless no date and
1276 : * no time values have been specified. This forces 6+
1277 : * digit years to be at the end of the string, or to use
1278 : * the ISO date specification.
1279 : */
1280 6106 : else if (flen >= 6 && (!(fmask & DTK_DATE_M) ||
1281 96 : !(fmask & DTK_TIME_M)))
1282 : {
1283 338 : dterr = DecodeNumberField(flen, field[i], fmask,
1284 : &tmask, tm,
1285 : fsec, &is2digits);
1286 338 : if (dterr < 0)
1287 0 : return dterr;
1288 : }
1289 : /* otherwise it is a single date/time field... */
1290 : else
1291 : {
1292 5768 : dterr = DecodeNumber(flen, field[i],
1293 : haveTextMonth, fmask,
1294 : &tmask, tm,
1295 : fsec, &is2digits);
1296 5768 : if (dterr)
1297 12 : return dterr;
1298 : }
1299 : }
1300 6256 : break;
1301 :
1302 9248 : case DTK_STRING:
1303 : case DTK_SPECIAL:
1304 : /* timezone abbrevs take precedence over built-in tokens */
1305 9248 : dterr = DecodeTimezoneAbbrev(i, field[i],
1306 : &type, &val, &valtz, extra);
1307 9248 : if (dterr)
1308 0 : return dterr;
1309 9248 : if (type == UNKNOWN_FIELD)
1310 6838 : type = DecodeSpecial(i, field[i], &val);
1311 9248 : if (type == IGNORE_DTF)
1312 0 : continue;
1313 :
1314 9248 : tmask = DTK_M(type);
1315 9248 : switch (type)
1316 : {
1317 1666 : case RESERV:
1318 1666 : switch (val)
1319 : {
1320 114 : case DTK_NOW:
1321 114 : tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1322 114 : *dtype = DTK_DATE;
1323 114 : GetCurrentTimeUsec(tm, fsec, tzp);
1324 114 : break;
1325 :
1326 102 : case DTK_YESTERDAY:
1327 102 : tmask = DTK_DATE_M;
1328 102 : *dtype = DTK_DATE;
1329 102 : GetCurrentDateTime(&cur_tm);
1330 102 : j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) - 1,
1331 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1332 102 : break;
1333 :
1334 144 : case DTK_TODAY:
1335 144 : tmask = DTK_DATE_M;
1336 144 : *dtype = DTK_DATE;
1337 144 : GetCurrentDateTime(&cur_tm);
1338 144 : tm->tm_year = cur_tm.tm_year;
1339 144 : tm->tm_mon = cur_tm.tm_mon;
1340 144 : tm->tm_mday = cur_tm.tm_mday;
1341 144 : break;
1342 :
1343 150 : case DTK_TOMORROW:
1344 150 : tmask = DTK_DATE_M;
1345 150 : *dtype = DTK_DATE;
1346 150 : GetCurrentDateTime(&cur_tm);
1347 150 : j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) + 1,
1348 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1349 150 : break;
1350 :
1351 6 : case DTK_ZULU:
1352 6 : tmask = (DTK_TIME_M | DTK_M(TZ));
1353 6 : *dtype = DTK_DATE;
1354 6 : tm->tm_hour = 0;
1355 6 : tm->tm_min = 0;
1356 6 : tm->tm_sec = 0;
1357 6 : if (tzp != NULL)
1358 6 : *tzp = 0;
1359 6 : break;
1360 :
1361 1150 : case DTK_EPOCH:
1362 : case DTK_LATE:
1363 : case DTK_EARLY:
1364 1150 : tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1365 1150 : *dtype = val;
1366 : /* caller ignores tm for these dtype codes */
1367 1150 : break;
1368 :
1369 0 : default:
1370 0 : elog(ERROR, "unrecognized RESERV datetime token: %d",
1371 : val);
1372 : }
1373 :
1374 1666 : break;
1375 :
1376 2644 : case MONTH:
1377 :
1378 : /*
1379 : * already have a (numeric) month? then see if we can
1380 : * substitute...
1381 : */
1382 2644 : if ((fmask & DTK_M(MONTH)) && !haveTextMonth &&
1383 92 : !(fmask & DTK_M(DAY)) && tm->tm_mon >= 1 &&
1384 80 : tm->tm_mon <= 31)
1385 : {
1386 74 : tm->tm_mday = tm->tm_mon;
1387 74 : tmask = DTK_M(DAY);
1388 : }
1389 2644 : haveTextMonth = true;
1390 2644 : tm->tm_mon = val;
1391 2644 : break;
1392 :
1393 6 : case DTZMOD:
1394 :
1395 : /*
1396 : * daylight savings time modifier (solves "MET DST"
1397 : * syntax)
1398 : */
1399 6 : tmask |= DTK_M(DTZ);
1400 6 : tm->tm_isdst = 1;
1401 6 : if (tzp == NULL)
1402 0 : return DTERR_BAD_FORMAT;
1403 6 : *tzp -= val;
1404 6 : break;
1405 :
1406 44 : case DTZ:
1407 :
1408 : /*
1409 : * set mask for TZ here _or_ check for DTZ later when
1410 : * getting default timezone
1411 : */
1412 44 : tmask |= DTK_M(TZ);
1413 44 : tm->tm_isdst = 1;
1414 44 : if (tzp == NULL)
1415 0 : return DTERR_BAD_FORMAT;
1416 44 : *tzp = -val;
1417 44 : break;
1418 :
1419 2282 : case TZ:
1420 2282 : tm->tm_isdst = 0;
1421 2282 : if (tzp == NULL)
1422 0 : return DTERR_BAD_FORMAT;
1423 2282 : *tzp = -val;
1424 2282 : break;
1425 :
1426 84 : case DYNTZ:
1427 84 : tmask |= DTK_M(TZ);
1428 84 : if (tzp == NULL)
1429 0 : return DTERR_BAD_FORMAT;
1430 : /* we'll determine the actual offset later */
1431 84 : abbrevTz = valtz;
1432 84 : abbrev = field[i];
1433 84 : break;
1434 :
1435 30 : case AMPM:
1436 30 : mer = val;
1437 30 : break;
1438 :
1439 306 : case ADBC:
1440 306 : bc = (val == BC);
1441 306 : break;
1442 :
1443 1952 : case DOW:
1444 1952 : tm->tm_wday = val;
1445 1952 : break;
1446 :
1447 126 : case UNITS:
1448 126 : tmask = 0;
1449 : /* reject consecutive unhandled units */
1450 126 : if (ptype != 0)
1451 12 : return DTERR_BAD_FORMAT;
1452 114 : ptype = val;
1453 114 : break;
1454 :
1455 60 : case ISOTIME:
1456 :
1457 : /*
1458 : * This is a filler field "t" indicating that the next
1459 : * field is time. Try to verify that this is sensible.
1460 : */
1461 60 : tmask = 0;
1462 :
1463 : /* No preceding date? Then quit... */
1464 60 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1465 0 : return DTERR_BAD_FORMAT;
1466 :
1467 : /* reject consecutive unhandled units */
1468 60 : if (ptype != 0)
1469 0 : return DTERR_BAD_FORMAT;
1470 60 : ptype = val;
1471 60 : break;
1472 :
1473 48 : case UNKNOWN_FIELD:
1474 :
1475 : /*
1476 : * Before giving up and declaring error, check to see
1477 : * if it is an all-alpha timezone name.
1478 : */
1479 48 : namedTz = pg_tzset(field[i]);
1480 48 : if (!namedTz)
1481 48 : return DTERR_BAD_FORMAT;
1482 : /* we'll apply the zone setting below */
1483 0 : tmask = DTK_M(TZ);
1484 0 : break;
1485 :
1486 0 : default:
1487 0 : return DTERR_BAD_FORMAT;
1488 : }
1489 9188 : break;
1490 :
1491 0 : default:
1492 0 : return DTERR_BAD_FORMAT;
1493 : }
1494 :
1495 173114 : if (tmask & fmask)
1496 144 : return DTERR_BAD_FORMAT;
1497 172970 : fmask |= tmask;
1498 : } /* end loop over fields */
1499 :
1500 : /* reject if prefix type appeared and was never handled */
1501 69464 : if (ptype != 0)
1502 0 : return DTERR_BAD_FORMAT;
1503 :
1504 : /* do additional checking for normal date specs (but not "infinity" etc) */
1505 69464 : if (*dtype == DTK_DATE)
1506 : {
1507 : /* do final checking/adjustment of Y/M/D fields */
1508 68458 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
1509 68458 : if (dterr)
1510 198 : return dterr;
1511 :
1512 : /* handle AM/PM */
1513 68260 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
1514 0 : return DTERR_FIELD_OVERFLOW;
1515 68260 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
1516 0 : tm->tm_hour = 0;
1517 68260 : else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
1518 30 : tm->tm_hour += HOURS_PER_DAY / 2;
1519 :
1520 : /* check for incomplete input */
1521 68260 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1522 : {
1523 6 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1524 0 : return 1;
1525 6 : return DTERR_BAD_FORMAT;
1526 : }
1527 :
1528 : /*
1529 : * If we had a full timezone spec, compute the offset (we could not do
1530 : * it before, because we need the date to resolve DST status).
1531 : */
1532 68254 : if (namedTz != NULL)
1533 : {
1534 : /* daylight savings time modifier disallowed with full TZ */
1535 1362 : if (fmask & DTK_M(DTZMOD))
1536 0 : return DTERR_BAD_FORMAT;
1537 :
1538 1362 : *tzp = DetermineTimeZoneOffset(tm, namedTz);
1539 : }
1540 :
1541 : /*
1542 : * Likewise, if we had a dynamic timezone abbreviation, resolve it
1543 : * now.
1544 : */
1545 68254 : if (abbrevTz != NULL)
1546 : {
1547 : /* daylight savings time modifier disallowed with dynamic TZ */
1548 84 : if (fmask & DTK_M(DTZMOD))
1549 0 : return DTERR_BAD_FORMAT;
1550 :
1551 84 : *tzp = DetermineTimeZoneAbbrevOffset(tm, abbrev, abbrevTz);
1552 : }
1553 :
1554 : /* timezone not specified? then use session timezone */
1555 68254 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
1556 : {
1557 : /*
1558 : * daylight savings time modifier but no standard timezone? then
1559 : * error
1560 : */
1561 27692 : if (fmask & DTK_M(DTZMOD))
1562 0 : return DTERR_BAD_FORMAT;
1563 :
1564 27692 : *tzp = DetermineTimeZoneOffset(tm, session_timezone);
1565 : }
1566 : }
1567 :
1568 69260 : return 0;
1569 : }
1570 :
1571 :
1572 : /* DetermineTimeZoneOffset()
1573 : *
1574 : * Given a struct pg_tm in which tm_year, tm_mon, tm_mday, tm_hour, tm_min,
1575 : * and tm_sec fields are set, and a zic-style time zone definition, determine
1576 : * the applicable GMT offset and daylight-savings status at that time.
1577 : * Set the struct pg_tm's tm_isdst field accordingly, and return the GMT
1578 : * offset as the function result.
1579 : *
1580 : * Note: if the date is out of the range we can deal with, we return zero
1581 : * as the GMT offset and set tm_isdst = 0. We don't throw an error here,
1582 : * though probably some higher-level code will.
1583 : */
1584 : int
1585 57064 : DetermineTimeZoneOffset(struct pg_tm *tm, pg_tz *tzp)
1586 : {
1587 : pg_time_t t;
1588 :
1589 57064 : return DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1590 : }
1591 :
1592 :
1593 : /* DetermineTimeZoneOffsetInternal()
1594 : *
1595 : * As above, but also return the actual UTC time imputed to the date/time
1596 : * into *tp.
1597 : *
1598 : * In event of an out-of-range date, we punt by returning zero into *tp.
1599 : * This is okay for the immediate callers but is a good reason for not
1600 : * exposing this worker function globally.
1601 : *
1602 : * Note: it might seem that we should use mktime() for this, but bitter
1603 : * experience teaches otherwise. This code is much faster than most versions
1604 : * of mktime(), anyway.
1605 : */
1606 : static int
1607 57250 : DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp, pg_time_t *tp)
1608 : {
1609 : int date,
1610 : sec;
1611 : pg_time_t day,
1612 : mytime,
1613 : prevtime,
1614 : boundary,
1615 : beforetime,
1616 : aftertime;
1617 : long int before_gmtoff,
1618 : after_gmtoff;
1619 : int before_isdst,
1620 : after_isdst;
1621 : int res;
1622 :
1623 : /*
1624 : * First, generate the pg_time_t value corresponding to the given
1625 : * y/m/d/h/m/s taken as GMT time. If this overflows, punt and decide the
1626 : * timezone is GMT. (For a valid Julian date, integer overflow should be
1627 : * impossible with 64-bit pg_time_t, but let's check for safety.)
1628 : */
1629 57250 : if (!IS_VALID_JULIAN(tm->tm_year, tm->tm_mon, tm->tm_mday))
1630 24 : goto overflow;
1631 57226 : date = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday) - UNIX_EPOCH_JDATE;
1632 :
1633 57226 : day = ((pg_time_t) date) * SECS_PER_DAY;
1634 57226 : if (day / SECS_PER_DAY != date)
1635 0 : goto overflow;
1636 57226 : sec = tm->tm_sec + (tm->tm_min + tm->tm_hour * MINS_PER_HOUR) * SECS_PER_MINUTE;
1637 57226 : mytime = day + sec;
1638 : /* since sec >= 0, overflow could only be from +day to -mytime */
1639 57226 : if (mytime < 0 && day > 0)
1640 0 : goto overflow;
1641 :
1642 : /*
1643 : * Find the DST time boundary just before or following the target time. We
1644 : * assume that all zones have GMT offsets less than 24 hours, and that DST
1645 : * boundaries can't be closer together than 48 hours, so backing up 24
1646 : * hours and finding the "next" boundary will work.
1647 : */
1648 57226 : prevtime = mytime - SECS_PER_DAY;
1649 57226 : if (mytime < 0 && prevtime > 0)
1650 0 : goto overflow;
1651 :
1652 57226 : res = pg_next_dst_boundary(&prevtime,
1653 : &before_gmtoff, &before_isdst,
1654 : &boundary,
1655 : &after_gmtoff, &after_isdst,
1656 : tzp);
1657 57226 : if (res < 0)
1658 0 : goto overflow; /* failure? */
1659 :
1660 57226 : if (res == 0)
1661 : {
1662 : /* Non-DST zone, life is simple */
1663 3592 : tm->tm_isdst = before_isdst;
1664 3592 : *tp = mytime - before_gmtoff;
1665 3592 : return -(int) before_gmtoff;
1666 : }
1667 :
1668 : /*
1669 : * Form the candidate pg_time_t values with local-time adjustment
1670 : */
1671 53634 : beforetime = mytime - before_gmtoff;
1672 53634 : if ((before_gmtoff > 0 &&
1673 12 : mytime < 0 && beforetime > 0) ||
1674 53634 : (before_gmtoff <= 0 &&
1675 40822 : mytime > 0 && beforetime < 0))
1676 0 : goto overflow;
1677 53634 : aftertime = mytime - after_gmtoff;
1678 53634 : if ((after_gmtoff > 0 &&
1679 12 : mytime < 0 && aftertime > 0) ||
1680 53634 : (after_gmtoff <= 0 &&
1681 40822 : mytime > 0 && aftertime < 0))
1682 0 : goto overflow;
1683 :
1684 : /*
1685 : * If both before or both after the boundary time, we know what to do. The
1686 : * boundary time itself is considered to be after the transition, which
1687 : * means we can accept aftertime == boundary in the second case.
1688 : */
1689 53634 : if (beforetime < boundary && aftertime < boundary)
1690 : {
1691 52756 : tm->tm_isdst = before_isdst;
1692 52756 : *tp = beforetime;
1693 52756 : return -(int) before_gmtoff;
1694 : }
1695 878 : if (beforetime > boundary && aftertime >= boundary)
1696 : {
1697 728 : tm->tm_isdst = after_isdst;
1698 728 : *tp = aftertime;
1699 728 : return -(int) after_gmtoff;
1700 : }
1701 :
1702 : /*
1703 : * It's an invalid or ambiguous time due to timezone transition. In a
1704 : * spring-forward transition, prefer the "before" interpretation; in a
1705 : * fall-back transition, prefer "after". (We used to define and implement
1706 : * this test as "prefer the standard-time interpretation", but that rule
1707 : * does not help to resolve the behavior when both times are reported as
1708 : * standard time; which does happen, eg Europe/Moscow in Oct 2014. Also,
1709 : * in some zones such as Europe/Dublin, there is widespread confusion
1710 : * about which time offset is "standard" time, so it's fortunate that our
1711 : * behavior doesn't depend on that.)
1712 : */
1713 150 : if (beforetime > aftertime)
1714 : {
1715 72 : tm->tm_isdst = before_isdst;
1716 72 : *tp = beforetime;
1717 72 : return -(int) before_gmtoff;
1718 : }
1719 78 : tm->tm_isdst = after_isdst;
1720 78 : *tp = aftertime;
1721 78 : return -(int) after_gmtoff;
1722 :
1723 24 : overflow:
1724 : /* Given date is out of range, so assume UTC */
1725 24 : tm->tm_isdst = 0;
1726 24 : *tp = 0;
1727 24 : return 0;
1728 : }
1729 :
1730 :
1731 : /* DetermineTimeZoneAbbrevOffset()
1732 : *
1733 : * Determine the GMT offset and DST flag to be attributed to a dynamic
1734 : * time zone abbreviation, that is one whose meaning has changed over time.
1735 : * *tm contains the local time at which the meaning should be determined,
1736 : * and tm->tm_isdst receives the DST flag.
1737 : *
1738 : * This differs from the behavior of DetermineTimeZoneOffset() in that a
1739 : * standard-time or daylight-time abbreviation forces use of the corresponding
1740 : * GMT offset even when the zone was then in DS or standard time respectively.
1741 : * (However, that happens only if we can match the given abbreviation to some
1742 : * abbreviation that appears in the IANA timezone data. Otherwise, we fall
1743 : * back to doing DetermineTimeZoneOffset().)
1744 : */
1745 : int
1746 186 : DetermineTimeZoneAbbrevOffset(struct pg_tm *tm, const char *abbr, pg_tz *tzp)
1747 : {
1748 : pg_time_t t;
1749 : int zone_offset;
1750 : int abbr_offset;
1751 : int abbr_isdst;
1752 :
1753 : /*
1754 : * Compute the UTC time we want to probe at. (In event of overflow, we'll
1755 : * probe at the epoch, which is a bit random but probably doesn't matter.)
1756 : */
1757 186 : zone_offset = DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1758 :
1759 : /*
1760 : * Try to match the abbreviation to something in the zone definition.
1761 : */
1762 186 : if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1763 : &abbr_offset, &abbr_isdst))
1764 : {
1765 : /* Success, so use the abbrev-specific answers. */
1766 186 : tm->tm_isdst = abbr_isdst;
1767 186 : return abbr_offset;
1768 : }
1769 :
1770 : /*
1771 : * No match, so use the answers we already got from
1772 : * DetermineTimeZoneOffsetInternal.
1773 : */
1774 0 : return zone_offset;
1775 : }
1776 :
1777 :
1778 : /* DetermineTimeZoneAbbrevOffsetTS()
1779 : *
1780 : * As above but the probe time is specified as a TimestampTz (hence, UTC time),
1781 : * and DST status is returned into *isdst rather than into tm->tm_isdst.
1782 : */
1783 : int
1784 966 : DetermineTimeZoneAbbrevOffsetTS(TimestampTz ts, const char *abbr,
1785 : pg_tz *tzp, int *isdst)
1786 : {
1787 966 : pg_time_t t = timestamptz_to_time_t(ts);
1788 : int zone_offset;
1789 : int abbr_offset;
1790 : int tz;
1791 : struct pg_tm tm;
1792 : fsec_t fsec;
1793 :
1794 : /*
1795 : * If the abbrev matches anything in the zone data, this is pretty easy.
1796 : */
1797 966 : if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1798 : &abbr_offset, isdst))
1799 90 : return abbr_offset;
1800 :
1801 : /*
1802 : * Else, break down the timestamp so we can use DetermineTimeZoneOffset.
1803 : */
1804 876 : if (timestamp2tm(ts, &tz, &tm, &fsec, NULL, tzp) != 0)
1805 0 : ereport(ERROR,
1806 : (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
1807 : errmsg("timestamp out of range")));
1808 :
1809 876 : zone_offset = DetermineTimeZoneOffset(&tm, tzp);
1810 876 : *isdst = tm.tm_isdst;
1811 876 : return zone_offset;
1812 : }
1813 :
1814 :
1815 : /* DetermineTimeZoneAbbrevOffsetInternal()
1816 : *
1817 : * Workhorse for above two functions: work from a pg_time_t probe instant.
1818 : * On success, return GMT offset and DST status into *offset and *isdst.
1819 : */
1820 : static bool
1821 1152 : DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t, const char *abbr, pg_tz *tzp,
1822 : int *offset, int *isdst)
1823 : {
1824 : char upabbr[TZ_STRLEN_MAX + 1];
1825 : unsigned char *p;
1826 : long int gmtoff;
1827 :
1828 : /* We need to force the abbrev to upper case */
1829 1152 : strlcpy(upabbr, abbr, sizeof(upabbr));
1830 5376 : for (p = (unsigned char *) upabbr; *p; p++)
1831 4224 : *p = pg_toupper(*p);
1832 :
1833 : /* Look up the abbrev's meaning at this time in this zone */
1834 1152 : if (pg_interpret_timezone_abbrev(upabbr,
1835 : &t,
1836 : &gmtoff,
1837 : isdst,
1838 : tzp))
1839 : {
1840 : /* Change sign to agree with DetermineTimeZoneOffset() */
1841 276 : *offset = (int) -gmtoff;
1842 276 : return true;
1843 : }
1844 876 : return false;
1845 : }
1846 :
1847 :
1848 : /* DecodeTimeOnly()
1849 : * Interpret parsed string as time fields only.
1850 : * Returns 0 if successful, DTERR code if bogus input detected.
1851 : *
1852 : * Inputs are field[] and ftype[] arrays, of length nf.
1853 : * Other arguments are outputs.
1854 : *
1855 : * Note that support for time zone is here for
1856 : * SQL TIME WITH TIME ZONE, but it reveals
1857 : * bogosity with SQL date/time standards, since
1858 : * we must infer a time zone from current time.
1859 : * - thomas 2000-03-10
1860 : * Allow specifying date to get a better time zone,
1861 : * if time zones are allowed. - thomas 2001-12-26
1862 : */
1863 : int
1864 4270 : DecodeTimeOnly(char **field, int *ftype, int nf,
1865 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
1866 : DateTimeErrorExtra *extra)
1867 : {
1868 4270 : int fmask = 0,
1869 : tmask,
1870 : type;
1871 4270 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
1872 : int i;
1873 : int val;
1874 : int dterr;
1875 4270 : bool isjulian = false;
1876 4270 : bool is2digits = false;
1877 4270 : bool bc = false;
1878 4270 : int mer = HR24;
1879 4270 : pg_tz *namedTz = NULL;
1880 4270 : pg_tz *abbrevTz = NULL;
1881 4270 : char *abbrev = NULL;
1882 : pg_tz *valtz;
1883 :
1884 4270 : *dtype = DTK_TIME;
1885 4270 : tm->tm_hour = 0;
1886 4270 : tm->tm_min = 0;
1887 4270 : tm->tm_sec = 0;
1888 4270 : *fsec = 0;
1889 : /* don't know daylight savings time status apriori */
1890 4270 : tm->tm_isdst = -1;
1891 :
1892 4270 : if (tzp != NULL)
1893 4270 : *tzp = 0;
1894 :
1895 10998 : for (i = 0; i < nf; i++)
1896 : {
1897 6752 : switch (ftype[i])
1898 : {
1899 1402 : case DTK_DATE:
1900 :
1901 : /*
1902 : * Time zone not allowed? Then should not accept dates or time
1903 : * zones no matter what else!
1904 : */
1905 1402 : if (tzp == NULL)
1906 0 : return DTERR_BAD_FORMAT;
1907 :
1908 : /* Under limited circumstances, we will accept a date... */
1909 1402 : if (i == 0 && nf >= 2 &&
1910 210 : (ftype[nf - 1] == DTK_DATE || ftype[1] == DTK_TIME))
1911 : {
1912 210 : dterr = DecodeDate(field[i], fmask,
1913 : &tmask, &is2digits, tm);
1914 210 : if (dterr)
1915 0 : return dterr;
1916 : }
1917 : /* otherwise, this is a time and/or time zone */
1918 : else
1919 : {
1920 1192 : if (isdigit((unsigned char) *field[i]))
1921 : {
1922 : char *cp;
1923 :
1924 : /*
1925 : * Starts with a digit but we already have a time
1926 : * field? Then we are in trouble with time already...
1927 : */
1928 0 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1929 0 : return DTERR_BAD_FORMAT;
1930 :
1931 : /*
1932 : * Should not get here and fail. Sanity check only...
1933 : */
1934 0 : if ((cp = strchr(field[i], '-')) == NULL)
1935 0 : return DTERR_BAD_FORMAT;
1936 :
1937 : /* Get the time zone from the end of the string */
1938 0 : dterr = DecodeTimezone(cp, tzp);
1939 0 : if (dterr)
1940 0 : return dterr;
1941 0 : *cp = '\0';
1942 :
1943 : /*
1944 : * Then read the rest of the field as a concatenated
1945 : * time
1946 : */
1947 0 : dterr = DecodeNumberField(strlen(field[i]), field[i],
1948 : (fmask | DTK_DATE_M),
1949 : &tmask, tm,
1950 : fsec, &is2digits);
1951 0 : if (dterr < 0)
1952 0 : return dterr;
1953 0 : ftype[i] = dterr;
1954 :
1955 0 : tmask |= DTK_M(TZ);
1956 : }
1957 : else
1958 : {
1959 1192 : namedTz = pg_tzset(field[i]);
1960 1192 : if (!namedTz)
1961 : {
1962 0 : extra->dtee_timezone = field[i];
1963 0 : return DTERR_BAD_TIMEZONE;
1964 : }
1965 : /* we'll apply the zone setting below */
1966 1192 : ftype[i] = DTK_TZ;
1967 1192 : tmask = DTK_M(TZ);
1968 : }
1969 : }
1970 1402 : break;
1971 :
1972 4096 : case DTK_TIME:
1973 :
1974 : /*
1975 : * This might be an ISO time following a "t" field.
1976 : */
1977 4096 : if (ptype != 0)
1978 : {
1979 36 : if (ptype != DTK_TIME)
1980 0 : return DTERR_BAD_FORMAT;
1981 36 : ptype = 0;
1982 : }
1983 :
1984 4096 : dterr = DecodeTime(field[i], (fmask | DTK_DATE_M),
1985 : INTERVAL_FULL_RANGE,
1986 : &tmask, tm, fsec);
1987 4096 : if (dterr)
1988 0 : return dterr;
1989 4096 : break;
1990 :
1991 630 : case DTK_TZ:
1992 : {
1993 : int tz;
1994 :
1995 630 : if (tzp == NULL)
1996 0 : return DTERR_BAD_FORMAT;
1997 :
1998 630 : dterr = DecodeTimezone(field[i], &tz);
1999 630 : if (dterr)
2000 0 : return dterr;
2001 630 : *tzp = tz;
2002 630 : tmask = DTK_M(TZ);
2003 : }
2004 630 : break;
2005 :
2006 180 : case DTK_NUMBER:
2007 :
2008 : /*
2009 : * Deal with cases where previous field labeled this one
2010 : */
2011 180 : if (ptype != 0)
2012 : {
2013 : char *cp;
2014 : int value;
2015 :
2016 120 : errno = 0;
2017 120 : value = strtoint(field[i], &cp, 10);
2018 120 : if (errno == ERANGE)
2019 24 : return DTERR_FIELD_OVERFLOW;
2020 120 : if (*cp != '.' && *cp != '\0')
2021 0 : return DTERR_BAD_FORMAT;
2022 :
2023 : switch (ptype)
2024 : {
2025 6 : case DTK_JULIAN:
2026 : /* previous field was a label for "julian date" */
2027 6 : if (tzp == NULL)
2028 0 : return DTERR_BAD_FORMAT;
2029 6 : if (value < 0)
2030 0 : return DTERR_FIELD_OVERFLOW;
2031 6 : tmask = DTK_DATE_M;
2032 6 : j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2033 6 : isjulian = true;
2034 :
2035 6 : if (*cp == '.')
2036 : {
2037 : double time;
2038 :
2039 0 : dterr = ParseFraction(cp, &time);
2040 0 : if (dterr)
2041 0 : return dterr;
2042 0 : time *= USECS_PER_DAY;
2043 0 : dt2time(time,
2044 : &tm->tm_hour, &tm->tm_min,
2045 : &tm->tm_sec, fsec);
2046 0 : tmask |= DTK_TIME_M;
2047 : }
2048 6 : break;
2049 :
2050 102 : case DTK_TIME:
2051 : /* previous field was "t" for ISO time */
2052 102 : dterr = DecodeNumberField(strlen(field[i]), field[i],
2053 : (fmask | DTK_DATE_M),
2054 : &tmask, tm,
2055 : fsec, &is2digits);
2056 102 : if (dterr < 0)
2057 12 : return dterr;
2058 90 : ftype[i] = dterr;
2059 :
2060 90 : if (tmask != DTK_TIME_M)
2061 0 : return DTERR_BAD_FORMAT;
2062 90 : break;
2063 :
2064 12 : default:
2065 12 : return DTERR_BAD_FORMAT;
2066 : break;
2067 : }
2068 :
2069 96 : ptype = 0;
2070 96 : *dtype = DTK_DATE;
2071 : }
2072 : else
2073 : {
2074 : char *cp;
2075 : int flen;
2076 :
2077 60 : flen = strlen(field[i]);
2078 60 : cp = strchr(field[i], '.');
2079 :
2080 : /* Embedded decimal? */
2081 60 : if (cp != NULL)
2082 : {
2083 : /*
2084 : * Under limited circumstances, we will accept a
2085 : * date...
2086 : */
2087 36 : if (i == 0 && nf >= 2 && ftype[nf - 1] == DTK_DATE)
2088 : {
2089 0 : dterr = DecodeDate(field[i], fmask,
2090 : &tmask, &is2digits, tm);
2091 0 : if (dterr)
2092 0 : return dterr;
2093 : }
2094 : /* embedded decimal and several digits before? */
2095 36 : else if (flen - strlen(cp) > 2)
2096 : {
2097 : /*
2098 : * Interpret as a concatenated date or time Set
2099 : * the type field to allow decoding other fields
2100 : * later. Example: 20011223 or 040506
2101 : */
2102 36 : dterr = DecodeNumberField(flen, field[i],
2103 : (fmask | DTK_DATE_M),
2104 : &tmask, tm,
2105 : fsec, &is2digits);
2106 36 : if (dterr < 0)
2107 0 : return dterr;
2108 36 : ftype[i] = dterr;
2109 : }
2110 : else
2111 0 : return DTERR_BAD_FORMAT;
2112 : }
2113 24 : else if (flen > 4)
2114 : {
2115 12 : dterr = DecodeNumberField(flen, field[i],
2116 : (fmask | DTK_DATE_M),
2117 : &tmask, tm,
2118 : fsec, &is2digits);
2119 12 : if (dterr < 0)
2120 0 : return dterr;
2121 12 : ftype[i] = dterr;
2122 : }
2123 : /* otherwise it is a single date/time field... */
2124 : else
2125 : {
2126 12 : dterr = DecodeNumber(flen, field[i],
2127 : false,
2128 : (fmask | DTK_DATE_M),
2129 : &tmask, tm,
2130 : fsec, &is2digits);
2131 12 : if (dterr)
2132 0 : return dterr;
2133 : }
2134 : }
2135 156 : break;
2136 :
2137 444 : case DTK_STRING:
2138 : case DTK_SPECIAL:
2139 : /* timezone abbrevs take precedence over built-in tokens */
2140 444 : dterr = DecodeTimezoneAbbrev(i, field[i],
2141 : &type, &val, &valtz, extra);
2142 444 : if (dterr)
2143 0 : return dterr;
2144 444 : if (type == UNKNOWN_FIELD)
2145 180 : type = DecodeSpecial(i, field[i], &val);
2146 444 : if (type == IGNORE_DTF)
2147 0 : continue;
2148 :
2149 444 : tmask = DTK_M(type);
2150 444 : switch (type)
2151 : {
2152 12 : case RESERV:
2153 12 : switch (val)
2154 : {
2155 12 : case DTK_NOW:
2156 12 : tmask = DTK_TIME_M;
2157 12 : *dtype = DTK_TIME;
2158 12 : GetCurrentTimeUsec(tm, fsec, NULL);
2159 12 : break;
2160 :
2161 0 : case DTK_ZULU:
2162 0 : tmask = (DTK_TIME_M | DTK_M(TZ));
2163 0 : *dtype = DTK_TIME;
2164 0 : tm->tm_hour = 0;
2165 0 : tm->tm_min = 0;
2166 0 : tm->tm_sec = 0;
2167 0 : tm->tm_isdst = 0;
2168 0 : break;
2169 :
2170 0 : default:
2171 0 : return DTERR_BAD_FORMAT;
2172 : }
2173 :
2174 12 : break;
2175 :
2176 0 : case DTZMOD:
2177 :
2178 : /*
2179 : * daylight savings time modifier (solves "MET DST"
2180 : * syntax)
2181 : */
2182 0 : tmask |= DTK_M(DTZ);
2183 0 : tm->tm_isdst = 1;
2184 0 : if (tzp == NULL)
2185 0 : return DTERR_BAD_FORMAT;
2186 0 : *tzp -= val;
2187 0 : break;
2188 :
2189 198 : case DTZ:
2190 :
2191 : /*
2192 : * set mask for TZ here _or_ check for DTZ later when
2193 : * getting default timezone
2194 : */
2195 198 : tmask |= DTK_M(TZ);
2196 198 : tm->tm_isdst = 1;
2197 198 : if (tzp == NULL)
2198 0 : return DTERR_BAD_FORMAT;
2199 198 : *tzp = -val;
2200 198 : ftype[i] = DTK_TZ;
2201 198 : break;
2202 :
2203 60 : case TZ:
2204 60 : tm->tm_isdst = 0;
2205 60 : if (tzp == NULL)
2206 0 : return DTERR_BAD_FORMAT;
2207 60 : *tzp = -val;
2208 60 : ftype[i] = DTK_TZ;
2209 60 : break;
2210 :
2211 6 : case DYNTZ:
2212 6 : tmask |= DTK_M(TZ);
2213 6 : if (tzp == NULL)
2214 0 : return DTERR_BAD_FORMAT;
2215 : /* we'll determine the actual offset later */
2216 6 : abbrevTz = valtz;
2217 6 : abbrev = field[i];
2218 6 : ftype[i] = DTK_TZ;
2219 6 : break;
2220 :
2221 12 : case AMPM:
2222 12 : mer = val;
2223 12 : break;
2224 :
2225 0 : case ADBC:
2226 0 : bc = (val == BC);
2227 0 : break;
2228 :
2229 18 : case UNITS:
2230 18 : tmask = 0;
2231 : /* reject consecutive unhandled units */
2232 18 : if (ptype != 0)
2233 0 : return DTERR_BAD_FORMAT;
2234 18 : ptype = val;
2235 18 : break;
2236 :
2237 138 : case ISOTIME:
2238 138 : tmask = 0;
2239 : /* reject consecutive unhandled units */
2240 138 : if (ptype != 0)
2241 0 : return DTERR_BAD_FORMAT;
2242 138 : ptype = val;
2243 138 : break;
2244 :
2245 0 : case UNKNOWN_FIELD:
2246 :
2247 : /*
2248 : * Before giving up and declaring error, check to see
2249 : * if it is an all-alpha timezone name.
2250 : */
2251 0 : namedTz = pg_tzset(field[i]);
2252 0 : if (!namedTz)
2253 0 : return DTERR_BAD_FORMAT;
2254 : /* we'll apply the zone setting below */
2255 0 : tmask = DTK_M(TZ);
2256 0 : break;
2257 :
2258 0 : default:
2259 0 : return DTERR_BAD_FORMAT;
2260 : }
2261 444 : break;
2262 :
2263 0 : default:
2264 0 : return DTERR_BAD_FORMAT;
2265 : }
2266 :
2267 6728 : if (tmask & fmask)
2268 0 : return DTERR_BAD_FORMAT;
2269 6728 : fmask |= tmask;
2270 : } /* end loop over fields */
2271 :
2272 : /* reject if prefix type appeared and was never handled */
2273 4246 : if (ptype != 0)
2274 0 : return DTERR_BAD_FORMAT;
2275 :
2276 : /* do final checking/adjustment of Y/M/D fields */
2277 4246 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
2278 4246 : if (dterr)
2279 0 : return dterr;
2280 :
2281 : /* handle AM/PM */
2282 4246 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
2283 0 : return DTERR_FIELD_OVERFLOW;
2284 4246 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
2285 0 : tm->tm_hour = 0;
2286 4246 : else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
2287 12 : tm->tm_hour += HOURS_PER_DAY / 2;
2288 :
2289 : /* check for time overflow */
2290 4246 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec, *fsec))
2291 72 : return DTERR_FIELD_OVERFLOW;
2292 :
2293 4174 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2294 0 : return DTERR_BAD_FORMAT;
2295 :
2296 : /*
2297 : * If we had a full timezone spec, compute the offset (we could not do it
2298 : * before, because we may need the date to resolve DST status).
2299 : */
2300 4174 : if (namedTz != NULL)
2301 : {
2302 : long int gmtoff;
2303 :
2304 : /* daylight savings time modifier disallowed with full TZ */
2305 1192 : if (fmask & DTK_M(DTZMOD))
2306 42 : return DTERR_BAD_FORMAT;
2307 :
2308 : /* if non-DST zone, we do not need to know the date */
2309 1192 : if (pg_get_timezone_offset(namedTz, &gmtoff))
2310 : {
2311 1114 : *tzp = -(int) gmtoff;
2312 : }
2313 : else
2314 : {
2315 : /* a date has to be specified */
2316 78 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2317 42 : return DTERR_BAD_FORMAT;
2318 36 : *tzp = DetermineTimeZoneOffset(tm, namedTz);
2319 : }
2320 : }
2321 :
2322 : /*
2323 : * Likewise, if we had a dynamic timezone abbreviation, resolve it now.
2324 : */
2325 4132 : if (abbrevTz != NULL)
2326 : {
2327 : struct pg_tm tt,
2328 0 : *tmp = &tt;
2329 :
2330 : /*
2331 : * daylight savings time modifier but no standard timezone? then error
2332 : */
2333 0 : if (fmask & DTK_M(DTZMOD))
2334 0 : return DTERR_BAD_FORMAT;
2335 :
2336 0 : if ((fmask & DTK_DATE_M) == 0)
2337 0 : GetCurrentDateTime(tmp);
2338 : else
2339 : {
2340 : /* a date has to be specified */
2341 0 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2342 0 : return DTERR_BAD_FORMAT;
2343 0 : tmp->tm_year = tm->tm_year;
2344 0 : tmp->tm_mon = tm->tm_mon;
2345 0 : tmp->tm_mday = tm->tm_mday;
2346 : }
2347 0 : tmp->tm_hour = tm->tm_hour;
2348 0 : tmp->tm_min = tm->tm_min;
2349 0 : tmp->tm_sec = tm->tm_sec;
2350 0 : *tzp = DetermineTimeZoneAbbrevOffset(tmp, abbrev, abbrevTz);
2351 0 : tm->tm_isdst = tmp->tm_isdst;
2352 : }
2353 :
2354 : /* timezone not specified? then use session timezone */
2355 4132 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
2356 : {
2357 : struct pg_tm tt,
2358 2130 : *tmp = &tt;
2359 :
2360 : /*
2361 : * daylight savings time modifier but no standard timezone? then error
2362 : */
2363 2130 : if (fmask & DTK_M(DTZMOD))
2364 0 : return DTERR_BAD_FORMAT;
2365 :
2366 2130 : if ((fmask & DTK_DATE_M) == 0)
2367 2052 : GetCurrentDateTime(tmp);
2368 : else
2369 : {
2370 : /* a date has to be specified */
2371 78 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2372 0 : return DTERR_BAD_FORMAT;
2373 78 : tmp->tm_year = tm->tm_year;
2374 78 : tmp->tm_mon = tm->tm_mon;
2375 78 : tmp->tm_mday = tm->tm_mday;
2376 : }
2377 2130 : tmp->tm_hour = tm->tm_hour;
2378 2130 : tmp->tm_min = tm->tm_min;
2379 2130 : tmp->tm_sec = tm->tm_sec;
2380 2130 : *tzp = DetermineTimeZoneOffset(tmp, session_timezone);
2381 2130 : tm->tm_isdst = tmp->tm_isdst;
2382 : }
2383 :
2384 4132 : return 0;
2385 : }
2386 :
2387 : /* DecodeDate()
2388 : * Decode date string which includes delimiters.
2389 : * Return 0 if okay, a DTERR code if not.
2390 : *
2391 : * str: field to be parsed
2392 : * fmask: bitmask for field types already seen
2393 : * *tmask: receives bitmask for fields found here
2394 : * *is2digits: set to true if we find 2-digit year
2395 : * *tm: field values are stored into appropriate members of this struct
2396 : */
2397 : static int
2398 65204 : DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
2399 : struct pg_tm *tm)
2400 : {
2401 : fsec_t fsec;
2402 65204 : int nf = 0;
2403 : int i,
2404 : len;
2405 : int dterr;
2406 65204 : bool haveTextMonth = false;
2407 : int type,
2408 : val,
2409 65204 : dmask = 0;
2410 : char *field[MAXDATEFIELDS];
2411 :
2412 65204 : *tmask = 0;
2413 :
2414 : /* parse this string... */
2415 260750 : while (*str != '\0' && nf < MAXDATEFIELDS)
2416 : {
2417 : /* skip field separators */
2418 195546 : while (*str != '\0' && !isalnum((unsigned char) *str))
2419 0 : str++;
2420 :
2421 195546 : if (*str == '\0')
2422 0 : return DTERR_BAD_FORMAT; /* end of string after separator */
2423 :
2424 195546 : field[nf] = str;
2425 195546 : if (isdigit((unsigned char) *str))
2426 : {
2427 716526 : while (isdigit((unsigned char) *str))
2428 521124 : str++;
2429 : }
2430 144 : else if (isalpha((unsigned char) *str))
2431 : {
2432 576 : while (isalpha((unsigned char) *str))
2433 432 : str++;
2434 : }
2435 :
2436 : /* Just get rid of any non-digit, non-alpha characters... */
2437 195546 : if (*str != '\0')
2438 130378 : *str++ = '\0';
2439 195546 : nf++;
2440 : }
2441 :
2442 : /* look first for text fields, since that will be unambiguous month */
2443 260750 : for (i = 0; i < nf; i++)
2444 : {
2445 195546 : if (isalpha((unsigned char) *field[i]))
2446 : {
2447 144 : type = DecodeSpecial(i, field[i], &val);
2448 144 : if (type == IGNORE_DTF)
2449 0 : continue;
2450 :
2451 144 : dmask = DTK_M(type);
2452 144 : switch (type)
2453 : {
2454 144 : case MONTH:
2455 144 : tm->tm_mon = val;
2456 144 : haveTextMonth = true;
2457 144 : break;
2458 :
2459 0 : default:
2460 0 : return DTERR_BAD_FORMAT;
2461 : }
2462 144 : if (fmask & dmask)
2463 0 : return DTERR_BAD_FORMAT;
2464 :
2465 144 : fmask |= dmask;
2466 144 : *tmask |= dmask;
2467 :
2468 : /* mark this field as being completed */
2469 144 : field[i] = NULL;
2470 : }
2471 : }
2472 :
2473 : /* now pick up remaining numeric fields */
2474 260750 : for (i = 0; i < nf; i++)
2475 : {
2476 195546 : if (field[i] == NULL)
2477 144 : continue;
2478 :
2479 195402 : if ((len = strlen(field[i])) <= 0)
2480 0 : return DTERR_BAD_FORMAT;
2481 :
2482 195402 : dterr = DecodeNumber(len, field[i], haveTextMonth, fmask,
2483 : &dmask, tm,
2484 : &fsec, is2digits);
2485 195402 : if (dterr)
2486 0 : return dterr;
2487 :
2488 195402 : if (fmask & dmask)
2489 0 : return DTERR_BAD_FORMAT;
2490 :
2491 195402 : fmask |= dmask;
2492 195402 : *tmask |= dmask;
2493 : }
2494 :
2495 65204 : if ((fmask & ~(DTK_M(DOY) | DTK_M(TZ))) != DTK_DATE_M)
2496 36 : return DTERR_BAD_FORMAT;
2497 :
2498 : /* validation of the field values must wait until ValidateDate() */
2499 :
2500 65168 : return 0;
2501 : }
2502 :
2503 : /* ValidateDate()
2504 : * Check valid year/month/day values, handle BC and DOY cases
2505 : * Return 0 if okay, a DTERR code if not.
2506 : */
2507 : int
2508 78950 : ValidateDate(int fmask, bool isjulian, bool is2digits, bool bc,
2509 : struct pg_tm *tm)
2510 : {
2511 78950 : if (fmask & DTK_M(YEAR))
2512 : {
2513 74908 : if (isjulian)
2514 : {
2515 : /* tm_year is correct and should not be touched */
2516 : }
2517 68836 : else if (bc)
2518 : {
2519 : /* there is no year zero in AD/BC notation */
2520 306 : if (tm->tm_year <= 0)
2521 0 : return DTERR_FIELD_OVERFLOW;
2522 : /* internally, we represent 1 BC as year zero, 2 BC as -1, etc */
2523 306 : tm->tm_year = -(tm->tm_year - 1);
2524 : }
2525 68530 : else if (is2digits)
2526 : {
2527 : /* process 1 or 2-digit input as 1970-2069 AD, allow '0' and '00' */
2528 354 : if (tm->tm_year < 0) /* just paranoia */
2529 0 : return DTERR_FIELD_OVERFLOW;
2530 354 : if (tm->tm_year < 70)
2531 174 : tm->tm_year += 2000;
2532 180 : else if (tm->tm_year < 100)
2533 180 : tm->tm_year += 1900;
2534 : }
2535 : else
2536 : {
2537 : /* there is no year zero in AD/BC notation */
2538 68176 : if (tm->tm_year <= 0)
2539 12 : return DTERR_FIELD_OVERFLOW;
2540 : }
2541 : }
2542 :
2543 : /* now that we have correct year, decode DOY */
2544 78938 : if (fmask & DTK_M(DOY))
2545 : {
2546 30 : j2date(date2j(tm->tm_year, 1, 1) + tm->tm_yday - 1,
2547 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2548 : }
2549 :
2550 : /* check for valid month */
2551 78938 : if (fmask & DTK_M(MONTH))
2552 : {
2553 74872 : if (tm->tm_mon < 1 || tm->tm_mon > MONTHS_PER_YEAR)
2554 78 : return DTERR_MD_FIELD_OVERFLOW;
2555 : }
2556 :
2557 : /* minimal check for valid day */
2558 78860 : if (fmask & DTK_M(DAY))
2559 : {
2560 74758 : if (tm->tm_mday < 1 || tm->tm_mday > 31)
2561 138 : return DTERR_MD_FIELD_OVERFLOW;
2562 : }
2563 :
2564 78722 : if ((fmask & DTK_DATE_M) == DTK_DATE_M)
2565 : {
2566 : /*
2567 : * Check for valid day of month, now that we know for sure the month
2568 : * and year. Note we don't use MD_FIELD_OVERFLOW here, since it seems
2569 : * unlikely that "Feb 29" is a YMD-order error.
2570 : */
2571 74602 : if (tm->tm_mday > day_tab[isleap(tm->tm_year)][tm->tm_mon - 1])
2572 48 : return DTERR_FIELD_OVERFLOW;
2573 : }
2574 :
2575 78674 : return 0;
2576 : }
2577 :
2578 :
2579 : /* DecodeTimeCommon()
2580 : * Decode time string which includes delimiters.
2581 : * Return 0 if okay, a DTERR code if not.
2582 : * tmask and itm are output parameters.
2583 : *
2584 : * This code is shared between the timestamp and interval cases.
2585 : * We return a struct pg_itm (of which only the tm_usec, tm_sec, tm_min,
2586 : * and tm_hour fields are used) and let the wrapper functions below
2587 : * convert and range-check as necessary.
2588 : */
2589 : static int
2590 60784 : DecodeTimeCommon(char *str, int fmask, int range,
2591 : int *tmask, struct pg_itm *itm)
2592 : {
2593 : char *cp;
2594 : int dterr;
2595 60784 : fsec_t fsec = 0;
2596 :
2597 60784 : *tmask = DTK_TIME_M;
2598 :
2599 60784 : errno = 0;
2600 60784 : itm->tm_hour = strtoi64(str, &cp, 10);
2601 60784 : if (errno == ERANGE)
2602 0 : return DTERR_FIELD_OVERFLOW;
2603 60784 : if (*cp != ':')
2604 0 : return DTERR_BAD_FORMAT;
2605 60784 : errno = 0;
2606 60784 : itm->tm_min = strtoint(cp + 1, &cp, 10);
2607 60784 : if (errno == ERANGE)
2608 0 : return DTERR_FIELD_OVERFLOW;
2609 60784 : if (*cp == '\0')
2610 : {
2611 1698 : itm->tm_sec = 0;
2612 : /* If it's a MINUTE TO SECOND interval, take 2 fields as being mm:ss */
2613 1698 : if (range == (INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND)))
2614 : {
2615 18 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2616 0 : return DTERR_FIELD_OVERFLOW;
2617 18 : itm->tm_sec = itm->tm_min;
2618 18 : itm->tm_min = (int) itm->tm_hour;
2619 18 : itm->tm_hour = 0;
2620 : }
2621 : }
2622 59086 : else if (*cp == '.')
2623 : {
2624 : /* always assume mm:ss.sss is MINUTE TO SECOND */
2625 48 : dterr = ParseFractionalSecond(cp, &fsec);
2626 48 : if (dterr)
2627 12 : return dterr;
2628 36 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2629 0 : return DTERR_FIELD_OVERFLOW;
2630 36 : itm->tm_sec = itm->tm_min;
2631 36 : itm->tm_min = (int) itm->tm_hour;
2632 36 : itm->tm_hour = 0;
2633 : }
2634 59038 : else if (*cp == ':')
2635 : {
2636 59038 : errno = 0;
2637 59038 : itm->tm_sec = strtoint(cp + 1, &cp, 10);
2638 59038 : if (errno == ERANGE)
2639 0 : return DTERR_FIELD_OVERFLOW;
2640 59038 : if (*cp == '.')
2641 : {
2642 22656 : dterr = ParseFractionalSecond(cp, &fsec);
2643 22656 : if (dterr)
2644 0 : return dterr;
2645 : }
2646 36382 : else if (*cp != '\0')
2647 0 : return DTERR_BAD_FORMAT;
2648 : }
2649 : else
2650 0 : return DTERR_BAD_FORMAT;
2651 :
2652 : /* do a sanity check; but caller must check the range of tm_hour */
2653 60772 : if (itm->tm_hour < 0 ||
2654 60772 : itm->tm_min < 0 || itm->tm_min > MINS_PER_HOUR - 1 ||
2655 60772 : itm->tm_sec < 0 || itm->tm_sec > SECS_PER_MINUTE ||
2656 60772 : fsec < 0 || fsec > USECS_PER_SEC)
2657 0 : return DTERR_FIELD_OVERFLOW;
2658 :
2659 60772 : itm->tm_usec = (int) fsec;
2660 :
2661 60772 : return 0;
2662 : }
2663 :
2664 : /* DecodeTime()
2665 : * Decode time string which includes delimiters.
2666 : * Return 0 if okay, a DTERR code if not.
2667 : *
2668 : * This version is used for timestamps. The results are returned into
2669 : * the tm_hour/tm_min/tm_sec fields of *tm, and microseconds into *fsec.
2670 : */
2671 : static int
2672 58782 : DecodeTime(char *str, int fmask, int range,
2673 : int *tmask, struct pg_tm *tm, fsec_t *fsec)
2674 : {
2675 : struct pg_itm itm;
2676 : int dterr;
2677 :
2678 58782 : dterr = DecodeTimeCommon(str, fmask, range,
2679 : tmask, &itm);
2680 58782 : if (dterr)
2681 0 : return dterr;
2682 :
2683 58782 : if (itm.tm_hour > INT_MAX)
2684 0 : return DTERR_FIELD_OVERFLOW;
2685 58782 : tm->tm_hour = (int) itm.tm_hour;
2686 58782 : tm->tm_min = itm.tm_min;
2687 58782 : tm->tm_sec = itm.tm_sec;
2688 58782 : *fsec = itm.tm_usec;
2689 :
2690 58782 : return 0;
2691 : }
2692 :
2693 : /* DecodeTimeForInterval()
2694 : * Decode time string which includes delimiters.
2695 : * Return 0 if okay, a DTERR code if not.
2696 : *
2697 : * This version is used for intervals. The results are returned into
2698 : * itm_in->tm_usec.
2699 : */
2700 : static int
2701 2002 : DecodeTimeForInterval(char *str, int fmask, int range,
2702 : int *tmask, struct pg_itm_in *itm_in)
2703 : {
2704 : struct pg_itm itm;
2705 : int dterr;
2706 :
2707 2002 : dterr = DecodeTimeCommon(str, fmask, range,
2708 : tmask, &itm);
2709 2002 : if (dterr)
2710 12 : return dterr;
2711 :
2712 1990 : itm_in->tm_usec = itm.tm_usec;
2713 1990 : if (!int64_multiply_add(itm.tm_hour, USECS_PER_HOUR, &itm_in->tm_usec) ||
2714 1990 : !int64_multiply_add(itm.tm_min, USECS_PER_MINUTE, &itm_in->tm_usec) ||
2715 1990 : !int64_multiply_add(itm.tm_sec, USECS_PER_SEC, &itm_in->tm_usec))
2716 6 : return DTERR_FIELD_OVERFLOW;
2717 :
2718 1984 : return 0;
2719 : }
2720 :
2721 :
2722 : /* DecodeNumber()
2723 : * Interpret plain numeric field as a date value in context.
2724 : * Return 0 if okay, a DTERR code if not.
2725 : */
2726 : static int
2727 201182 : DecodeNumber(int flen, char *str, bool haveTextMonth, int fmask,
2728 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2729 : {
2730 : int val;
2731 : char *cp;
2732 : int dterr;
2733 :
2734 201182 : *tmask = 0;
2735 :
2736 201182 : errno = 0;
2737 201182 : val = strtoint(str, &cp, 10);
2738 201182 : if (errno == ERANGE)
2739 0 : return DTERR_FIELD_OVERFLOW;
2740 201182 : if (cp == str)
2741 0 : return DTERR_BAD_FORMAT;
2742 :
2743 201182 : if (*cp == '.')
2744 : {
2745 : /*
2746 : * More than two digits before decimal point? Then could be a date or
2747 : * a run-together time: 2001.360 20011225 040506.789
2748 : */
2749 0 : if (cp - str > 2)
2750 : {
2751 0 : dterr = DecodeNumberField(flen, str,
2752 : (fmask | DTK_DATE_M),
2753 : tmask, tm,
2754 : fsec, is2digits);
2755 0 : if (dterr < 0)
2756 0 : return dterr;
2757 0 : return 0;
2758 : }
2759 :
2760 0 : dterr = ParseFractionalSecond(cp, fsec);
2761 0 : if (dterr)
2762 0 : return dterr;
2763 : }
2764 201182 : else if (*cp != '\0')
2765 0 : return DTERR_BAD_FORMAT;
2766 :
2767 : /* Special case for day of year */
2768 201182 : if (flen == 3 && (fmask & DTK_DATE_M) == DTK_M(YEAR) && val >= 1 &&
2769 : val <= 366)
2770 : {
2771 54 : *tmask = (DTK_M(DOY) | DTK_M(MONTH) | DTK_M(DAY));
2772 54 : tm->tm_yday = val;
2773 : /* tm_mon and tm_mday can't actually be set yet ... */
2774 54 : return 0;
2775 : }
2776 :
2777 : /* Switch based on what we have so far */
2778 201128 : switch (fmask & DTK_DATE_M)
2779 : {
2780 65394 : case 0:
2781 :
2782 : /*
2783 : * Nothing so far; make a decision about what we think the input
2784 : * is. There used to be lots of heuristics here, but the
2785 : * consensus now is to be paranoid. It *must* be either
2786 : * YYYY-MM-DD (with a more-than-two-digit year field), or the
2787 : * field order defined by DateOrder.
2788 : */
2789 65394 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2790 : {
2791 63654 : *tmask = DTK_M(YEAR);
2792 63654 : tm->tm_year = val;
2793 : }
2794 1740 : else if (DateOrder == DATEORDER_DMY)
2795 : {
2796 164 : *tmask = DTK_M(DAY);
2797 164 : tm->tm_mday = val;
2798 : }
2799 : else
2800 : {
2801 1576 : *tmask = DTK_M(MONTH);
2802 1576 : tm->tm_mon = val;
2803 : }
2804 65394 : break;
2805 :
2806 63546 : case (DTK_M(YEAR)):
2807 : /* Must be at second field of YY-MM-DD */
2808 63546 : *tmask = DTK_M(MONTH);
2809 63546 : tm->tm_mon = val;
2810 63546 : break;
2811 :
2812 4144 : case (DTK_M(MONTH)):
2813 4144 : if (haveTextMonth)
2814 : {
2815 : /*
2816 : * We are at the first numeric field of a date that included a
2817 : * textual month name. We want to support the variants
2818 : * MON-DD-YYYY, DD-MON-YYYY, and YYYY-MON-DD as unambiguous
2819 : * inputs. We will also accept MON-DD-YY or DD-MON-YY in
2820 : * either DMY or MDY modes, as well as YY-MON-DD in YMD mode.
2821 : */
2822 2624 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2823 : {
2824 72 : *tmask = DTK_M(YEAR);
2825 72 : tm->tm_year = val;
2826 : }
2827 : else
2828 : {
2829 2552 : *tmask = DTK_M(DAY);
2830 2552 : tm->tm_mday = val;
2831 : }
2832 : }
2833 : else
2834 : {
2835 : /* Must be at second field of MM-DD-YY */
2836 1520 : *tmask = DTK_M(DAY);
2837 1520 : tm->tm_mday = val;
2838 : }
2839 4144 : break;
2840 :
2841 63624 : case (DTK_M(YEAR) | DTK_M(MONTH)):
2842 63624 : if (haveTextMonth)
2843 : {
2844 : /* Need to accept DD-MON-YYYY even in YMD mode */
2845 126 : if (flen >= 3 && *is2digits)
2846 : {
2847 : /* Guess that first numeric field is day was wrong */
2848 30 : *tmask = DTK_M(DAY); /* YEAR is already set */
2849 30 : tm->tm_mday = tm->tm_year;
2850 30 : tm->tm_year = val;
2851 30 : *is2digits = false;
2852 : }
2853 : else
2854 : {
2855 96 : *tmask = DTK_M(DAY);
2856 96 : tm->tm_mday = val;
2857 : }
2858 : }
2859 : else
2860 : {
2861 : /* Must be at third field of YY-MM-DD */
2862 63498 : *tmask = DTK_M(DAY);
2863 63498 : tm->tm_mday = val;
2864 : }
2865 63624 : break;
2866 :
2867 146 : case (DTK_M(DAY)):
2868 : /* Must be at second field of DD-MM-YY */
2869 146 : *tmask = DTK_M(MONTH);
2870 146 : tm->tm_mon = val;
2871 146 : break;
2872 :
2873 4250 : case (DTK_M(MONTH) | DTK_M(DAY)):
2874 : /* Must be at third field of DD-MM-YY or MM-DD-YY */
2875 4250 : *tmask = DTK_M(YEAR);
2876 4250 : tm->tm_year = val;
2877 4250 : break;
2878 :
2879 24 : case (DTK_M(YEAR) | DTK_M(MONTH) | DTK_M(DAY)):
2880 : /* we have all the date, so it must be a time field */
2881 24 : dterr = DecodeNumberField(flen, str, fmask,
2882 : tmask, tm,
2883 : fsec, is2digits);
2884 24 : if (dterr < 0)
2885 12 : return dterr;
2886 12 : return 0;
2887 :
2888 0 : default:
2889 : /* Anything else is bogus input */
2890 0 : return DTERR_BAD_FORMAT;
2891 : }
2892 :
2893 : /*
2894 : * When processing a year field, mark it for adjustment if it's only one
2895 : * or two digits.
2896 : */
2897 201104 : if (*tmask == DTK_M(YEAR))
2898 67976 : *is2digits = (flen <= 2);
2899 :
2900 201104 : return 0;
2901 : }
2902 :
2903 :
2904 : /* DecodeNumberField()
2905 : * Interpret numeric string as a concatenated date or time field.
2906 : * Return a DTK token (>= 0) if successful, a DTERR code (< 0) if not.
2907 : *
2908 : * Use the context of previously decoded fields to help with
2909 : * the interpretation.
2910 : */
2911 : static int
2912 578 : DecodeNumberField(int len, char *str, int fmask,
2913 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2914 : {
2915 : char *cp;
2916 :
2917 : /*
2918 : * Have a decimal point? Then this is a date or something with a seconds
2919 : * field...
2920 : */
2921 578 : if ((cp = strchr(str, '.')) != NULL)
2922 : {
2923 : /*
2924 : * Can we use ParseFractionalSecond here? Not clear whether trailing
2925 : * junk should be rejected ...
2926 : */
2927 138 : if (cp[1] == '\0')
2928 : {
2929 : /* avoid assuming that strtod will accept "." */
2930 0 : *fsec = 0;
2931 : }
2932 : else
2933 : {
2934 : double frac;
2935 :
2936 138 : errno = 0;
2937 138 : frac = strtod(cp, NULL);
2938 138 : if (errno != 0)
2939 0 : return DTERR_BAD_FORMAT;
2940 138 : *fsec = rint(frac * 1000000);
2941 : }
2942 : /* Now truncate off the fraction for further processing */
2943 138 : *cp = '\0';
2944 138 : len = strlen(str);
2945 : }
2946 : /* No decimal point and no complete date yet? */
2947 440 : else if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2948 : {
2949 248 : if (len >= 6)
2950 : {
2951 248 : *tmask = DTK_DATE_M;
2952 :
2953 : /*
2954 : * Start from end and consider first 2 as Day, next 2 as Month,
2955 : * and the rest as Year.
2956 : */
2957 248 : tm->tm_mday = atoi(str + (len - 2));
2958 248 : *(str + (len - 2)) = '\0';
2959 248 : tm->tm_mon = atoi(str + (len - 4));
2960 248 : *(str + (len - 4)) = '\0';
2961 248 : tm->tm_year = atoi(str);
2962 248 : if ((len - 4) == 2)
2963 18 : *is2digits = true;
2964 :
2965 248 : return DTK_DATE;
2966 : }
2967 : }
2968 :
2969 : /* not all time fields are specified? */
2970 330 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2971 : {
2972 : /* hhmmss */
2973 330 : if (len == 6)
2974 : {
2975 282 : *tmask = DTK_TIME_M;
2976 282 : tm->tm_sec = atoi(str + 4);
2977 282 : *(str + 4) = '\0';
2978 282 : tm->tm_min = atoi(str + 2);
2979 282 : *(str + 2) = '\0';
2980 282 : tm->tm_hour = atoi(str);
2981 :
2982 282 : return DTK_TIME;
2983 : }
2984 : /* hhmm? */
2985 48 : else if (len == 4)
2986 : {
2987 24 : *tmask = DTK_TIME_M;
2988 24 : tm->tm_sec = 0;
2989 24 : tm->tm_min = atoi(str + 2);
2990 24 : *(str + 2) = '\0';
2991 24 : tm->tm_hour = atoi(str);
2992 :
2993 24 : return DTK_TIME;
2994 : }
2995 : }
2996 :
2997 24 : return DTERR_BAD_FORMAT;
2998 : }
2999 :
3000 :
3001 : /* DecodeTimezone()
3002 : * Interpret string as a numeric timezone.
3003 : *
3004 : * Return 0 if okay (and set *tzp), a DTERR code if not okay.
3005 : */
3006 : int
3007 37522 : DecodeTimezone(const char *str, int *tzp)
3008 : {
3009 : int tz;
3010 : int hr,
3011 : min,
3012 37522 : sec = 0;
3013 : char *cp;
3014 :
3015 : /* leading character must be "+" or "-" */
3016 37522 : if (*str != '+' && *str != '-')
3017 66 : return DTERR_BAD_FORMAT;
3018 :
3019 37456 : errno = 0;
3020 37456 : hr = strtoint(str + 1, &cp, 10);
3021 37456 : if (errno == ERANGE)
3022 0 : return DTERR_TZDISP_OVERFLOW;
3023 :
3024 : /* explicit delimiter? */
3025 37456 : if (*cp == ':')
3026 : {
3027 108 : errno = 0;
3028 108 : min = strtoint(cp + 1, &cp, 10);
3029 108 : if (errno == ERANGE)
3030 0 : return DTERR_TZDISP_OVERFLOW;
3031 108 : if (*cp == ':')
3032 : {
3033 24 : errno = 0;
3034 24 : sec = strtoint(cp + 1, &cp, 10);
3035 24 : if (errno == ERANGE)
3036 0 : return DTERR_TZDISP_OVERFLOW;
3037 : }
3038 : }
3039 : /* otherwise, might have run things together... */
3040 37348 : else if (*cp == '\0' && strlen(str) > 3)
3041 : {
3042 72 : min = hr % 100;
3043 72 : hr = hr / 100;
3044 : /* we could, but don't, support a run-together hhmmss format */
3045 : }
3046 : else
3047 37276 : min = 0;
3048 :
3049 : /* Range-check the values; see notes in datatype/timestamp.h */
3050 37456 : if (hr < 0 || hr > MAX_TZDISP_HOUR)
3051 12 : return DTERR_TZDISP_OVERFLOW;
3052 37444 : if (min < 0 || min >= MINS_PER_HOUR)
3053 12 : return DTERR_TZDISP_OVERFLOW;
3054 37432 : if (sec < 0 || sec >= SECS_PER_MINUTE)
3055 0 : return DTERR_TZDISP_OVERFLOW;
3056 :
3057 37432 : tz = (hr * MINS_PER_HOUR + min) * SECS_PER_MINUTE + sec;
3058 37432 : if (*str == '-')
3059 826 : tz = -tz;
3060 :
3061 37432 : *tzp = -tz;
3062 :
3063 37432 : if (*cp != '\0')
3064 0 : return DTERR_BAD_FORMAT;
3065 :
3066 37432 : return 0;
3067 : }
3068 :
3069 :
3070 : /* DecodeTimezoneAbbrev()
3071 : * Interpret string as a timezone abbreviation, if possible.
3072 : *
3073 : * Sets *ftype to an abbreviation type (TZ, DTZ, or DYNTZ), or UNKNOWN_FIELD if
3074 : * string is not any known abbreviation. On success, set *offset and *tz to
3075 : * represent the UTC offset (for TZ or DTZ) or underlying zone (for DYNTZ).
3076 : * Note that full timezone names (such as America/New_York) are not handled
3077 : * here, mostly for historical reasons.
3078 : *
3079 : * The function result is 0 or a DTERR code; in the latter case, *extra
3080 : * is filled as needed. Note that unknown-abbreviation is not considered
3081 : * an error case. Also note that many callers assume that the DTERR code
3082 : * is one that DateTimeParseError does not require "str" or "datatype"
3083 : * strings for.
3084 : *
3085 : * Given string must be lowercased already.
3086 : *
3087 : * Implement a cache lookup since it is likely that dates
3088 : * will be related in format.
3089 : */
3090 : int
3091 10490 : DecodeTimezoneAbbrev(int field, const char *lowtoken,
3092 : int *ftype, int *offset, pg_tz **tz,
3093 : DateTimeErrorExtra *extra)
3094 : {
3095 : const datetkn *tp;
3096 :
3097 10490 : tp = abbrevcache[field];
3098 : /* use strncmp so that we match truncated tokens */
3099 10490 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3100 : {
3101 7788 : if (zoneabbrevtbl)
3102 7788 : tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3103 7788 : zoneabbrevtbl->numabbrevs);
3104 : else
3105 0 : tp = NULL;
3106 : }
3107 10490 : if (tp == NULL)
3108 : {
3109 7366 : *ftype = UNKNOWN_FIELD;
3110 7366 : *offset = 0;
3111 7366 : *tz = NULL;
3112 : }
3113 : else
3114 : {
3115 3124 : abbrevcache[field] = tp;
3116 3124 : *ftype = tp->type;
3117 3124 : if (tp->type == DYNTZ)
3118 : {
3119 264 : *offset = 0;
3120 264 : *tz = FetchDynamicTimeZone(zoneabbrevtbl, tp, extra);
3121 264 : if (*tz == NULL)
3122 0 : return DTERR_BAD_ZONE_ABBREV;
3123 : }
3124 : else
3125 : {
3126 2860 : *offset = tp->value;
3127 2860 : *tz = NULL;
3128 : }
3129 : }
3130 :
3131 10490 : return 0;
3132 : }
3133 :
3134 :
3135 : /* DecodeSpecial()
3136 : * Decode text string using lookup table.
3137 : *
3138 : * Recognizes the keywords listed in datetktbl.
3139 : * Note: at one time this would also recognize timezone abbreviations,
3140 : * but no more; use DecodeTimezoneAbbrev for that.
3141 : *
3142 : * Given string must be lowercased already.
3143 : *
3144 : * Implement a cache lookup since it is likely that dates
3145 : * will be related in format.
3146 : */
3147 : int
3148 42186 : DecodeSpecial(int field, const char *lowtoken, int *val)
3149 : {
3150 : int type;
3151 : const datetkn *tp;
3152 :
3153 42186 : tp = datecache[field];
3154 : /* use strncmp so that we match truncated tokens */
3155 42186 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3156 : {
3157 9778 : tp = datebsearch(lowtoken, datetktbl, szdatetktbl);
3158 : }
3159 42186 : if (tp == NULL)
3160 : {
3161 108 : type = UNKNOWN_FIELD;
3162 108 : *val = 0;
3163 : }
3164 : else
3165 : {
3166 42078 : datecache[field] = tp;
3167 42078 : type = tp->type;
3168 42078 : *val = tp->value;
3169 : }
3170 :
3171 42186 : return type;
3172 : }
3173 :
3174 :
3175 : /* DecodeTimezoneName()
3176 : * Interpret string as a timezone abbreviation or name.
3177 : * Throw error if the name is not recognized.
3178 : *
3179 : * The return value indicates what kind of zone identifier it is:
3180 : * TZNAME_FIXED_OFFSET: fixed offset from UTC
3181 : * TZNAME_DYNTZ: dynamic timezone abbreviation
3182 : * TZNAME_ZONE: full tzdb zone name
3183 : *
3184 : * For TZNAME_FIXED_OFFSET, *offset receives the UTC offset (in seconds,
3185 : * with ISO sign convention: positive is east of Greenwich).
3186 : * For the other two cases, *tz receives the timezone struct representing
3187 : * the zone name or the abbreviation's underlying zone.
3188 : */
3189 : int
3190 798 : DecodeTimezoneName(const char *tzname, int *offset, pg_tz **tz)
3191 : {
3192 : char *lowzone;
3193 : int dterr,
3194 : type;
3195 : DateTimeErrorExtra extra;
3196 :
3197 : /*
3198 : * First we look in the timezone abbreviation table (to handle cases like
3199 : * "EST"), and if that fails, we look in the timezone database (to handle
3200 : * cases like "America/New_York"). This matches the order in which
3201 : * timestamp input checks the cases; it's important because the timezone
3202 : * database unwisely uses a few zone names that are identical to offset
3203 : * abbreviations.
3204 : */
3205 :
3206 : /* DecodeTimezoneAbbrev requires lowercase input */
3207 798 : lowzone = downcase_truncate_identifier(tzname,
3208 798 : strlen(tzname),
3209 : false);
3210 :
3211 798 : dterr = DecodeTimezoneAbbrev(0, lowzone, &type, offset, tz, &extra);
3212 798 : if (dterr)
3213 0 : DateTimeParseError(dterr, &extra, NULL, NULL, NULL);
3214 :
3215 798 : if (type == TZ || type == DTZ)
3216 : {
3217 : /* fixed-offset abbreviation, return the offset */
3218 276 : return TZNAME_FIXED_OFFSET;
3219 : }
3220 522 : else if (type == DYNTZ)
3221 : {
3222 : /* dynamic-offset abbreviation, return its referenced timezone */
3223 174 : return TZNAME_DYNTZ;
3224 : }
3225 : else
3226 : {
3227 : /* try it as a full zone name */
3228 348 : *tz = pg_tzset(tzname);
3229 348 : if (*tz == NULL)
3230 12 : ereport(ERROR,
3231 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
3232 : errmsg("time zone \"%s\" not recognized", tzname)));
3233 336 : return TZNAME_ZONE;
3234 : }
3235 : }
3236 :
3237 : /* DecodeTimezoneNameToTz()
3238 : * Interpret string as a timezone abbreviation or name.
3239 : * Throw error if the name is not recognized.
3240 : *
3241 : * This is a simple wrapper for DecodeTimezoneName that produces a pg_tz *
3242 : * result in all cases.
3243 : */
3244 : pg_tz *
3245 78 : DecodeTimezoneNameToTz(const char *tzname)
3246 : {
3247 : pg_tz *result;
3248 : int offset;
3249 :
3250 78 : if (DecodeTimezoneName(tzname, &offset, &result) == TZNAME_FIXED_OFFSET)
3251 : {
3252 : /* fixed-offset abbreviation, get a pg_tz descriptor for that */
3253 12 : result = pg_tzset_offset(-offset); /* flip to POSIX sign convention */
3254 : }
3255 78 : return result;
3256 : }
3257 :
3258 : /* DecodeTimezoneAbbrevPrefix()
3259 : * Interpret prefix of string as a timezone abbreviation, if possible.
3260 : *
3261 : * This has roughly the same functionality as DecodeTimezoneAbbrev(),
3262 : * but the API is adapted to the needs of formatting.c. Notably,
3263 : * we will match the longest possible prefix of the given string
3264 : * rather than insisting on a complete match, and downcasing is applied
3265 : * here rather than in the caller.
3266 : *
3267 : * Returns the length of the timezone abbreviation, or -1 if not recognized.
3268 : * On success, sets *offset to the GMT offset for the abbreviation if it
3269 : * is a fixed-offset abbreviation, or sets *tz to the pg_tz struct for
3270 : * a dynamic abbreviation.
3271 : */
3272 : int
3273 3552 : DecodeTimezoneAbbrevPrefix(const char *str, int *offset, pg_tz **tz)
3274 : {
3275 : char lowtoken[TOKMAXLEN + 1];
3276 : int len;
3277 :
3278 3552 : *offset = 0; /* avoid uninitialized vars on failure */
3279 3552 : *tz = NULL;
3280 :
3281 3552 : if (!zoneabbrevtbl)
3282 0 : return -1; /* no abbrevs known, so fail immediately */
3283 :
3284 : /* Downcase as much of the string as we could need */
3285 3672 : for (len = 0; len < TOKMAXLEN; len++)
3286 : {
3287 3672 : if (*str == '\0' || !isalpha((unsigned char) *str))
3288 : break;
3289 120 : lowtoken[len] = pg_tolower((unsigned char) *str++);
3290 : }
3291 3552 : lowtoken[len] = '\0';
3292 :
3293 : /*
3294 : * We could avoid doing repeated binary searches if we cared to duplicate
3295 : * datebsearch here, but it's not clear that such an optimization would be
3296 : * worth the trouble. In common cases there's probably not anything after
3297 : * the zone abbrev anyway. So just search with successively truncated
3298 : * strings.
3299 : */
3300 3594 : while (len > 0)
3301 : {
3302 72 : const datetkn *tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3303 72 : zoneabbrevtbl->numabbrevs);
3304 :
3305 72 : if (tp != NULL)
3306 : {
3307 30 : if (tp->type == DYNTZ)
3308 : {
3309 : DateTimeErrorExtra extra;
3310 6 : pg_tz *tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp,
3311 : &extra);
3312 :
3313 6 : if (tzp != NULL)
3314 : {
3315 : /* Caller must resolve the abbrev's current meaning */
3316 6 : *tz = tzp;
3317 6 : return len;
3318 : }
3319 : }
3320 : else
3321 : {
3322 : /* Fixed-offset zone abbrev, so it's easy */
3323 24 : *offset = tp->value;
3324 24 : return len;
3325 : }
3326 : }
3327 42 : lowtoken[--len] = '\0';
3328 : }
3329 :
3330 : /* Did not find a match */
3331 3522 : return -1;
3332 : }
3333 :
3334 :
3335 : /* ClearPgItmIn
3336 : *
3337 : * Zero out a pg_itm_in
3338 : */
3339 : static inline void
3340 12434 : ClearPgItmIn(struct pg_itm_in *itm_in)
3341 : {
3342 12434 : itm_in->tm_usec = 0;
3343 12434 : itm_in->tm_mday = 0;
3344 12434 : itm_in->tm_mon = 0;
3345 12434 : itm_in->tm_year = 0;
3346 12434 : }
3347 :
3348 :
3349 : /* DecodeInterval()
3350 : * Interpret previously parsed fields for general time interval.
3351 : * Returns 0 if successful, DTERR code if bogus input detected.
3352 : * dtype and itm_in are output parameters.
3353 : *
3354 : * Allow "date" field DTK_DATE since this could be just
3355 : * an unsigned floating point number. - thomas 1997-11-16
3356 : *
3357 : * Allow ISO-style time span, with implicit units on number of days
3358 : * preceding an hh:mm:ss field. - thomas 1998-04-30
3359 : *
3360 : * itm_in remains undefined for infinite interval values for which dtype alone
3361 : * suffices.
3362 : */
3363 : int
3364 11822 : DecodeInterval(char **field, int *ftype, int nf, int range,
3365 : int *dtype, struct pg_itm_in *itm_in)
3366 : {
3367 11822 : bool force_negative = false;
3368 11822 : bool is_before = false;
3369 11822 : bool parsing_unit_val = false;
3370 : char *cp;
3371 11822 : int fmask = 0,
3372 : tmask,
3373 : type,
3374 : uval;
3375 : int i;
3376 : int dterr;
3377 : int64 val;
3378 : double fval;
3379 :
3380 11822 : *dtype = DTK_DELTA;
3381 11822 : type = IGNORE_DTF;
3382 11822 : ClearPgItmIn(itm_in);
3383 :
3384 : /*----------
3385 : * The SQL standard defines the interval literal
3386 : * '-1 1:00:00'
3387 : * to mean "negative 1 days and negative 1 hours", while Postgres
3388 : * traditionally treats this as meaning "negative 1 days and positive
3389 : * 1 hours". In SQL_STANDARD intervalstyle, we apply the leading sign
3390 : * to all fields if there are no other explicit signs.
3391 : *
3392 : * We leave the signs alone if there are additional explicit signs.
3393 : * This protects us against misinterpreting postgres-style dump output,
3394 : * since the postgres-style output code has always put an explicit sign on
3395 : * all fields following a negative field. But note that SQL-spec output
3396 : * is ambiguous and can be misinterpreted on load! (So it's best practice
3397 : * to dump in postgres style, not SQL style.)
3398 : *----------
3399 : */
3400 11822 : if (IntervalStyle == INTSTYLE_SQL_STANDARD && nf > 0 && *field[0] == '-')
3401 : {
3402 38 : force_negative = true;
3403 : /* Check for additional explicit signs */
3404 244 : for (i = 1; i < nf; i++)
3405 : {
3406 224 : if (*field[i] == '-' || *field[i] == '+')
3407 : {
3408 18 : force_negative = false;
3409 18 : break;
3410 : }
3411 : }
3412 : }
3413 :
3414 : /* read through list backwards to pick up units before values */
3415 37106 : for (i = nf - 1; i >= 0; i--)
3416 : {
3417 26376 : switch (ftype[i])
3418 : {
3419 1234 : case DTK_TIME:
3420 1234 : dterr = DecodeTimeForInterval(field[i], fmask, range,
3421 : &tmask, itm_in);
3422 1234 : if (dterr)
3423 18 : return dterr;
3424 1216 : if (force_negative &&
3425 2 : itm_in->tm_usec > 0)
3426 2 : itm_in->tm_usec = -itm_in->tm_usec;
3427 1216 : type = DTK_DAY;
3428 1216 : parsing_unit_val = false;
3429 1216 : break;
3430 :
3431 2728 : case DTK_TZ:
3432 :
3433 : /*
3434 : * Timezone means a token with a leading sign character and at
3435 : * least one digit; there could be ':', '.', '-' embedded in
3436 : * it as well.
3437 : */
3438 : Assert(*field[i] == '-' || *field[i] == '+');
3439 :
3440 : /*
3441 : * Check for signed hh:mm or hh:mm:ss. If so, process exactly
3442 : * like DTK_TIME case above, plus handling the sign.
3443 : */
3444 3496 : if (strchr(field[i] + 1, ':') != NULL &&
3445 768 : DecodeTimeForInterval(field[i] + 1, fmask, range,
3446 : &tmask, itm_in) == 0)
3447 : {
3448 768 : if (*field[i] == '-')
3449 : {
3450 : /* flip the sign on time field */
3451 708 : if (itm_in->tm_usec == PG_INT64_MIN)
3452 0 : return DTERR_FIELD_OVERFLOW;
3453 708 : itm_in->tm_usec = -itm_in->tm_usec;
3454 : }
3455 :
3456 768 : if (force_negative &&
3457 0 : itm_in->tm_usec > 0)
3458 0 : itm_in->tm_usec = -itm_in->tm_usec;
3459 :
3460 : /*
3461 : * Set the next type to be a day, if units are not
3462 : * specified. This handles the case of '1 +02:03' since we
3463 : * are reading right to left.
3464 : */
3465 768 : type = DTK_DAY;
3466 768 : parsing_unit_val = false;
3467 768 : break;
3468 : }
3469 :
3470 : /*
3471 : * Otherwise, fall through to DTK_NUMBER case, which can
3472 : * handle signed float numbers and signed year-month values.
3473 : */
3474 :
3475 : /* FALLTHROUGH */
3476 :
3477 : case DTK_DATE:
3478 : case DTK_NUMBER:
3479 12044 : if (type == IGNORE_DTF)
3480 : {
3481 : /* use typmod to decide what rightmost field is */
3482 : switch (range)
3483 : {
3484 6 : case INTERVAL_MASK(YEAR):
3485 6 : type = DTK_YEAR;
3486 6 : break;
3487 30 : case INTERVAL_MASK(MONTH):
3488 : case INTERVAL_MASK(YEAR) | INTERVAL_MASK(MONTH):
3489 30 : type = DTK_MONTH;
3490 30 : break;
3491 18 : case INTERVAL_MASK(DAY):
3492 18 : type = DTK_DAY;
3493 18 : break;
3494 24 : case INTERVAL_MASK(HOUR):
3495 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR):
3496 24 : type = DTK_HOUR;
3497 24 : break;
3498 24 : case INTERVAL_MASK(MINUTE):
3499 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3500 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3501 24 : type = DTK_MINUTE;
3502 24 : break;
3503 60 : case INTERVAL_MASK(SECOND):
3504 : case INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3505 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3506 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3507 60 : type = DTK_SECOND;
3508 60 : break;
3509 438 : default:
3510 438 : type = DTK_SECOND;
3511 438 : break;
3512 : }
3513 11444 : }
3514 :
3515 12044 : errno = 0;
3516 12044 : val = strtoi64(field[i], &cp, 10);
3517 12044 : if (errno == ERANGE)
3518 12 : return DTERR_FIELD_OVERFLOW;
3519 :
3520 12032 : if (*cp == '-')
3521 : {
3522 : /* SQL "years-months" syntax */
3523 : int val2;
3524 :
3525 60 : val2 = strtoint(cp + 1, &cp, 10);
3526 60 : if (errno == ERANGE || val2 < 0 || val2 >= MONTHS_PER_YEAR)
3527 0 : return DTERR_FIELD_OVERFLOW;
3528 60 : if (*cp != '\0')
3529 0 : return DTERR_BAD_FORMAT;
3530 60 : type = DTK_MONTH;
3531 60 : if (*field[i] == '-')
3532 6 : val2 = -val2;
3533 60 : if (pg_mul_s64_overflow(val, MONTHS_PER_YEAR, &val))
3534 0 : return DTERR_FIELD_OVERFLOW;
3535 60 : if (pg_add_s64_overflow(val, val2, &val))
3536 0 : return DTERR_FIELD_OVERFLOW;
3537 60 : fval = 0;
3538 : }
3539 11972 : else if (*cp == '.')
3540 : {
3541 486 : dterr = ParseFraction(cp, &fval);
3542 486 : if (dterr)
3543 0 : return dterr;
3544 486 : if (*field[i] == '-')
3545 138 : fval = -fval;
3546 : }
3547 11486 : else if (*cp == '\0')
3548 11102 : fval = 0;
3549 : else
3550 384 : return DTERR_BAD_FORMAT;
3551 :
3552 11648 : tmask = 0; /* DTK_M(type); */
3553 :
3554 11648 : if (force_negative)
3555 : {
3556 : /* val and fval should be of same sign, but test anyway */
3557 80 : if (val > 0)
3558 60 : val = -val;
3559 80 : if (fval > 0)
3560 18 : fval = -fval;
3561 : }
3562 :
3563 : switch (type)
3564 : {
3565 330 : case DTK_MICROSEC:
3566 330 : if (!AdjustMicroseconds(val, fval, 1, itm_in))
3567 36 : return DTERR_FIELD_OVERFLOW;
3568 294 : tmask = DTK_M(MICROSECOND);
3569 294 : break;
3570 :
3571 104 : case DTK_MILLISEC:
3572 104 : if (!AdjustMicroseconds(val, fval, 1000, itm_in))
3573 12 : return DTERR_FIELD_OVERFLOW;
3574 92 : tmask = DTK_M(MILLISECOND);
3575 92 : break;
3576 :
3577 920 : case DTK_SECOND:
3578 920 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
3579 12 : return DTERR_FIELD_OVERFLOW;
3580 :
3581 : /*
3582 : * If any subseconds were specified, consider this
3583 : * microsecond and millisecond input as well.
3584 : */
3585 908 : if (fval == 0)
3586 746 : tmask = DTK_M(SECOND);
3587 : else
3588 162 : tmask = DTK_ALL_SECS_M;
3589 908 : break;
3590 :
3591 370 : case DTK_MINUTE:
3592 370 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
3593 12 : return DTERR_FIELD_OVERFLOW;
3594 358 : tmask = DTK_M(MINUTE);
3595 358 : break;
3596 :
3597 692 : case DTK_HOUR:
3598 692 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3599 12 : return DTERR_FIELD_OVERFLOW;
3600 680 : tmask = DTK_M(HOUR);
3601 680 : type = DTK_DAY; /* set for next field */
3602 680 : break;
3603 :
3604 7052 : case DTK_DAY:
3605 7052 : if (!AdjustDays(val, 1, itm_in) ||
3606 6980 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3607 90 : return DTERR_FIELD_OVERFLOW;
3608 6962 : tmask = DTK_M(DAY);
3609 6962 : break;
3610 :
3611 96 : case DTK_WEEK:
3612 96 : if (!AdjustDays(val, 7, itm_in) ||
3613 72 : !AdjustFractDays(fval, 7, itm_in))
3614 48 : return DTERR_FIELD_OVERFLOW;
3615 48 : tmask = DTK_M(WEEK);
3616 48 : break;
3617 :
3618 900 : case DTK_MONTH:
3619 900 : if (!AdjustMonths(val, itm_in) ||
3620 840 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3621 84 : return DTERR_FIELD_OVERFLOW;
3622 816 : tmask = DTK_M(MONTH);
3623 816 : break;
3624 :
3625 986 : case DTK_YEAR:
3626 986 : if (!AdjustYears(val, 1, itm_in) ||
3627 938 : !AdjustFractYears(fval, 1, itm_in))
3628 60 : return DTERR_FIELD_OVERFLOW;
3629 926 : tmask = DTK_M(YEAR);
3630 926 : break;
3631 :
3632 66 : case DTK_DECADE:
3633 66 : if (!AdjustYears(val, 10, itm_in) ||
3634 42 : !AdjustFractYears(fval, 10, itm_in))
3635 36 : return DTERR_FIELD_OVERFLOW;
3636 30 : tmask = DTK_M(DECADE);
3637 30 : break;
3638 :
3639 66 : case DTK_CENTURY:
3640 66 : if (!AdjustYears(val, 100, itm_in) ||
3641 42 : !AdjustFractYears(fval, 100, itm_in))
3642 36 : return DTERR_FIELD_OVERFLOW;
3643 30 : tmask = DTK_M(CENTURY);
3644 30 : break;
3645 :
3646 66 : case DTK_MILLENNIUM:
3647 66 : if (!AdjustYears(val, 1000, itm_in) ||
3648 42 : !AdjustFractYears(fval, 1000, itm_in))
3649 36 : return DTERR_FIELD_OVERFLOW;
3650 30 : tmask = DTK_M(MILLENNIUM);
3651 30 : break;
3652 :
3653 0 : default:
3654 0 : return DTERR_BAD_FORMAT;
3655 : }
3656 11174 : parsing_unit_val = false;
3657 11174 : break;
3658 :
3659 12330 : case DTK_STRING:
3660 : case DTK_SPECIAL:
3661 : /* reject consecutive unhandled units */
3662 12330 : if (parsing_unit_val)
3663 12 : return DTERR_BAD_FORMAT;
3664 12318 : type = DecodeUnits(i, field[i], &uval);
3665 12318 : if (type == UNKNOWN_FIELD)
3666 1062 : type = DecodeSpecial(i, field[i], &uval);
3667 12318 : if (type == IGNORE_DTF)
3668 0 : continue;
3669 :
3670 12318 : tmask = 0; /* DTK_M(type); */
3671 : switch (type)
3672 : {
3673 11154 : case UNITS:
3674 11154 : type = uval;
3675 11154 : parsing_unit_val = true;
3676 11154 : break;
3677 :
3678 102 : case AGO:
3679 :
3680 : /*
3681 : * "ago" is only allowed to appear at the end of the
3682 : * interval.
3683 : */
3684 102 : if (i != nf - 1)
3685 12 : return DTERR_BAD_FORMAT;
3686 90 : is_before = true;
3687 90 : type = uval;
3688 90 : break;
3689 :
3690 1026 : case RESERV:
3691 1026 : tmask = (DTK_DATE_M | DTK_TIME_M);
3692 :
3693 : /*
3694 : * Only reserved words corresponding to infinite
3695 : * intervals are accepted.
3696 : */
3697 1026 : if (uval != DTK_LATE && uval != DTK_EARLY)
3698 36 : return DTERR_BAD_FORMAT;
3699 :
3700 : /*
3701 : * Infinity cannot be followed by anything else. We
3702 : * could allow "ago" to reverse the sign of infinity
3703 : * but using signed infinity is more intuitive.
3704 : */
3705 990 : if (i != nf - 1)
3706 12 : return DTERR_BAD_FORMAT;
3707 :
3708 978 : *dtype = uval;
3709 978 : break;
3710 :
3711 36 : default:
3712 36 : return DTERR_BAD_FORMAT;
3713 : }
3714 12222 : break;
3715 :
3716 0 : default:
3717 0 : return DTERR_BAD_FORMAT;
3718 : }
3719 :
3720 25380 : if (tmask & fmask)
3721 96 : return DTERR_BAD_FORMAT;
3722 25284 : fmask |= tmask;
3723 : }
3724 :
3725 : /* ensure that at least one time field has been found */
3726 10730 : if (fmask == 0)
3727 6 : return DTERR_BAD_FORMAT;
3728 :
3729 : /* reject if unit appeared and was never handled */
3730 10724 : if (parsing_unit_val)
3731 6 : return DTERR_BAD_FORMAT;
3732 :
3733 : /* finally, AGO negates everything */
3734 10718 : if (is_before)
3735 : {
3736 42 : if (itm_in->tm_usec == PG_INT64_MIN ||
3737 30 : itm_in->tm_mday == INT_MIN ||
3738 24 : itm_in->tm_mon == INT_MIN ||
3739 18 : itm_in->tm_year == INT_MIN)
3740 24 : return DTERR_FIELD_OVERFLOW;
3741 :
3742 18 : itm_in->tm_usec = -itm_in->tm_usec;
3743 18 : itm_in->tm_mday = -itm_in->tm_mday;
3744 18 : itm_in->tm_mon = -itm_in->tm_mon;
3745 18 : itm_in->tm_year = -itm_in->tm_year;
3746 : }
3747 :
3748 10694 : return 0;
3749 : }
3750 :
3751 :
3752 : /*
3753 : * Helper functions to avoid duplicated code in DecodeISO8601Interval.
3754 : *
3755 : * Parse a decimal value and break it into integer and fractional parts.
3756 : * Set *endptr to end+1 of the parsed substring.
3757 : * Returns 0 or DTERR code.
3758 : */
3759 : static int
3760 954 : ParseISO8601Number(char *str, char **endptr, int64 *ipart, double *fpart)
3761 : {
3762 : double val;
3763 :
3764 : /*
3765 : * Historically this has accepted anything that strtod() would take,
3766 : * notably including "e" notation, so continue doing that. This is
3767 : * slightly annoying because the precision of double is less than that of
3768 : * int64, so we would lose accuracy for inputs larger than 2^53 or so.
3769 : * However, historically we rejected inputs outside the int32 range,
3770 : * making that concern moot. What we do now is reject abs(val) above
3771 : * 1.0e15 (a round number a bit less than 2^50), so that any accepted
3772 : * value will have an exact integer part, and thereby a fraction part with
3773 : * abs(*fpart) less than 1. In the absence of field complaints it doesn't
3774 : * seem worth working harder.
3775 : */
3776 954 : if (!(isdigit((unsigned char) *str) || *str == '-' || *str == '.'))
3777 0 : return DTERR_BAD_FORMAT;
3778 954 : errno = 0;
3779 954 : val = strtod(str, endptr);
3780 : /* did we not see anything that looks like a double? */
3781 954 : if (*endptr == str || errno != 0)
3782 6 : return DTERR_BAD_FORMAT;
3783 : /* watch out for overflow, including infinities; reject NaN too */
3784 948 : if (isnan(val) || val < -1.0e15 || val > 1.0e15)
3785 0 : return DTERR_FIELD_OVERFLOW;
3786 : /* be very sure we truncate towards zero (cf dtrunc()) */
3787 948 : if (val >= 0)
3788 732 : *ipart = (int64) floor(val);
3789 : else
3790 216 : *ipart = (int64) -floor(-val);
3791 948 : *fpart = val - *ipart;
3792 : /* Callers expect this to hold */
3793 : Assert(*fpart > -1.0 && *fpart < 1.0);
3794 948 : return 0;
3795 : }
3796 :
3797 : /*
3798 : * Determine number of integral digits in a valid ISO 8601 number field
3799 : * (we should ignore sign and any fraction part)
3800 : */
3801 : static int
3802 66 : ISO8601IntegerWidth(char *fieldstart)
3803 : {
3804 : /* We might have had a leading '-' */
3805 66 : if (*fieldstart == '-')
3806 18 : fieldstart++;
3807 66 : return strspn(fieldstart, "0123456789");
3808 : }
3809 :
3810 :
3811 : /* DecodeISO8601Interval()
3812 : * Decode an ISO 8601 time interval of the "format with designators"
3813 : * (section 4.4.3.2) or "alternative format" (section 4.4.3.3)
3814 : * Examples: P1D for 1 day
3815 : * PT1H for 1 hour
3816 : * P2Y6M7DT1H30M for 2 years, 6 months, 7 days 1 hour 30 min
3817 : * P0002-06-07T01:30:00 the same value in alternative format
3818 : *
3819 : * Returns 0 if successful, DTERR code if bogus input detected.
3820 : * Note: error code should be DTERR_BAD_FORMAT if input doesn't look like
3821 : * ISO8601, otherwise this could cause unexpected error messages.
3822 : * dtype and itm_in are output parameters.
3823 : *
3824 : * A couple exceptions from the spec:
3825 : * - a week field ('W') may coexist with other units
3826 : * - allows decimals in fields other than the least significant unit.
3827 : */
3828 : int
3829 612 : DecodeISO8601Interval(char *str,
3830 : int *dtype, struct pg_itm_in *itm_in)
3831 : {
3832 612 : bool datepart = true;
3833 612 : bool havefield = false;
3834 :
3835 612 : *dtype = DTK_DELTA;
3836 612 : ClearPgItmIn(itm_in);
3837 :
3838 612 : if (strlen(str) < 2 || str[0] != 'P')
3839 228 : return DTERR_BAD_FORMAT;
3840 :
3841 384 : str++;
3842 1122 : while (*str)
3843 : {
3844 : char *fieldstart;
3845 : int64 val;
3846 : double fval;
3847 : char unit;
3848 : int dterr;
3849 :
3850 1014 : if (*str == 'T') /* T indicates the beginning of the time part */
3851 : {
3852 198 : datepart = false;
3853 198 : havefield = false;
3854 198 : str++;
3855 240 : continue;
3856 : }
3857 :
3858 816 : fieldstart = str;
3859 816 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3860 816 : if (dterr)
3861 276 : return dterr;
3862 :
3863 : /*
3864 : * Note: we could step off the end of the string here. Code below
3865 : * *must* exit the loop if unit == '\0'.
3866 : */
3867 810 : unit = *str++;
3868 :
3869 810 : if (datepart)
3870 : {
3871 468 : switch (unit) /* before T: Y M W D */
3872 : {
3873 84 : case 'Y':
3874 84 : if (!AdjustYears(val, 1, itm_in) ||
3875 84 : !AdjustFractYears(fval, 1, itm_in))
3876 12 : return DTERR_FIELD_OVERFLOW;
3877 72 : break;
3878 108 : case 'M':
3879 108 : if (!AdjustMonths(val, itm_in) ||
3880 96 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3881 24 : return DTERR_FIELD_OVERFLOW;
3882 84 : break;
3883 54 : case 'W':
3884 54 : if (!AdjustDays(val, 7, itm_in) ||
3885 42 : !AdjustFractDays(fval, 7, itm_in))
3886 24 : return DTERR_FIELD_OVERFLOW;
3887 30 : break;
3888 132 : case 'D':
3889 132 : if (!AdjustDays(val, 1, itm_in) ||
3890 96 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3891 36 : return DTERR_FIELD_OVERFLOW;
3892 96 : break;
3893 30 : case 'T': /* ISO 8601 4.4.3.3 Alternative Format / Basic */
3894 : case '\0':
3895 30 : if (ISO8601IntegerWidth(fieldstart) == 8 && !havefield)
3896 : {
3897 6 : if (!AdjustYears(val / 10000, 1, itm_in) ||
3898 6 : !AdjustMonths((val / 100) % 100, itm_in) ||
3899 6 : !AdjustDays(val % 100, 1, itm_in) ||
3900 6 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3901 0 : return DTERR_FIELD_OVERFLOW;
3902 6 : if (unit == '\0')
3903 0 : return 0;
3904 6 : datepart = false;
3905 6 : havefield = false;
3906 6 : continue;
3907 : }
3908 : /* Else fall through to extended alternative format */
3909 : /* FALLTHROUGH */
3910 : case '-': /* ISO 8601 4.4.3.3 Alternative Format,
3911 : * Extended */
3912 84 : if (havefield)
3913 0 : return DTERR_BAD_FORMAT;
3914 :
3915 84 : if (!AdjustYears(val, 1, itm_in) ||
3916 72 : !AdjustFractYears(fval, 1, itm_in))
3917 12 : return DTERR_FIELD_OVERFLOW;
3918 72 : if (unit == '\0')
3919 6 : return 0;
3920 66 : if (unit == 'T')
3921 : {
3922 6 : datepart = false;
3923 6 : havefield = false;
3924 6 : continue;
3925 : }
3926 :
3927 60 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3928 60 : if (dterr)
3929 0 : return dterr;
3930 60 : if (!AdjustMonths(val, itm_in) ||
3931 54 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3932 6 : return DTERR_FIELD_OVERFLOW;
3933 54 : if (*str == '\0')
3934 6 : return 0;
3935 48 : if (*str == 'T')
3936 : {
3937 6 : datepart = false;
3938 6 : havefield = false;
3939 6 : continue;
3940 : }
3941 42 : if (*str != '-')
3942 0 : return DTERR_BAD_FORMAT;
3943 42 : str++;
3944 :
3945 42 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3946 42 : if (dterr)
3947 0 : return dterr;
3948 42 : if (!AdjustDays(val, 1, itm_in) ||
3949 36 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3950 6 : return DTERR_FIELD_OVERFLOW;
3951 36 : if (*str == '\0')
3952 12 : return 0;
3953 24 : if (*str == 'T')
3954 : {
3955 24 : datepart = false;
3956 24 : havefield = false;
3957 24 : continue;
3958 : }
3959 0 : return DTERR_BAD_FORMAT;
3960 0 : default:
3961 : /* not a valid date unit suffix */
3962 0 : return DTERR_BAD_FORMAT;
3963 : }
3964 : }
3965 : else
3966 : {
3967 342 : switch (unit) /* after T: H M S */
3968 : {
3969 108 : case 'H':
3970 108 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3971 36 : return DTERR_FIELD_OVERFLOW;
3972 72 : break;
3973 60 : case 'M':
3974 60 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
3975 0 : return DTERR_FIELD_OVERFLOW;
3976 60 : break;
3977 96 : case 'S':
3978 96 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
3979 12 : return DTERR_FIELD_OVERFLOW;
3980 84 : break;
3981 36 : case '\0': /* ISO 8601 4.4.3.3 Alternative Format */
3982 36 : if (ISO8601IntegerWidth(fieldstart) == 6 && !havefield)
3983 : {
3984 6 : if (!AdjustMicroseconds(val / 10000, 0, USECS_PER_HOUR, itm_in) ||
3985 6 : !AdjustMicroseconds((val / 100) % 100, 0, USECS_PER_MINUTE, itm_in) ||
3986 6 : !AdjustMicroseconds(val % 100, 0, USECS_PER_SEC, itm_in) ||
3987 6 : !AdjustFractMicroseconds(fval, 1, itm_in))
3988 0 : return DTERR_FIELD_OVERFLOW;
3989 6 : return 0;
3990 : }
3991 : /* Else fall through to extended alternative format */
3992 : /* FALLTHROUGH */
3993 : case ':': /* ISO 8601 4.4.3.3 Alternative Format,
3994 : * Extended */
3995 72 : if (havefield)
3996 0 : return DTERR_BAD_FORMAT;
3997 :
3998 72 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3999 30 : return DTERR_FIELD_OVERFLOW;
4000 42 : if (unit == '\0')
4001 18 : return 0;
4002 :
4003 24 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4004 24 : if (dterr)
4005 0 : return dterr;
4006 24 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
4007 6 : return DTERR_FIELD_OVERFLOW;
4008 18 : if (*str == '\0')
4009 6 : return 0;
4010 12 : if (*str != ':')
4011 0 : return DTERR_BAD_FORMAT;
4012 12 : str++;
4013 :
4014 12 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4015 12 : if (dterr)
4016 0 : return dterr;
4017 12 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
4018 0 : return DTERR_FIELD_OVERFLOW;
4019 12 : if (*str == '\0')
4020 12 : return 0;
4021 0 : return DTERR_BAD_FORMAT;
4022 :
4023 0 : default:
4024 : /* not a valid time unit suffix */
4025 0 : return DTERR_BAD_FORMAT;
4026 : }
4027 : }
4028 :
4029 498 : havefield = true;
4030 : }
4031 :
4032 108 : return 0;
4033 : }
4034 :
4035 :
4036 : /* DecodeUnits()
4037 : * Decode text string using lookup table.
4038 : *
4039 : * This routine recognizes keywords associated with time interval units.
4040 : *
4041 : * Given string must be lowercased already.
4042 : *
4043 : * Implement a cache lookup since it is likely that dates
4044 : * will be related in format.
4045 : */
4046 : int
4047 66344 : DecodeUnits(int field, const char *lowtoken, int *val)
4048 : {
4049 : int type;
4050 : const datetkn *tp;
4051 :
4052 66344 : tp = deltacache[field];
4053 : /* use strncmp so that we match truncated tokens */
4054 66344 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
4055 : {
4056 54344 : tp = datebsearch(lowtoken, deltatktbl, szdeltatktbl);
4057 : }
4058 66344 : if (tp == NULL)
4059 : {
4060 35024 : type = UNKNOWN_FIELD;
4061 35024 : *val = 0;
4062 : }
4063 : else
4064 : {
4065 31320 : deltacache[field] = tp;
4066 31320 : type = tp->type;
4067 31320 : *val = tp->value;
4068 : }
4069 :
4070 66344 : return type;
4071 : } /* DecodeUnits() */
4072 :
4073 : /*
4074 : * Report an error detected by one of the datetime input processing routines.
4075 : *
4076 : * dterr is the error code, and *extra contains any auxiliary info we need
4077 : * for the error report. extra can be NULL if not needed for the particular
4078 : * dterr value.
4079 : *
4080 : * str is the original input string, and datatype is the name of the datatype
4081 : * we were trying to accept. (For some DTERR codes, these are not used and
4082 : * can be NULL.)
4083 : *
4084 : * If escontext points to an ErrorSaveContext node, that is filled instead
4085 : * of throwing an error.
4086 : *
4087 : * Note: it might seem useless to distinguish DTERR_INTERVAL_OVERFLOW and
4088 : * DTERR_TZDISP_OVERFLOW from DTERR_FIELD_OVERFLOW, but SQL99 mandates three
4089 : * separate SQLSTATE codes, so ...
4090 : */
4091 : void
4092 1692 : DateTimeParseError(int dterr, DateTimeErrorExtra *extra,
4093 : const char *str, const char *datatype,
4094 : Node *escontext)
4095 : {
4096 1692 : switch (dterr)
4097 : {
4098 174 : case DTERR_FIELD_OVERFLOW:
4099 174 : errsave(escontext,
4100 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4101 : errmsg("date/time field value out of range: \"%s\"",
4102 : str)));
4103 24 : break;
4104 180 : case DTERR_MD_FIELD_OVERFLOW:
4105 : /* <nanny>same as above, but add hint about DateStyle</nanny> */
4106 180 : errsave(escontext,
4107 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4108 : errmsg("date/time field value out of range: \"%s\"",
4109 : str),
4110 : errhint("Perhaps you need a different \"DateStyle\" setting.")));
4111 0 : break;
4112 720 : case DTERR_INTERVAL_OVERFLOW:
4113 720 : errsave(escontext,
4114 : (errcode(ERRCODE_INTERVAL_FIELD_OVERFLOW),
4115 : errmsg("interval field value out of range: \"%s\"",
4116 : str)));
4117 0 : break;
4118 12 : case DTERR_TZDISP_OVERFLOW:
4119 12 : errsave(escontext,
4120 : (errcode(ERRCODE_INVALID_TIME_ZONE_DISPLACEMENT_VALUE),
4121 : errmsg("time zone displacement out of range: \"%s\"",
4122 : str)));
4123 0 : break;
4124 36 : case DTERR_BAD_TIMEZONE:
4125 36 : errsave(escontext,
4126 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
4127 : errmsg("time zone \"%s\" not recognized",
4128 : extra->dtee_timezone)));
4129 24 : break;
4130 0 : case DTERR_BAD_ZONE_ABBREV:
4131 0 : errsave(escontext,
4132 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
4133 : errmsg("time zone \"%s\" not recognized",
4134 : extra->dtee_timezone),
4135 : errdetail("This time zone name appears in the configuration file for time zone abbreviation \"%s\".",
4136 : extra->dtee_abbrev)));
4137 0 : break;
4138 570 : case DTERR_BAD_FORMAT:
4139 : default:
4140 570 : errsave(escontext,
4141 : (errcode(ERRCODE_INVALID_DATETIME_FORMAT),
4142 : errmsg("invalid input syntax for type %s: \"%s\"",
4143 : datatype, str)));
4144 84 : break;
4145 : }
4146 132 : }
4147 :
4148 : /* datebsearch()
4149 : * Binary search -- from Knuth (6.2.1) Algorithm B. Special case like this
4150 : * is WAY faster than the generic bsearch().
4151 : */
4152 : static const datetkn *
4153 73884 : datebsearch(const char *key, const datetkn *base, int nel)
4154 : {
4155 73884 : if (nel > 0)
4156 : {
4157 73884 : const datetkn *last = base + nel - 1,
4158 : *position;
4159 : int result;
4160 :
4161 467454 : while (last >= base)
4162 : {
4163 423402 : position = base + ((last - base) >> 1);
4164 : /* precheck the first character for a bit of extra speed */
4165 423402 : result = (int) key[0] - (int) position->token[0];
4166 423402 : if (result == 0)
4167 : {
4168 : /* use strncmp so that we match truncated tokens */
4169 108058 : result = strncmp(key, position->token, TOKMAXLEN);
4170 108058 : if (result == 0)
4171 29832 : return position;
4172 : }
4173 393570 : if (result < 0)
4174 195204 : last = position - 1;
4175 : else
4176 198366 : base = position + 1;
4177 : }
4178 : }
4179 44052 : return NULL;
4180 : }
4181 :
4182 : /* EncodeTimezone()
4183 : * Copies representation of a numeric timezone offset to str.
4184 : *
4185 : * Returns a pointer to the new end of string. No NUL terminator is put
4186 : * there; callers are responsible for NUL terminating str themselves.
4187 : */
4188 : static char *
4189 59996 : EncodeTimezone(char *str, int tz, int style)
4190 : {
4191 : int hour,
4192 : min,
4193 : sec;
4194 :
4195 59996 : sec = abs(tz);
4196 59996 : min = sec / SECS_PER_MINUTE;
4197 59996 : sec -= min * SECS_PER_MINUTE;
4198 59996 : hour = min / MINS_PER_HOUR;
4199 59996 : min -= hour * MINS_PER_HOUR;
4200 :
4201 : /* TZ is negated compared to sign we wish to display ... */
4202 59996 : *str++ = (tz <= 0 ? '+' : '-');
4203 :
4204 59996 : if (sec != 0)
4205 : {
4206 0 : str = pg_ultostr_zeropad(str, hour, 2);
4207 0 : *str++ = ':';
4208 0 : str = pg_ultostr_zeropad(str, min, 2);
4209 0 : *str++ = ':';
4210 0 : str = pg_ultostr_zeropad(str, sec, 2);
4211 : }
4212 59996 : else if (min != 0 || style == USE_XSD_DATES)
4213 : {
4214 852 : str = pg_ultostr_zeropad(str, hour, 2);
4215 852 : *str++ = ':';
4216 852 : str = pg_ultostr_zeropad(str, min, 2);
4217 : }
4218 : else
4219 59144 : str = pg_ultostr_zeropad(str, hour, 2);
4220 59996 : return str;
4221 : }
4222 :
4223 : /* EncodeDateOnly()
4224 : * Encode date as local time.
4225 : */
4226 : void
4227 12230 : EncodeDateOnly(struct pg_tm *tm, int style, char *str)
4228 : {
4229 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4230 :
4231 12230 : switch (style)
4232 : {
4233 7002 : case USE_ISO_DATES:
4234 : case USE_XSD_DATES:
4235 : /* compatible with ISO date formats */
4236 7002 : str = pg_ultostr_zeropad(str,
4237 7002 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4238 7002 : *str++ = '-';
4239 7002 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4240 7002 : *str++ = '-';
4241 7002 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4242 7002 : break;
4243 :
4244 0 : case USE_SQL_DATES:
4245 : /* compatible with Oracle/Ingres date formats */
4246 0 : if (DateOrder == DATEORDER_DMY)
4247 : {
4248 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4249 0 : *str++ = '/';
4250 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4251 : }
4252 : else
4253 : {
4254 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4255 0 : *str++ = '/';
4256 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4257 : }
4258 0 : *str++ = '/';
4259 0 : str = pg_ultostr_zeropad(str,
4260 0 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4261 0 : break;
4262 :
4263 8 : case USE_GERMAN_DATES:
4264 : /* German-style date format */
4265 8 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4266 8 : *str++ = '.';
4267 8 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4268 8 : *str++ = '.';
4269 8 : str = pg_ultostr_zeropad(str,
4270 8 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4271 8 : break;
4272 :
4273 5220 : case USE_POSTGRES_DATES:
4274 : default:
4275 : /* traditional date-only style for Postgres */
4276 5220 : if (DateOrder == DATEORDER_DMY)
4277 : {
4278 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4279 0 : *str++ = '-';
4280 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4281 : }
4282 : else
4283 : {
4284 5220 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4285 5220 : *str++ = '-';
4286 5220 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4287 : }
4288 5220 : *str++ = '-';
4289 5220 : str = pg_ultostr_zeropad(str,
4290 5220 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4291 5220 : break;
4292 : }
4293 :
4294 12230 : if (tm->tm_year <= 0)
4295 : {
4296 80 : memcpy(str, " BC", 3); /* Don't copy NUL */
4297 80 : str += 3;
4298 : }
4299 12230 : *str = '\0';
4300 12230 : }
4301 :
4302 :
4303 : /* EncodeTimeOnly()
4304 : * Encode time fields only.
4305 : *
4306 : * tm and fsec are the value to encode, print_tz determines whether to include
4307 : * a time zone (the difference between time and timetz types), tz is the
4308 : * numeric time zone offset, style is the date style, str is where to write the
4309 : * output.
4310 : */
4311 : void
4312 12600 : EncodeTimeOnly(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, int style, char *str)
4313 : {
4314 12600 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4315 12600 : *str++ = ':';
4316 12600 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4317 12600 : *str++ = ':';
4318 12600 : str = AppendSeconds(str, tm->tm_sec, fsec, MAX_TIME_PRECISION, true);
4319 12600 : if (print_tz)
4320 6734 : str = EncodeTimezone(str, tz, style);
4321 12600 : *str = '\0';
4322 12600 : }
4323 :
4324 :
4325 : /* EncodeDateTime()
4326 : * Encode date and time interpreted as local time.
4327 : *
4328 : * tm and fsec are the value to encode, print_tz determines whether to include
4329 : * a time zone (the difference between timestamp and timestamptz types), tz is
4330 : * the numeric time zone offset, tzn is the textual time zone, which if
4331 : * specified will be used instead of tz by some styles, style is the date
4332 : * style, str is where to write the output.
4333 : *
4334 : * Supported date styles:
4335 : * Postgres - day mon hh:mm:ss yyyy tz
4336 : * SQL - mm/dd/yyyy hh:mm:ss.ss tz
4337 : * ISO - yyyy-mm-dd hh:mm:ss+/-tz
4338 : * German - dd.mm.yyyy hh:mm:ss tz
4339 : * XSD - yyyy-mm-ddThh:mm:ss.ss+/-tz
4340 : */
4341 : void
4342 113750 : EncodeDateTime(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, const char *tzn, int style, char *str)
4343 : {
4344 : int day;
4345 :
4346 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4347 :
4348 : /*
4349 : * Negative tm_isdst means we have no valid time zone translation.
4350 : */
4351 113750 : if (tm->tm_isdst < 0)
4352 41874 : print_tz = false;
4353 :
4354 113750 : switch (style)
4355 : {
4356 84008 : case USE_ISO_DATES:
4357 : case USE_XSD_DATES:
4358 : /* Compatible with ISO-8601 date formats */
4359 84008 : str = pg_ultostr_zeropad(str,
4360 84008 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4361 84008 : *str++ = '-';
4362 84008 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4363 84008 : *str++ = '-';
4364 84008 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4365 84008 : *str++ = (style == USE_ISO_DATES) ? ' ' : 'T';
4366 84008 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4367 84008 : *str++ = ':';
4368 84008 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4369 84008 : *str++ = ':';
4370 84008 : str = AppendTimestampSeconds(str, tm, fsec);
4371 84008 : if (print_tz)
4372 53262 : str = EncodeTimezone(str, tz, style);
4373 84008 : break;
4374 :
4375 780 : case USE_SQL_DATES:
4376 : /* Compatible with Oracle/Ingres date formats */
4377 780 : if (DateOrder == DATEORDER_DMY)
4378 : {
4379 384 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4380 384 : *str++ = '/';
4381 384 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4382 : }
4383 : else
4384 : {
4385 396 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4386 396 : *str++ = '/';
4387 396 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4388 : }
4389 780 : *str++ = '/';
4390 780 : str = pg_ultostr_zeropad(str,
4391 780 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4392 780 : *str++ = ' ';
4393 780 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4394 780 : *str++ = ':';
4395 780 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4396 780 : *str++ = ':';
4397 780 : str = AppendTimestampSeconds(str, tm, fsec);
4398 :
4399 : /*
4400 : * Note: the uses of %.*s in this function would be risky if the
4401 : * timezone names ever contain non-ASCII characters, since we are
4402 : * not being careful to do encoding-aware clipping. However, all
4403 : * TZ abbreviations in the IANA database are plain ASCII.
4404 : */
4405 780 : if (print_tz)
4406 : {
4407 18 : if (tzn)
4408 : {
4409 18 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4410 18 : str += strlen(str);
4411 : }
4412 : else
4413 0 : str = EncodeTimezone(str, tz, style);
4414 : }
4415 780 : break;
4416 :
4417 24 : case USE_GERMAN_DATES:
4418 : /* German variant on European style */
4419 24 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4420 24 : *str++ = '.';
4421 24 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4422 24 : *str++ = '.';
4423 24 : str = pg_ultostr_zeropad(str,
4424 24 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4425 24 : *str++ = ' ';
4426 24 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4427 24 : *str++ = ':';
4428 24 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4429 24 : *str++ = ':';
4430 24 : str = AppendTimestampSeconds(str, tm, fsec);
4431 :
4432 24 : if (print_tz)
4433 : {
4434 24 : if (tzn)
4435 : {
4436 24 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4437 24 : str += strlen(str);
4438 : }
4439 : else
4440 0 : str = EncodeTimezone(str, tz, style);
4441 : }
4442 24 : break;
4443 :
4444 28938 : case USE_POSTGRES_DATES:
4445 : default:
4446 : /* Backward-compatible with traditional Postgres abstime dates */
4447 28938 : day = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday);
4448 28938 : tm->tm_wday = j2day(day);
4449 28938 : memcpy(str, days[tm->tm_wday], 3);
4450 28938 : str += 3;
4451 28938 : *str++ = ' ';
4452 28938 : if (DateOrder == DATEORDER_DMY)
4453 : {
4454 398 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4455 398 : *str++ = ' ';
4456 398 : memcpy(str, months[tm->tm_mon - 1], 3);
4457 398 : str += 3;
4458 : }
4459 : else
4460 : {
4461 28540 : memcpy(str, months[tm->tm_mon - 1], 3);
4462 28540 : str += 3;
4463 28540 : *str++ = ' ';
4464 28540 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4465 : }
4466 28938 : *str++ = ' ';
4467 28938 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4468 28938 : *str++ = ':';
4469 28938 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4470 28938 : *str++ = ':';
4471 28938 : str = AppendTimestampSeconds(str, tm, fsec);
4472 28938 : *str++ = ' ';
4473 28938 : str = pg_ultostr_zeropad(str,
4474 28938 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4475 :
4476 28938 : if (print_tz)
4477 : {
4478 18572 : if (tzn)
4479 : {
4480 18572 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4481 18572 : str += strlen(str);
4482 : }
4483 : else
4484 : {
4485 : /*
4486 : * We have a time zone, but no string version. Use the
4487 : * numeric form, but be sure to include a leading space to
4488 : * avoid formatting something which would be rejected by
4489 : * the date/time parser later. - thomas 2001-10-19
4490 : */
4491 0 : *str++ = ' ';
4492 0 : str = EncodeTimezone(str, tz, style);
4493 : }
4494 : }
4495 28938 : break;
4496 : }
4497 :
4498 113750 : if (tm->tm_year <= 0)
4499 : {
4500 274 : memcpy(str, " BC", 3); /* Don't copy NUL */
4501 274 : str += 3;
4502 : }
4503 113750 : *str = '\0';
4504 113750 : }
4505 :
4506 :
4507 : /*
4508 : * Helper functions to avoid duplicated code in EncodeInterval.
4509 : */
4510 :
4511 : /* Append an ISO-8601-style interval field, but only if value isn't zero */
4512 : static char *
4513 210 : AddISO8601IntPart(char *cp, int64 value, char units)
4514 : {
4515 210 : if (value == 0)
4516 54 : return cp;
4517 156 : sprintf(cp, "%lld%c", (long long) value, units);
4518 156 : return cp + strlen(cp);
4519 : }
4520 :
4521 : /* Append a postgres-style interval field, but only if value isn't zero */
4522 : static char *
4523 14436 : AddPostgresIntPart(char *cp, int64 value, const char *units,
4524 : bool *is_zero, bool *is_before)
4525 : {
4526 14436 : if (value == 0)
4527 7998 : return cp;
4528 19314 : sprintf(cp, "%s%s%lld %s%s",
4529 6438 : (!*is_zero) ? " " : "",
4530 6438 : (*is_before && value > 0) ? "+" : "",
4531 : (long long) value,
4532 : units,
4533 : (value != 1) ? "s" : "");
4534 :
4535 : /*
4536 : * Each nonzero field sets is_before for (only) the next one. This is a
4537 : * tad bizarre but it's how it worked before...
4538 : */
4539 6438 : *is_before = (value < 0);
4540 6438 : *is_zero = false;
4541 6438 : return cp + strlen(cp);
4542 : }
4543 :
4544 : /* Append a verbose-style interval field, but only if value isn't zero */
4545 : static char *
4546 42490 : AddVerboseIntPart(char *cp, int64 value, const char *units,
4547 : bool *is_zero, bool *is_before)
4548 : {
4549 42490 : if (value == 0)
4550 28208 : return cp;
4551 : /* first nonzero value sets is_before */
4552 14282 : if (*is_zero)
4553 : {
4554 7874 : *is_before = (value < 0);
4555 7874 : value = i64abs(value);
4556 : }
4557 6408 : else if (*is_before)
4558 1350 : value = -value;
4559 14282 : sprintf(cp, " %lld %s%s", (long long) value, units, (value == 1) ? "" : "s");
4560 14282 : *is_zero = false;
4561 14282 : return cp + strlen(cp);
4562 : }
4563 :
4564 :
4565 : /* EncodeInterval()
4566 : * Interpret time structure as a delta time and convert to string.
4567 : *
4568 : * Support "traditional Postgres" and ISO-8601 styles.
4569 : * Actually, afaik ISO does not address time interval formatting,
4570 : * but this looks similar to the spec for absolute date/time.
4571 : * - thomas 1998-04-30
4572 : *
4573 : * Actually, afaik, ISO 8601 does specify formats for "time
4574 : * intervals...[of the]...format with time-unit designators", which
4575 : * are pretty ugly. The format looks something like
4576 : * P1Y1M1DT1H1M1.12345S
4577 : * but useful for exchanging data with computers instead of humans.
4578 : * - ron 2003-07-14
4579 : *
4580 : * And ISO's SQL 2008 standard specifies standards for
4581 : * "year-month literal"s (that look like '2-3') and
4582 : * "day-time literal"s (that look like ('4 5:6:7')
4583 : */
4584 : void
4585 13484 : EncodeInterval(struct pg_itm *itm, int style, char *str)
4586 : {
4587 13484 : char *cp = str;
4588 13484 : int year = itm->tm_year;
4589 13484 : int mon = itm->tm_mon;
4590 13484 : int64 mday = itm->tm_mday; /* tm_mday could be INT_MIN */
4591 13484 : int64 hour = itm->tm_hour;
4592 13484 : int min = itm->tm_min;
4593 13484 : int sec = itm->tm_sec;
4594 13484 : int fsec = itm->tm_usec;
4595 13484 : bool is_before = false;
4596 13484 : bool is_zero = true;
4597 :
4598 : /*
4599 : * The sign of year and month are guaranteed to match, since they are
4600 : * stored internally as "month". But we'll need to check for is_before and
4601 : * is_zero when determining the signs of day and hour/minute/seconds
4602 : * fields.
4603 : */
4604 13484 : switch (style)
4605 : {
4606 : /* SQL Standard interval format */
4607 126 : case INTSTYLE_SQL_STANDARD:
4608 : {
4609 96 : bool has_negative = year < 0 || mon < 0 ||
4610 66 : mday < 0 || hour < 0 ||
4611 222 : min < 0 || sec < 0 || fsec < 0;
4612 102 : bool has_positive = year > 0 || mon > 0 ||
4613 66 : mday > 0 || hour > 0 ||
4614 228 : min > 0 || sec > 0 || fsec > 0;
4615 126 : bool has_year_month = year != 0 || mon != 0;
4616 42 : bool has_day_time = mday != 0 || hour != 0 ||
4617 168 : min != 0 || sec != 0 || fsec != 0;
4618 126 : bool has_day = mday != 0;
4619 222 : bool sql_standard_value = !(has_negative && has_positive) &&
4620 96 : !(has_year_month && has_day_time);
4621 :
4622 : /*
4623 : * SQL Standard wants only 1 "<sign>" preceding the whole
4624 : * interval ... but can't do that if mixed signs.
4625 : */
4626 126 : if (has_negative && sql_standard_value)
4627 : {
4628 30 : *cp++ = '-';
4629 30 : year = -year;
4630 30 : mon = -mon;
4631 30 : mday = -mday;
4632 30 : hour = -hour;
4633 30 : min = -min;
4634 30 : sec = -sec;
4635 30 : fsec = -fsec;
4636 : }
4637 :
4638 126 : if (!has_negative && !has_positive)
4639 : {
4640 12 : sprintf(cp, "0");
4641 : }
4642 114 : else if (!sql_standard_value)
4643 : {
4644 : /*
4645 : * For non sql-standard interval values, force outputting
4646 : * the signs to avoid ambiguities with intervals with
4647 : * mixed sign components.
4648 : */
4649 54 : char year_sign = (year < 0 || mon < 0) ? '-' : '+';
4650 54 : char day_sign = (mday < 0) ? '-' : '+';
4651 78 : char sec_sign = (hour < 0 || min < 0 ||
4652 24 : sec < 0 || fsec < 0) ? '-' : '+';
4653 :
4654 54 : sprintf(cp, "%c%d-%d %c%lld %c%lld:%02d:",
4655 : year_sign, abs(year), abs(mon),
4656 54 : day_sign, (long long) i64abs(mday),
4657 54 : sec_sign, (long long) i64abs(hour), abs(min));
4658 54 : cp += strlen(cp);
4659 54 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4660 54 : *cp = '\0';
4661 : }
4662 60 : else if (has_year_month)
4663 : {
4664 18 : sprintf(cp, "%d-%d", year, mon);
4665 : }
4666 42 : else if (has_day)
4667 : {
4668 30 : sprintf(cp, "%lld %lld:%02d:",
4669 : (long long) mday, (long long) hour, min);
4670 30 : cp += strlen(cp);
4671 30 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4672 30 : *cp = '\0';
4673 : }
4674 : else
4675 : {
4676 12 : sprintf(cp, "%lld:%02d:", (long long) hour, min);
4677 12 : cp += strlen(cp);
4678 12 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4679 12 : *cp = '\0';
4680 : }
4681 : }
4682 126 : break;
4683 :
4684 : /* ISO 8601 "time-intervals by duration only" */
4685 48 : case INTSTYLE_ISO_8601:
4686 : /* special-case zero to avoid printing nothing */
4687 48 : if (year == 0 && mon == 0 && mday == 0 &&
4688 6 : hour == 0 && min == 0 && sec == 0 && fsec == 0)
4689 : {
4690 6 : sprintf(cp, "PT0S");
4691 6 : break;
4692 : }
4693 42 : *cp++ = 'P';
4694 42 : cp = AddISO8601IntPart(cp, year, 'Y');
4695 42 : cp = AddISO8601IntPart(cp, mon, 'M');
4696 42 : cp = AddISO8601IntPart(cp, mday, 'D');
4697 42 : if (hour != 0 || min != 0 || sec != 0 || fsec != 0)
4698 36 : *cp++ = 'T';
4699 42 : cp = AddISO8601IntPart(cp, hour, 'H');
4700 42 : cp = AddISO8601IntPart(cp, min, 'M');
4701 42 : if (sec != 0 || fsec != 0)
4702 : {
4703 36 : if (sec < 0 || fsec < 0)
4704 12 : *cp++ = '-';
4705 36 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4706 36 : *cp++ = 'S';
4707 36 : *cp++ = '\0';
4708 : }
4709 42 : break;
4710 :
4711 : /* Compatible with postgresql < 8.4 when DateStyle = 'iso' */
4712 4812 : case INTSTYLE_POSTGRES:
4713 4812 : cp = AddPostgresIntPart(cp, year, "year", &is_zero, &is_before);
4714 :
4715 : /*
4716 : * Ideally we should spell out "month" like we do for "year" and
4717 : * "day". However, for backward compatibility, we can't easily
4718 : * fix this. bjm 2011-05-24
4719 : */
4720 4812 : cp = AddPostgresIntPart(cp, mon, "mon", &is_zero, &is_before);
4721 4812 : cp = AddPostgresIntPart(cp, mday, "day", &is_zero, &is_before);
4722 4812 : if (is_zero || hour != 0 || min != 0 || sec != 0 || fsec != 0)
4723 : {
4724 3868 : bool minus = (hour < 0 || min < 0 || sec < 0 || fsec < 0);
4725 :
4726 7736 : sprintf(cp, "%s%s%02lld:%02d:",
4727 3868 : is_zero ? "" : " ",
4728 3654 : (minus ? "-" : (is_before ? "+" : "")),
4729 3868 : (long long) i64abs(hour), abs(min));
4730 3868 : cp += strlen(cp);
4731 3868 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4732 3868 : *cp = '\0';
4733 : }
4734 4812 : break;
4735 :
4736 : /* Compatible with postgresql < 8.4 when DateStyle != 'iso' */
4737 8498 : case INTSTYLE_POSTGRES_VERBOSE:
4738 : default:
4739 8498 : strcpy(cp, "@");
4740 8498 : cp++;
4741 8498 : cp = AddVerboseIntPart(cp, year, "year", &is_zero, &is_before);
4742 8498 : cp = AddVerboseIntPart(cp, mon, "mon", &is_zero, &is_before);
4743 8498 : cp = AddVerboseIntPart(cp, mday, "day", &is_zero, &is_before);
4744 8498 : cp = AddVerboseIntPart(cp, hour, "hour", &is_zero, &is_before);
4745 8498 : cp = AddVerboseIntPart(cp, min, "min", &is_zero, &is_before);
4746 8498 : if (sec != 0 || fsec != 0)
4747 : {
4748 3276 : *cp++ = ' ';
4749 3276 : if (sec < 0 || (sec == 0 && fsec < 0))
4750 : {
4751 1014 : if (is_zero)
4752 342 : is_before = true;
4753 672 : else if (!is_before)
4754 6 : *cp++ = '-';
4755 : }
4756 2262 : else if (is_before)
4757 12 : *cp++ = '-';
4758 3276 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4759 : /* We output "ago", not negatives, so use abs(). */
4760 3276 : sprintf(cp, " sec%s",
4761 3276 : (abs(sec) != 1 || fsec != 0) ? "s" : "");
4762 3276 : is_zero = false;
4763 : }
4764 : /* identically zero? then put in a unitless zero... */
4765 8498 : if (is_zero)
4766 230 : strcat(cp, " 0");
4767 8498 : if (is_before)
4768 1408 : strcat(cp, " ago");
4769 8498 : break;
4770 : }
4771 13484 : }
4772 :
4773 :
4774 : /*
4775 : * We've been burnt by stupid errors in the ordering of the datetkn tables
4776 : * once too often. Arrange to check them during postmaster start.
4777 : */
4778 : static bool
4779 2980 : CheckDateTokenTable(const char *tablename, const datetkn *base, int nel)
4780 : {
4781 2980 : bool ok = true;
4782 : int i;
4783 :
4784 201150 : for (i = 0; i < nel; i++)
4785 : {
4786 : /* check for token strings that don't fit */
4787 198170 : if (strlen(base[i].token) > TOKMAXLEN)
4788 : {
4789 : /* %.*s is safe since all our tokens are ASCII */
4790 0 : elog(LOG, "token too long in %s table: \"%.*s\"",
4791 : tablename,
4792 : TOKMAXLEN + 1, base[i].token);
4793 0 : ok = false;
4794 0 : break; /* don't risk applying strcmp */
4795 : }
4796 : /* check for out of order */
4797 198170 : if (i > 0 &&
4798 195190 : strcmp(base[i - 1].token, base[i].token) >= 0)
4799 : {
4800 0 : elog(LOG, "ordering error in %s table: \"%s\" >= \"%s\"",
4801 : tablename,
4802 : base[i - 1].token,
4803 : base[i].token);
4804 0 : ok = false;
4805 : }
4806 : }
4807 2980 : return ok;
4808 : }
4809 :
4810 : bool
4811 1490 : CheckDateTokenTables(void)
4812 : {
4813 1490 : bool ok = true;
4814 :
4815 : Assert(UNIX_EPOCH_JDATE == date2j(1970, 1, 1));
4816 : Assert(POSTGRES_EPOCH_JDATE == date2j(2000, 1, 1));
4817 :
4818 1490 : ok &= CheckDateTokenTable("datetktbl", datetktbl, szdatetktbl);
4819 1490 : ok &= CheckDateTokenTable("deltatktbl", deltatktbl, szdeltatktbl);
4820 1490 : return ok;
4821 : }
4822 :
4823 : /*
4824 : * Common code for temporal prosupport functions: simplify, if possible,
4825 : * a call to a temporal type's length-coercion function.
4826 : *
4827 : * Types time, timetz, timestamp and timestamptz each have a range of allowed
4828 : * precisions. An unspecified precision is rigorously equivalent to the
4829 : * highest specifiable precision. We can replace the function call with a
4830 : * no-op RelabelType if it is coercing to the same or higher precision as the
4831 : * input is known to have.
4832 : *
4833 : * The input Node is always a FuncExpr, but to reduce the #include footprint
4834 : * of datetime.h, we declare it as Node *.
4835 : *
4836 : * Note: timestamp_scale throws an error when the typmod is out of range, but
4837 : * we can't get there from a cast: our typmodin will have caught it already.
4838 : */
4839 : Node *
4840 24 : TemporalSimplify(int32 max_precis, Node *node)
4841 : {
4842 24 : FuncExpr *expr = castNode(FuncExpr, node);
4843 24 : Node *ret = NULL;
4844 : Node *typmod;
4845 :
4846 : Assert(list_length(expr->args) >= 2);
4847 :
4848 24 : typmod = (Node *) lsecond(expr->args);
4849 :
4850 24 : if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
4851 : {
4852 24 : Node *source = (Node *) linitial(expr->args);
4853 24 : int32 old_precis = exprTypmod(source);
4854 24 : int32 new_precis = DatumGetInt32(((Const *) typmod)->constvalue);
4855 :
4856 24 : if (new_precis < 0 || new_precis == max_precis ||
4857 0 : (old_precis >= 0 && new_precis >= old_precis))
4858 0 : ret = relabel_to_typmod(source, new_precis);
4859 : }
4860 :
4861 24 : return ret;
4862 : }
4863 :
4864 : /*
4865 : * This function gets called during timezone config file load or reload
4866 : * to create the final array of timezone tokens. The argument array
4867 : * is already sorted in name order.
4868 : *
4869 : * The result is a TimeZoneAbbrevTable (which must be a single guc_malloc'd
4870 : * chunk) or NULL on alloc failure. No other error conditions are defined.
4871 : */
4872 : TimeZoneAbbrevTable *
4873 12052 : ConvertTimeZoneAbbrevs(struct tzEntry *abbrevs, int n)
4874 : {
4875 : TimeZoneAbbrevTable *tbl;
4876 : Size tbl_size;
4877 : int i;
4878 :
4879 : /* Space for fixed fields and datetkn array */
4880 12052 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
4881 12052 : n * sizeof(datetkn);
4882 12052 : tbl_size = MAXALIGN(tbl_size);
4883 : /* Count up space for dynamic abbreviations */
4884 2362204 : for (i = 0; i < n; i++)
4885 : {
4886 2350152 : struct tzEntry *abbr = abbrevs + i;
4887 :
4888 2350152 : if (abbr->zone != NULL)
4889 : {
4890 : Size dsize;
4891 :
4892 602594 : dsize = offsetof(DynamicZoneAbbrev, zone) +
4893 602594 : strlen(abbr->zone) + 1;
4894 602594 : tbl_size += MAXALIGN(dsize);
4895 : }
4896 : }
4897 :
4898 : /* Alloc the result ... */
4899 12052 : tbl = guc_malloc(LOG, tbl_size);
4900 12052 : if (!tbl)
4901 0 : return NULL;
4902 :
4903 : /* ... and fill it in */
4904 12052 : tbl->tblsize = tbl_size;
4905 12052 : tbl->numabbrevs = n;
4906 : /* in this loop, tbl_size reprises the space calculation above */
4907 12052 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
4908 12052 : n * sizeof(datetkn);
4909 12052 : tbl_size = MAXALIGN(tbl_size);
4910 2362204 : for (i = 0; i < n; i++)
4911 : {
4912 2350152 : struct tzEntry *abbr = abbrevs + i;
4913 2350152 : datetkn *dtoken = tbl->abbrevs + i;
4914 :
4915 : /* use strlcpy to truncate name if necessary */
4916 2350152 : strlcpy(dtoken->token, abbr->abbrev, TOKMAXLEN + 1);
4917 2350152 : if (abbr->zone != NULL)
4918 : {
4919 : /* Allocate a DynamicZoneAbbrev for this abbreviation */
4920 : DynamicZoneAbbrev *dtza;
4921 : Size dsize;
4922 :
4923 602594 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tbl_size);
4924 602594 : dtza->tz = NULL;
4925 602594 : strcpy(dtza->zone, abbr->zone);
4926 :
4927 602594 : dtoken->type = DYNTZ;
4928 : /* value is offset from table start to DynamicZoneAbbrev */
4929 602594 : dtoken->value = (int32) tbl_size;
4930 :
4931 602594 : dsize = offsetof(DynamicZoneAbbrev, zone) +
4932 602594 : strlen(abbr->zone) + 1;
4933 602594 : tbl_size += MAXALIGN(dsize);
4934 : }
4935 : else
4936 : {
4937 1747558 : dtoken->type = abbr->is_dst ? DTZ : TZ;
4938 1747558 : dtoken->value = abbr->offset;
4939 : }
4940 : }
4941 :
4942 : /* Assert the two loops above agreed on size calculations */
4943 : Assert(tbl->tblsize == tbl_size);
4944 :
4945 : /* Check the ordering, if testing */
4946 : Assert(CheckDateTokenTable("timezone abbreviations", tbl->abbrevs, n));
4947 :
4948 12052 : return tbl;
4949 : }
4950 :
4951 : /*
4952 : * Install a TimeZoneAbbrevTable as the active table.
4953 : *
4954 : * Caller is responsible that the passed table doesn't go away while in use.
4955 : */
4956 : void
4957 11872 : InstallTimeZoneAbbrevs(TimeZoneAbbrevTable *tbl)
4958 : {
4959 11872 : zoneabbrevtbl = tbl;
4960 : /* reset abbrevcache, which may contain pointers into old table */
4961 11872 : memset(abbrevcache, 0, sizeof(abbrevcache));
4962 11872 : }
4963 :
4964 : /*
4965 : * Helper subroutine to locate pg_tz timezone for a dynamic abbreviation.
4966 : *
4967 : * On failure, returns NULL and fills *extra for a DTERR_BAD_ZONE_ABBREV error.
4968 : */
4969 : static pg_tz *
4970 1164 : FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp,
4971 : DateTimeErrorExtra *extra)
4972 : {
4973 : DynamicZoneAbbrev *dtza;
4974 :
4975 : /* Just some sanity checks to prevent indexing off into nowhere */
4976 : Assert(tp->type == DYNTZ);
4977 : Assert(tp->value > 0 && tp->value < tbl->tblsize);
4978 :
4979 1164 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tp->value);
4980 :
4981 : /* Look up the underlying zone if we haven't already */
4982 1164 : if (dtza->tz == NULL)
4983 : {
4984 918 : dtza->tz = pg_tzset(dtza->zone);
4985 918 : if (dtza->tz == NULL)
4986 : {
4987 : /* Ooops, bogus zone name in config file entry */
4988 0 : extra->dtee_timezone = dtza->zone;
4989 0 : extra->dtee_abbrev = tp->token;
4990 : }
4991 : }
4992 1164 : return dtza->tz;
4993 : }
4994 :
4995 :
4996 : /*
4997 : * This set-returning function reads all the available time zone abbreviations
4998 : * and returns a set of (abbrev, utc_offset, is_dst).
4999 : */
5000 : Datum
5001 3540 : pg_timezone_abbrevs(PG_FUNCTION_ARGS)
5002 : {
5003 : FuncCallContext *funcctx;
5004 : int *pindex;
5005 : Datum result;
5006 : HeapTuple tuple;
5007 : Datum values[3];
5008 3540 : bool nulls[3] = {0};
5009 : const datetkn *tp;
5010 : char buffer[TOKMAXLEN + 1];
5011 : int gmtoffset;
5012 : bool is_dst;
5013 : unsigned char *p;
5014 : struct pg_itm_in itm_in;
5015 : Interval *resInterval;
5016 :
5017 : /* stuff done only on the first call of the function */
5018 3540 : if (SRF_IS_FIRSTCALL())
5019 : {
5020 : TupleDesc tupdesc;
5021 : MemoryContext oldcontext;
5022 :
5023 : /* create a function context for cross-call persistence */
5024 18 : funcctx = SRF_FIRSTCALL_INIT();
5025 :
5026 : /*
5027 : * switch to memory context appropriate for multiple function calls
5028 : */
5029 18 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
5030 :
5031 : /* allocate memory for user context */
5032 18 : pindex = (int *) palloc(sizeof(int));
5033 18 : *pindex = 0;
5034 18 : funcctx->user_fctx = (void *) pindex;
5035 :
5036 18 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
5037 0 : elog(ERROR, "return type must be a row type");
5038 18 : funcctx->tuple_desc = tupdesc;
5039 :
5040 18 : MemoryContextSwitchTo(oldcontext);
5041 : }
5042 :
5043 : /* stuff done on every call of the function */
5044 3540 : funcctx = SRF_PERCALL_SETUP();
5045 3540 : pindex = (int *) funcctx->user_fctx;
5046 :
5047 3540 : if (zoneabbrevtbl == NULL ||
5048 3540 : *pindex >= zoneabbrevtbl->numabbrevs)
5049 18 : SRF_RETURN_DONE(funcctx);
5050 :
5051 3522 : tp = zoneabbrevtbl->abbrevs + *pindex;
5052 :
5053 3522 : switch (tp->type)
5054 : {
5055 1764 : case TZ:
5056 1764 : gmtoffset = tp->value;
5057 1764 : is_dst = false;
5058 1764 : break;
5059 864 : case DTZ:
5060 864 : gmtoffset = tp->value;
5061 864 : is_dst = true;
5062 864 : break;
5063 894 : case DYNTZ:
5064 : {
5065 : /* Determine the current meaning of the abbrev */
5066 : pg_tz *tzp;
5067 : DateTimeErrorExtra extra;
5068 : TimestampTz now;
5069 : int isdst;
5070 :
5071 894 : tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp, &extra);
5072 894 : if (tzp == NULL)
5073 0 : DateTimeParseError(DTERR_BAD_ZONE_ABBREV, &extra,
5074 : NULL, NULL, NULL);
5075 894 : now = GetCurrentTransactionStartTimestamp();
5076 1788 : gmtoffset = -DetermineTimeZoneAbbrevOffsetTS(now,
5077 894 : tp->token,
5078 : tzp,
5079 : &isdst);
5080 894 : is_dst = (bool) isdst;
5081 894 : break;
5082 : }
5083 0 : default:
5084 0 : elog(ERROR, "unrecognized timezone type %d", (int) tp->type);
5085 : gmtoffset = 0; /* keep compiler quiet */
5086 : is_dst = false;
5087 : break;
5088 : }
5089 :
5090 : /*
5091 : * Convert name to text, using upcasing conversion that is the inverse of
5092 : * what ParseDateTime() uses.
5093 : */
5094 3522 : strlcpy(buffer, tp->token, sizeof(buffer));
5095 16320 : for (p = (unsigned char *) buffer; *p; p++)
5096 12798 : *p = pg_toupper(*p);
5097 :
5098 3522 : values[0] = CStringGetTextDatum(buffer);
5099 :
5100 : /* Convert offset (in seconds) to an interval; can't overflow */
5101 14088 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5102 3522 : itm_in.tm_usec = (int64) gmtoffset * USECS_PER_SEC;
5103 3522 : resInterval = (Interval *) palloc(sizeof(Interval));
5104 3522 : (void) itmin2interval(&itm_in, resInterval);
5105 3522 : values[1] = IntervalPGetDatum(resInterval);
5106 :
5107 3522 : values[2] = BoolGetDatum(is_dst);
5108 :
5109 3522 : (*pindex)++;
5110 :
5111 3522 : tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
5112 3522 : result = HeapTupleGetDatum(tuple);
5113 :
5114 3522 : SRF_RETURN_NEXT(funcctx, result);
5115 : }
5116 :
5117 : /*
5118 : * This set-returning function reads all the available full time zones
5119 : * and returns a set of (name, abbrev, utc_offset, is_dst).
5120 : */
5121 : Datum
5122 16 : pg_timezone_names(PG_FUNCTION_ARGS)
5123 : {
5124 16 : ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
5125 : pg_tzenum *tzenum;
5126 : pg_tz *tz;
5127 : Datum values[4];
5128 16 : bool nulls[4] = {0};
5129 : int tzoff;
5130 : struct pg_tm tm;
5131 : fsec_t fsec;
5132 : const char *tzn;
5133 : Interval *resInterval;
5134 : struct pg_itm_in itm_in;
5135 :
5136 16 : InitMaterializedSRF(fcinfo, 0);
5137 :
5138 : /* initialize timezone scanning code */
5139 16 : tzenum = pg_tzenumerate_start();
5140 :
5141 : /* search for another zone to display */
5142 : for (;;)
5143 : {
5144 9568 : tz = pg_tzenumerate_next(tzenum);
5145 9568 : if (!tz)
5146 16 : break;
5147 :
5148 : /* Convert now() to local time in this zone */
5149 9552 : if (timestamp2tm(GetCurrentTransactionStartTimestamp(),
5150 : &tzoff, &tm, &fsec, &tzn, tz) != 0)
5151 0 : continue; /* ignore if conversion fails */
5152 :
5153 : /*
5154 : * IANA's rather silly "Factory" time zone used to emit ridiculously
5155 : * long "abbreviations" such as "Local time zone must be set--see zic
5156 : * manual page" or "Local time zone must be set--use tzsetup". While
5157 : * modern versions of tzdb emit the much saner "-00", it seems some
5158 : * benighted packagers are hacking the IANA data so that it continues
5159 : * to produce these strings. To prevent producing a weirdly wide
5160 : * abbrev column, reject ridiculously long abbreviations.
5161 : */
5162 9552 : if (tzn && strlen(tzn) > 31)
5163 0 : continue;
5164 :
5165 9552 : values[0] = CStringGetTextDatum(pg_get_timezone_name(tz));
5166 9552 : values[1] = CStringGetTextDatum(tzn ? tzn : "");
5167 :
5168 : /* Convert tzoff to an interval; can't overflow */
5169 38208 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5170 9552 : itm_in.tm_usec = (int64) -tzoff * USECS_PER_SEC;
5171 9552 : resInterval = (Interval *) palloc(sizeof(Interval));
5172 9552 : (void) itmin2interval(&itm_in, resInterval);
5173 9552 : values[2] = IntervalPGetDatum(resInterval);
5174 :
5175 9552 : values[3] = BoolGetDatum(tm.tm_isdst > 0);
5176 :
5177 9552 : tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls);
5178 : }
5179 :
5180 16 : pg_tzenumerate_end(tzenum);
5181 16 : return (Datum) 0;
5182 : }
|