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 341854 : date2j(int year, int month, int day)
287 : {
288 : int julian;
289 : int century;
290 :
291 341854 : if (month > 2)
292 : {
293 187660 : month += 1;
294 187660 : year += 4800;
295 : }
296 : else
297 : {
298 154194 : month += 13;
299 154194 : year += 4799;
300 : }
301 :
302 341854 : century = year / 100;
303 341854 : julian = year * 365 - 32167;
304 341854 : julian += year / 4 - century + century / 4;
305 341854 : julian += 7834 * month / 256 + day;
306 :
307 341854 : return julian;
308 : } /* date2j() */
309 :
310 : void
311 273676 : 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 273676 : julian = jd;
319 273676 : julian += 32044;
320 273676 : quad = julian / 146097;
321 273676 : extra = (julian - quad * 146097) * 4 + 3;
322 273676 : julian += 60 + quad * 3 + extra / 146097;
323 273676 : quad = julian / 1461;
324 273676 : julian -= quad * 1461;
325 273676 : y = julian * 4 / 1461;
326 547352 : julian = ((y != 0) ? ((julian + 305) % 365) : ((julian + 306) % 366))
327 273676 : + 123;
328 273676 : y += quad * 4;
329 273676 : *year = y - 4800;
330 273676 : quad = julian * 2141 / 65536;
331 273676 : *day = julian - 7834 * quad / 256;
332 273676 : *month = (quad + 10) % MONTHS_PER_YEAR + 1;
333 273676 : } /* 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 51558 : j2day(int date)
345 : {
346 51558 : date += 1;
347 51558 : date %= 7;
348 : /* Cope if division truncates towards zero, as it probably does */
349 51558 : if (date < 0)
350 0 : date += 7;
351 :
352 51558 : 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 2732 : GetCurrentDateTime(struct pg_tm *tm)
367 : {
368 : fsec_t fsec;
369 :
370 2732 : GetCurrentTimeUsec(tm, &fsec, NULL);
371 2732 : }
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 2906 : GetCurrentTimeUsec(struct pg_tm *tm, fsec_t *fsec, int *tzp)
388 : {
389 2906 : 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 2906 : 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 : */
411 886 : 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 : */
418 886 : 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. */
425 886 : cache_ts = cur_ts;
426 886 : cache_timezone = session_timezone;
427 : }
428 :
429 2906 : *tm = cache_tm;
430 2906 : *fsec = cache_fsec;
431 2906 : if (tzp != NULL)
432 162 : *tzp = cache_tz;
433 2906 : }
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 133250 : AppendSeconds(char *cp, int sec, fsec_t fsec, int precision, bool fillzeros)
449 : {
450 : Assert(precision >= 0);
451 :
452 133250 : if (fillzeros)
453 129950 : cp = pg_ultostr_zeropad(cp, abs(sec), 2);
454 : else
455 3300 : cp = pg_ultostr(cp, abs(sec));
456 :
457 : /* fsec_t is just an int32 */
458 133250 : if (fsec != 0)
459 : {
460 21776 : int32 value = abs(fsec);
461 21776 : char *end = &cp[precision + 1];
462 21776 : bool gotnonzero = false;
463 :
464 21776 : *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 152432 : while (precision--)
472 : {
473 130656 : int32 oldval = value;
474 : int32 remainder;
475 :
476 130656 : value /= 10;
477 130656 : remainder = oldval - value * 10;
478 :
479 : /* check if we got a non-zero */
480 130656 : if (remainder)
481 96958 : gotnonzero = true;
482 :
483 130656 : if (gotnonzero)
484 113778 : cp[precision] = '0' + remainder;
485 : else
486 16878 : 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 21776 : if (value)
495 0 : return pg_ultostr(cp, abs(fsec));
496 :
497 21776 : return end;
498 : }
499 : else
500 111474 : 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 113548 : AppendTimestampSeconds(char *cp, struct pg_tm *tm, fsec_t fsec)
512 : {
513 113548 : 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 8762 : int64_multiply_add(int64 val, int64 multiplier, int64 *sum)
523 : {
524 : int64 product;
525 :
526 17464 : if (pg_mul_s64_overflow(val, multiplier, &product) ||
527 8702 : pg_add_s64_overflow(*sum, product, sum))
528 156 : return false;
529 8606 : 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 9856 : AdjustFractMicroseconds(double frac, int64 scale,
538 : struct pg_itm_in *itm_in)
539 : {
540 : int64 usec;
541 :
542 : /* Fast path for common case */
543 9856 : if (frac == 0)
544 9394 : 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 2792 : AdjustMicroseconds(int64 val, double fval, int64 scale,
619 : struct pg_itm_in *itm_in)
620 : {
621 : /* Handle the integer part */
622 2792 : if (!int64_multiply_add(val, scale, &itm_in->tm_usec))
623 150 : return false;
624 : /* Handle the float part */
625 2642 : 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 7334 : AdjustDays(int64 val, int scale, struct pg_itm_in *itm_in)
634 : {
635 : int days;
636 :
637 7334 : if (val < INT_MIN || val > INT_MAX)
638 36 : return false;
639 14584 : return !pg_mul_s32_overflow((int32) val, scale, &days) &&
640 7286 : !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 23178 : 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 23178 : if (cp[1] == '\0')
690 : {
691 0 : *frac = 0;
692 : }
693 : else
694 : {
695 23178 : errno = 0;
696 23178 : *frac = strtod(cp, &cp);
697 : /* check for parse failure */
698 23178 : if (*cp != '\0' || errno != 0)
699 12 : return DTERR_BAD_FORMAT;
700 : }
701 23166 : 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 22680 : ParseFractionalSecond(char *cp, fsec_t *fsec)
710 : {
711 : double frac;
712 : int dterr;
713 :
714 22680 : dterr = ParseFraction(cp, &frac);
715 22680 : if (dterr)
716 12 : return dterr;
717 22668 : *fsec = rint(frac * 1000000);
718 22668 : 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 85296 : ParseDateTime(const char *timestr, char *workbuf, size_t buflen,
755 : char **field, int *ftype, int maxfields, int *numfields)
756 : {
757 85296 : int nf = 0;
758 85296 : const char *cp = timestr;
759 85296 : char *bufp = workbuf;
760 85296 : 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 375042 : while (*cp != '\0')
778 : {
779 : /* Ignore spaces between fields */
780 289746 : if (isspace((unsigned char) *cp))
781 : {
782 83316 : cp++;
783 83316 : continue;
784 : }
785 :
786 : /* Record start of current field */
787 206430 : if (nf >= maxfields)
788 0 : return DTERR_BAD_FORMAT;
789 206430 : field[nf] = bufp;
790 :
791 : /* leading digit? then date or time */
792 206430 : if (isdigit((unsigned char) *cp))
793 : {
794 140874 : APPEND_CHAR(bufp, bufend, *cp++);
795 420796 : while (isdigit((unsigned char) *cp))
796 279922 : APPEND_CHAR(bufp, bufend, *cp++);
797 :
798 : /* time field? */
799 140874 : if (*cp == ':')
800 : {
801 59920 : ftype[nf] = DTK_TIME;
802 59920 : APPEND_CHAR(bufp, bufend, *cp++);
803 511644 : while (isdigit((unsigned char) *cp) ||
804 140864 : (*cp == ':') || (*cp == '.'))
805 451724 : APPEND_CHAR(bufp, bufend, *cp++);
806 : }
807 : /* date field? allow embedded text month */
808 80954 : else if (*cp == '-' || *cp == '/' || *cp == '.')
809 65366 : {
810 : /* save delimiting character to use later */
811 65366 : char delim = *cp;
812 :
813 65366 : APPEND_CHAR(bufp, bufend, *cp++);
814 : /* second field is all digits? then no embedded text month */
815 65366 : if (isdigit((unsigned char) *cp))
816 : {
817 65270 : ftype[nf] = ((delim == '.') ? DTK_NUMBER : DTK_DATE);
818 195846 : while (isdigit((unsigned char) *cp))
819 130576 : APPEND_CHAR(bufp, bufend, *cp++);
820 :
821 : /*
822 : * insist that the delimiters match to get a three-field
823 : * date.
824 : */
825 65270 : if (*cp == delim)
826 : {
827 64688 : ftype[nf] = DTK_DATE;
828 64688 : APPEND_CHAR(bufp, bufend, *cp++);
829 196852 : while (isdigit((unsigned char) *cp) || *cp == delim)
830 132164 : 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 15588 : ftype[nf] = DTK_NUMBER;
847 : }
848 : /* Leading decimal point? Then fractional seconds... */
849 65556 : 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 65556 : else if (isalpha((unsigned char) *cp))
862 : {
863 : bool is_date;
864 :
865 24158 : ftype[nf] = DTK_STRING;
866 24158 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
867 96842 : while (isalpha((unsigned char) *cp))
868 72684 : 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 24158 : is_date = false;
879 24158 : if (*cp == '-' || *cp == '/' || *cp == '.')
880 1510 : is_date = true;
881 22648 : else if (*cp == '+' || isdigit((unsigned char) *cp))
882 : {
883 1818 : *bufp = '\0'; /* null-terminate current field value */
884 : /* we need search only the core token table, not TZ names */
885 1818 : if (datebsearch(field[nf], datetktbl, szdatetktbl) == NULL)
886 1512 : is_date = true;
887 : }
888 24158 : 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 41398 : else if (*cp == '+' || *cp == '-')
902 : {
903 40834 : APPEND_CHAR(bufp, bufend, *cp++);
904 : /* soak up leading whitespace */
905 40858 : 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 40834 : if (isdigit((unsigned char) *cp))
910 : {
911 39974 : ftype[nf] = DTK_TZ;
912 39974 : APPEND_CHAR(bufp, bufend, *cp++);
913 93736 : while (isdigit((unsigned char) *cp) ||
914 41672 : *cp == ':' || *cp == '.' || *cp == '-')
915 53762 : APPEND_CHAR(bufp, bufend, *cp++);
916 : }
917 : /* special? */
918 860 : else if (isalpha((unsigned char) *cp))
919 : {
920 860 : ftype[nf] = DTK_SPECIAL;
921 860 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
922 6880 : while (isalpha((unsigned char) *cp))
923 6020 : 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 205866 : *bufp++ = '\0';
941 205866 : nf++;
942 : }
943 :
944 85296 : *numfields = nf;
945 :
946 85296 : 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 69422 : DecodeDateTime(char **field, int *ftype, int nf,
979 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
980 : DateTimeErrorExtra *extra)
981 : {
982 69422 : int fmask = 0,
983 : tmask,
984 : type;
985 69422 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
986 : int i;
987 : int val;
988 : int dterr;
989 69422 : int mer = HR24;
990 69422 : bool haveTextMonth = false;
991 69422 : bool isjulian = false;
992 69422 : bool is2digits = false;
993 69422 : bool bc = false;
994 69422 : pg_tz *namedTz = NULL;
995 69422 : pg_tz *abbrevTz = NULL;
996 : pg_tz *valtz;
997 69422 : 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 69422 : *dtype = DTK_DATE;
1005 69422 : tm->tm_hour = 0;
1006 69422 : tm->tm_min = 0;
1007 69422 : tm->tm_sec = 0;
1008 69422 : *fsec = 0;
1009 : /* don't know daylight savings time status apriori */
1010 69422 : tm->tm_isdst = -1;
1011 69422 : if (tzp != NULL)
1012 69422 : *tzp = 0;
1013 :
1014 242008 : for (i = 0; i < nf; i++)
1015 : {
1016 172898 : switch (ftype[i])
1017 : {
1018 66044 : 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 66044 : 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 66038 : else if (ptype != 0 ||
1062 66026 : ((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 64622 : dterr = DecodeDate(field[i], fmask,
1133 : &tmask, &is2digits, tm);
1134 64622 : if (dterr)
1135 36 : return dterr;
1136 : }
1137 65966 : break;
1138 :
1139 54662 : case DTK_TIME:
1140 :
1141 : /*
1142 : * This might be an ISO time following a "t" field.
1143 : */
1144 54662 : 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 54662 : dterr = DecodeTime(field[i], fmask, INTERVAL_FULL_RANGE,
1152 : &tmask, tm, fsec);
1153 54662 : if (dterr)
1154 0 : return dterr;
1155 :
1156 : /* check for time overflow */
1157 54662 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec,
1158 : *fsec))
1159 0 : return DTERR_FIELD_OVERFLOW;
1160 54662 : 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 9230 : case DTK_STRING:
1303 : case DTK_SPECIAL:
1304 : /* timezone abbrevs take precedence over built-in tokens */
1305 9230 : dterr = DecodeTimezoneAbbrev(i, field[i],
1306 : &type, &val, &valtz, extra);
1307 9230 : if (dterr)
1308 0 : return dterr;
1309 9230 : if (type == UNKNOWN_FIELD)
1310 6826 : type = DecodeSpecial(i, field[i], &val);
1311 9230 : if (type == IGNORE_DTF)
1312 0 : continue;
1313 :
1314 9230 : tmask = DTK_M(type);
1315 9230 : switch (type)
1316 : {
1317 1654 : case RESERV:
1318 1654 : 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 1138 : case DTK_EPOCH:
1362 : case DTK_LATE:
1363 : case DTK_EARLY:
1364 1138 : tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1365 1138 : *dtype = val;
1366 : /* caller ignores tm for these dtype codes */
1367 1138 : break;
1368 :
1369 0 : default:
1370 0 : elog(ERROR, "unrecognized RESERV datetime token: %d",
1371 : val);
1372 : }
1373 :
1374 1654 : 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 1778 : case DTZ:
1407 :
1408 : /*
1409 : * set mask for TZ here _or_ check for DTZ later when
1410 : * getting default timezone
1411 : */
1412 1778 : tmask |= DTK_M(TZ);
1413 1778 : tm->tm_isdst = 1;
1414 1778 : if (tzp == NULL)
1415 0 : return DTERR_BAD_FORMAT;
1416 1778 : *tzp = -val;
1417 1778 : break;
1418 :
1419 542 : case TZ:
1420 542 : tm->tm_isdst = 0;
1421 542 : if (tzp == NULL)
1422 0 : return DTERR_BAD_FORMAT;
1423 542 : *tzp = -val;
1424 542 : 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 9170 : break;
1490 :
1491 0 : default:
1492 0 : return DTERR_BAD_FORMAT;
1493 : }
1494 :
1495 172730 : if (tmask & fmask)
1496 144 : return DTERR_BAD_FORMAT;
1497 172586 : fmask |= tmask;
1498 : } /* end loop over fields */
1499 :
1500 : /* reject if prefix type appeared and was never handled */
1501 69110 : if (ptype != 0)
1502 0 : return DTERR_BAD_FORMAT;
1503 :
1504 : /* do additional checking for normal date specs (but not "infinity" etc) */
1505 69110 : if (*dtype == DTK_DATE)
1506 : {
1507 : /* do final checking/adjustment of Y/M/D fields */
1508 68116 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
1509 68116 : if (dterr)
1510 198 : return dterr;
1511 :
1512 : /* handle AM/PM */
1513 67918 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
1514 0 : return DTERR_FIELD_OVERFLOW;
1515 67918 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
1516 0 : tm->tm_hour = 0;
1517 67918 : 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 67918 : 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 67912 : 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 67912 : 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 67912 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
1556 : {
1557 : /*
1558 : * daylight savings time modifier but no standard timezone? then
1559 : * error
1560 : */
1561 27356 : if (fmask & DTK_M(DTZMOD))
1562 0 : return DTERR_BAD_FORMAT;
1563 :
1564 27356 : *tzp = DetermineTimeZoneOffset(tm, session_timezone);
1565 : }
1566 : }
1567 :
1568 68906 : 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 55908 : DetermineTimeZoneOffset(struct pg_tm *tm, pg_tz *tzp)
1586 : {
1587 : pg_time_t t;
1588 :
1589 55908 : 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 56094 : 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 56094 : if (!IS_VALID_JULIAN(tm->tm_year, tm->tm_mon, tm->tm_mday))
1630 24 : goto overflow;
1631 56070 : date = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday) - UNIX_EPOCH_JDATE;
1632 :
1633 56070 : day = ((pg_time_t) date) * SECS_PER_DAY;
1634 56070 : if (day / SECS_PER_DAY != date)
1635 0 : goto overflow;
1636 56070 : sec = tm->tm_sec + (tm->tm_min + tm->tm_hour * MINS_PER_HOUR) * SECS_PER_MINUTE;
1637 56070 : mytime = day + sec;
1638 : /* since sec >= 0, overflow could only be from +day to -mytime */
1639 56070 : 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 56070 : prevtime = mytime - SECS_PER_DAY;
1649 56070 : if (mytime < 0 && prevtime > 0)
1650 0 : goto overflow;
1651 :
1652 56070 : res = pg_next_dst_boundary(&prevtime,
1653 : &before_gmtoff, &before_isdst,
1654 : &boundary,
1655 : &after_gmtoff, &after_isdst,
1656 : tzp);
1657 56070 : if (res < 0)
1658 0 : goto overflow; /* failure? */
1659 :
1660 56070 : if (res == 0)
1661 : {
1662 : /* Non-DST zone, life is simple */
1663 3264 : tm->tm_isdst = before_isdst;
1664 3264 : *tp = mytime - before_gmtoff;
1665 3264 : return -(int) before_gmtoff;
1666 : }
1667 :
1668 : /*
1669 : * Form the candidate pg_time_t values with local-time adjustment
1670 : */
1671 52806 : beforetime = mytime - before_gmtoff;
1672 52806 : if ((before_gmtoff > 0 &&
1673 12 : mytime < 0 && beforetime > 0) ||
1674 52806 : (before_gmtoff <= 0 &&
1675 39994 : mytime > 0 && beforetime < 0))
1676 0 : goto overflow;
1677 52806 : aftertime = mytime - after_gmtoff;
1678 52806 : if ((after_gmtoff > 0 &&
1679 12 : mytime < 0 && aftertime > 0) ||
1680 52806 : (after_gmtoff <= 0 &&
1681 39994 : 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 52806 : if (beforetime < boundary && aftertime < boundary)
1690 : {
1691 52228 : tm->tm_isdst = before_isdst;
1692 52228 : *tp = beforetime;
1693 52228 : return -(int) before_gmtoff;
1694 : }
1695 578 : if (beforetime > boundary && aftertime >= boundary)
1696 : {
1697 428 : tm->tm_isdst = after_isdst;
1698 428 : *tp = aftertime;
1699 428 : 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 4114 : DecodeTimeOnly(char **field, int *ftype, int nf,
1865 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
1866 : DateTimeErrorExtra *extra)
1867 : {
1868 4114 : int fmask = 0,
1869 : tmask,
1870 : type;
1871 4114 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
1872 : int i;
1873 : int val;
1874 : int dterr;
1875 4114 : bool isjulian = false;
1876 4114 : bool is2digits = false;
1877 4114 : bool bc = false;
1878 4114 : int mer = HR24;
1879 4114 : pg_tz *namedTz = NULL;
1880 4114 : pg_tz *abbrevTz = NULL;
1881 4114 : char *abbrev = NULL;
1882 : pg_tz *valtz;
1883 :
1884 4114 : *dtype = DTK_TIME;
1885 4114 : tm->tm_hour = 0;
1886 4114 : tm->tm_min = 0;
1887 4114 : tm->tm_sec = 0;
1888 4114 : *fsec = 0;
1889 : /* don't know daylight savings time status apriori */
1890 4114 : tm->tm_isdst = -1;
1891 :
1892 4114 : if (tzp != NULL)
1893 4114 : *tzp = 0;
1894 :
1895 10542 : for (i = 0; i < nf; i++)
1896 : {
1897 6440 : 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 4024 : case DTK_TIME:
1973 4024 : dterr = DecodeTime(field[i], (fmask | DTK_DATE_M),
1974 : INTERVAL_FULL_RANGE,
1975 : &tmask, tm, fsec);
1976 4024 : if (dterr)
1977 0 : return dterr;
1978 4024 : break;
1979 :
1980 558 : case DTK_TZ:
1981 : {
1982 : int tz;
1983 :
1984 558 : if (tzp == NULL)
1985 0 : return DTERR_BAD_FORMAT;
1986 :
1987 558 : dterr = DecodeTimezone(field[i], &tz);
1988 558 : if (dterr)
1989 0 : return dterr;
1990 558 : *tzp = tz;
1991 558 : tmask = DTK_M(TZ);
1992 : }
1993 558 : break;
1994 :
1995 96 : case DTK_NUMBER:
1996 :
1997 : /*
1998 : * Deal with cases where previous field labeled this one
1999 : */
2000 96 : if (ptype != 0)
2001 : {
2002 : char *cp;
2003 : int value;
2004 :
2005 72 : errno = 0;
2006 72 : value = strtoint(field[i], &cp, 10);
2007 72 : if (errno == ERANGE)
2008 12 : return DTERR_FIELD_OVERFLOW;
2009 72 : if (*cp != '.' && *cp != '\0')
2010 0 : return DTERR_BAD_FORMAT;
2011 :
2012 : switch (ptype)
2013 : {
2014 6 : case DTK_JULIAN:
2015 : /* previous field was a label for "julian date" */
2016 6 : if (tzp == NULL)
2017 0 : return DTERR_BAD_FORMAT;
2018 6 : if (value < 0)
2019 0 : return DTERR_FIELD_OVERFLOW;
2020 6 : tmask = DTK_DATE_M;
2021 6 : j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2022 6 : isjulian = true;
2023 :
2024 6 : if (*cp == '.')
2025 : {
2026 : double time;
2027 :
2028 0 : dterr = ParseFraction(cp, &time);
2029 0 : if (dterr)
2030 0 : return dterr;
2031 0 : time *= USECS_PER_DAY;
2032 0 : dt2time(time,
2033 : &tm->tm_hour, &tm->tm_min,
2034 : &tm->tm_sec, fsec);
2035 0 : tmask |= DTK_TIME_M;
2036 : }
2037 6 : break;
2038 :
2039 54 : case DTK_TIME:
2040 : /* previous field was "t" for ISO time */
2041 54 : dterr = DecodeNumberField(strlen(field[i]), field[i],
2042 : (fmask | DTK_DATE_M),
2043 : &tmask, tm,
2044 : fsec, &is2digits);
2045 54 : if (dterr < 0)
2046 0 : return dterr;
2047 54 : ftype[i] = dterr;
2048 :
2049 54 : if (tmask != DTK_TIME_M)
2050 0 : return DTERR_BAD_FORMAT;
2051 54 : break;
2052 :
2053 12 : default:
2054 12 : return DTERR_BAD_FORMAT;
2055 : break;
2056 : }
2057 :
2058 60 : ptype = 0;
2059 60 : *dtype = DTK_DATE;
2060 : }
2061 : else
2062 : {
2063 : char *cp;
2064 : int flen;
2065 :
2066 24 : flen = strlen(field[i]);
2067 24 : cp = strchr(field[i], '.');
2068 :
2069 : /* Embedded decimal? */
2070 24 : if (cp != NULL)
2071 : {
2072 : /*
2073 : * Under limited circumstances, we will accept a
2074 : * date...
2075 : */
2076 24 : if (i == 0 && nf >= 2 && ftype[nf - 1] == DTK_DATE)
2077 : {
2078 0 : dterr = DecodeDate(field[i], fmask,
2079 : &tmask, &is2digits, tm);
2080 0 : if (dterr)
2081 0 : return dterr;
2082 : }
2083 : /* embedded decimal and several digits before? */
2084 24 : else if (flen - strlen(cp) > 2)
2085 : {
2086 : /*
2087 : * Interpret as a concatenated date or time Set
2088 : * the type field to allow decoding other fields
2089 : * later. Example: 20011223 or 040506
2090 : */
2091 24 : dterr = DecodeNumberField(flen, field[i],
2092 : (fmask | DTK_DATE_M),
2093 : &tmask, tm,
2094 : fsec, &is2digits);
2095 24 : if (dterr < 0)
2096 0 : return dterr;
2097 24 : ftype[i] = dterr;
2098 : }
2099 : else
2100 0 : return DTERR_BAD_FORMAT;
2101 : }
2102 0 : else if (flen > 4)
2103 : {
2104 0 : dterr = DecodeNumberField(flen, field[i],
2105 : (fmask | DTK_DATE_M),
2106 : &tmask, tm,
2107 : fsec, &is2digits);
2108 0 : if (dterr < 0)
2109 0 : return dterr;
2110 0 : ftype[i] = dterr;
2111 : }
2112 : /* otherwise it is a single date/time field... */
2113 : else
2114 : {
2115 0 : dterr = DecodeNumber(flen, field[i],
2116 : false,
2117 : (fmask | DTK_DATE_M),
2118 : &tmask, tm,
2119 : fsec, &is2digits);
2120 0 : if (dterr)
2121 0 : return dterr;
2122 : }
2123 : }
2124 84 : break;
2125 :
2126 360 : case DTK_STRING:
2127 : case DTK_SPECIAL:
2128 : /* timezone abbrevs take precedence over built-in tokens */
2129 360 : dterr = DecodeTimezoneAbbrev(i, field[i],
2130 : &type, &val, &valtz, extra);
2131 360 : if (dterr)
2132 0 : return dterr;
2133 360 : if (type == UNKNOWN_FIELD)
2134 96 : type = DecodeSpecial(i, field[i], &val);
2135 360 : if (type == IGNORE_DTF)
2136 0 : continue;
2137 :
2138 360 : tmask = DTK_M(type);
2139 360 : switch (type)
2140 : {
2141 12 : case RESERV:
2142 12 : switch (val)
2143 : {
2144 12 : case DTK_NOW:
2145 12 : tmask = DTK_TIME_M;
2146 12 : *dtype = DTK_TIME;
2147 12 : GetCurrentTimeUsec(tm, fsec, NULL);
2148 12 : break;
2149 :
2150 0 : case DTK_ZULU:
2151 0 : tmask = (DTK_TIME_M | DTK_M(TZ));
2152 0 : *dtype = DTK_TIME;
2153 0 : tm->tm_hour = 0;
2154 0 : tm->tm_min = 0;
2155 0 : tm->tm_sec = 0;
2156 0 : tm->tm_isdst = 0;
2157 0 : break;
2158 :
2159 0 : default:
2160 0 : return DTERR_BAD_FORMAT;
2161 : }
2162 :
2163 12 : break;
2164 :
2165 0 : case DTZMOD:
2166 :
2167 : /*
2168 : * daylight savings time modifier (solves "MET DST"
2169 : * syntax)
2170 : */
2171 0 : tmask |= DTK_M(DTZ);
2172 0 : tm->tm_isdst = 1;
2173 0 : if (tzp == NULL)
2174 0 : return DTERR_BAD_FORMAT;
2175 0 : *tzp -= val;
2176 0 : break;
2177 :
2178 198 : case DTZ:
2179 :
2180 : /*
2181 : * set mask for TZ here _or_ check for DTZ later when
2182 : * getting default timezone
2183 : */
2184 198 : tmask |= DTK_M(TZ);
2185 198 : tm->tm_isdst = 1;
2186 198 : if (tzp == NULL)
2187 0 : return DTERR_BAD_FORMAT;
2188 198 : *tzp = -val;
2189 198 : ftype[i] = DTK_TZ;
2190 198 : break;
2191 :
2192 60 : case TZ:
2193 60 : tm->tm_isdst = 0;
2194 60 : if (tzp == NULL)
2195 0 : return DTERR_BAD_FORMAT;
2196 60 : *tzp = -val;
2197 60 : ftype[i] = DTK_TZ;
2198 60 : break;
2199 :
2200 6 : case DYNTZ:
2201 6 : tmask |= DTK_M(TZ);
2202 6 : if (tzp == NULL)
2203 0 : return DTERR_BAD_FORMAT;
2204 : /* we'll determine the actual offset later */
2205 6 : abbrevTz = valtz;
2206 6 : abbrev = field[i];
2207 6 : ftype[i] = DTK_TZ;
2208 6 : break;
2209 :
2210 12 : case AMPM:
2211 12 : mer = val;
2212 12 : break;
2213 :
2214 0 : case ADBC:
2215 0 : bc = (val == BC);
2216 0 : break;
2217 :
2218 18 : case UNITS:
2219 18 : tmask = 0;
2220 : /* reject consecutive unhandled units */
2221 18 : if (ptype != 0)
2222 0 : return DTERR_BAD_FORMAT;
2223 18 : ptype = val;
2224 18 : break;
2225 :
2226 54 : case ISOTIME:
2227 54 : tmask = 0;
2228 : /* reject consecutive unhandled units */
2229 54 : if (ptype != 0)
2230 0 : return DTERR_BAD_FORMAT;
2231 54 : ptype = val;
2232 54 : break;
2233 :
2234 0 : case UNKNOWN_FIELD:
2235 :
2236 : /*
2237 : * Before giving up and declaring error, check to see
2238 : * if it is an all-alpha timezone name.
2239 : */
2240 0 : namedTz = pg_tzset(field[i]);
2241 0 : if (!namedTz)
2242 0 : return DTERR_BAD_FORMAT;
2243 : /* we'll apply the zone setting below */
2244 0 : tmask = DTK_M(TZ);
2245 0 : break;
2246 :
2247 0 : default:
2248 0 : return DTERR_BAD_FORMAT;
2249 : }
2250 360 : break;
2251 :
2252 0 : default:
2253 0 : return DTERR_BAD_FORMAT;
2254 : }
2255 :
2256 6428 : if (tmask & fmask)
2257 0 : return DTERR_BAD_FORMAT;
2258 6428 : fmask |= tmask;
2259 : } /* end loop over fields */
2260 :
2261 : /* reject if prefix type appeared and was never handled */
2262 4102 : if (ptype != 0)
2263 0 : return DTERR_BAD_FORMAT;
2264 :
2265 : /* do final checking/adjustment of Y/M/D fields */
2266 4102 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
2267 4102 : if (dterr)
2268 0 : return dterr;
2269 :
2270 : /* handle AM/PM */
2271 4102 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
2272 0 : return DTERR_FIELD_OVERFLOW;
2273 4102 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
2274 0 : tm->tm_hour = 0;
2275 4102 : else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
2276 12 : tm->tm_hour += HOURS_PER_DAY / 2;
2277 :
2278 : /* check for time overflow */
2279 4102 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec, *fsec))
2280 72 : return DTERR_FIELD_OVERFLOW;
2281 :
2282 4030 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2283 0 : return DTERR_BAD_FORMAT;
2284 :
2285 : /*
2286 : * If we had a full timezone spec, compute the offset (we could not do it
2287 : * before, because we may need the date to resolve DST status).
2288 : */
2289 4030 : if (namedTz != NULL)
2290 : {
2291 : long int gmtoff;
2292 :
2293 : /* daylight savings time modifier disallowed with full TZ */
2294 1192 : if (fmask & DTK_M(DTZMOD))
2295 42 : return DTERR_BAD_FORMAT;
2296 :
2297 : /* if non-DST zone, we do not need to know the date */
2298 1192 : if (pg_get_timezone_offset(namedTz, &gmtoff))
2299 : {
2300 1114 : *tzp = -(int) gmtoff;
2301 : }
2302 : else
2303 : {
2304 : /* a date has to be specified */
2305 78 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2306 42 : return DTERR_BAD_FORMAT;
2307 36 : *tzp = DetermineTimeZoneOffset(tm, namedTz);
2308 : }
2309 : }
2310 :
2311 : /*
2312 : * Likewise, if we had a dynamic timezone abbreviation, resolve it now.
2313 : */
2314 3988 : if (abbrevTz != NULL)
2315 : {
2316 : struct pg_tm tt,
2317 0 : *tmp = &tt;
2318 :
2319 : /*
2320 : * daylight savings time modifier but no standard timezone? then error
2321 : */
2322 0 : if (fmask & DTK_M(DTZMOD))
2323 0 : return DTERR_BAD_FORMAT;
2324 :
2325 0 : if ((fmask & DTK_DATE_M) == 0)
2326 0 : GetCurrentDateTime(tmp);
2327 : else
2328 : {
2329 : /* a date has to be specified */
2330 0 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2331 0 : return DTERR_BAD_FORMAT;
2332 0 : tmp->tm_year = tm->tm_year;
2333 0 : tmp->tm_mon = tm->tm_mon;
2334 0 : tmp->tm_mday = tm->tm_mday;
2335 : }
2336 0 : tmp->tm_hour = tm->tm_hour;
2337 0 : tmp->tm_min = tm->tm_min;
2338 0 : tmp->tm_sec = tm->tm_sec;
2339 0 : *tzp = DetermineTimeZoneAbbrevOffset(tmp, abbrev, abbrevTz);
2340 0 : tm->tm_isdst = tmp->tm_isdst;
2341 : }
2342 :
2343 : /* timezone not specified? then use session timezone */
2344 3988 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
2345 : {
2346 : struct pg_tm tt,
2347 2058 : *tmp = &tt;
2348 :
2349 : /*
2350 : * daylight savings time modifier but no standard timezone? then error
2351 : */
2352 2058 : if (fmask & DTK_M(DTZMOD))
2353 0 : return DTERR_BAD_FORMAT;
2354 :
2355 2058 : if ((fmask & DTK_DATE_M) == 0)
2356 1980 : GetCurrentDateTime(tmp);
2357 : else
2358 : {
2359 : /* a date has to be specified */
2360 78 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2361 0 : return DTERR_BAD_FORMAT;
2362 78 : tmp->tm_year = tm->tm_year;
2363 78 : tmp->tm_mon = tm->tm_mon;
2364 78 : tmp->tm_mday = tm->tm_mday;
2365 : }
2366 2058 : tmp->tm_hour = tm->tm_hour;
2367 2058 : tmp->tm_min = tm->tm_min;
2368 2058 : tmp->tm_sec = tm->tm_sec;
2369 2058 : *tzp = DetermineTimeZoneOffset(tmp, session_timezone);
2370 2058 : tm->tm_isdst = tmp->tm_isdst;
2371 : }
2372 :
2373 3988 : return 0;
2374 : }
2375 :
2376 : /* DecodeDate()
2377 : * Decode date string which includes delimiters.
2378 : * Return 0 if okay, a DTERR code if not.
2379 : *
2380 : * str: field to be parsed
2381 : * fmask: bitmask for field types already seen
2382 : * *tmask: receives bitmask for fields found here
2383 : * *is2digits: set to true if we find 2-digit year
2384 : * *tm: field values are stored into appropriate members of this struct
2385 : */
2386 : static int
2387 64862 : DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
2388 : struct pg_tm *tm)
2389 : {
2390 : fsec_t fsec;
2391 64862 : int nf = 0;
2392 : int i,
2393 : len;
2394 : int dterr;
2395 64862 : bool haveTextMonth = false;
2396 : int type,
2397 : val,
2398 64862 : dmask = 0;
2399 : char *field[MAXDATEFIELDS];
2400 :
2401 64862 : *tmask = 0;
2402 :
2403 : /* parse this string... */
2404 259382 : while (*str != '\0' && nf < MAXDATEFIELDS)
2405 : {
2406 : /* skip field separators */
2407 194520 : while (*str != '\0' && !isalnum((unsigned char) *str))
2408 0 : str++;
2409 :
2410 194520 : if (*str == '\0')
2411 0 : return DTERR_BAD_FORMAT; /* end of string after separator */
2412 :
2413 194520 : field[nf] = str;
2414 194520 : if (isdigit((unsigned char) *str))
2415 : {
2416 712746 : while (isdigit((unsigned char) *str))
2417 518370 : str++;
2418 : }
2419 144 : else if (isalpha((unsigned char) *str))
2420 : {
2421 576 : while (isalpha((unsigned char) *str))
2422 432 : str++;
2423 : }
2424 :
2425 : /* Just get rid of any non-digit, non-alpha characters... */
2426 194520 : if (*str != '\0')
2427 129694 : *str++ = '\0';
2428 194520 : nf++;
2429 : }
2430 :
2431 : /* look first for text fields, since that will be unambiguous month */
2432 259382 : for (i = 0; i < nf; i++)
2433 : {
2434 194520 : if (isalpha((unsigned char) *field[i]))
2435 : {
2436 144 : type = DecodeSpecial(i, field[i], &val);
2437 144 : if (type == IGNORE_DTF)
2438 0 : continue;
2439 :
2440 144 : dmask = DTK_M(type);
2441 144 : switch (type)
2442 : {
2443 144 : case MONTH:
2444 144 : tm->tm_mon = val;
2445 144 : haveTextMonth = true;
2446 144 : break;
2447 :
2448 0 : default:
2449 0 : return DTERR_BAD_FORMAT;
2450 : }
2451 144 : if (fmask & dmask)
2452 0 : return DTERR_BAD_FORMAT;
2453 :
2454 144 : fmask |= dmask;
2455 144 : *tmask |= dmask;
2456 :
2457 : /* mark this field as being completed */
2458 144 : field[i] = NULL;
2459 : }
2460 : }
2461 :
2462 : /* now pick up remaining numeric fields */
2463 259382 : for (i = 0; i < nf; i++)
2464 : {
2465 194520 : if (field[i] == NULL)
2466 144 : continue;
2467 :
2468 194376 : if ((len = strlen(field[i])) <= 0)
2469 0 : return DTERR_BAD_FORMAT;
2470 :
2471 194376 : dterr = DecodeNumber(len, field[i], haveTextMonth, fmask,
2472 : &dmask, tm,
2473 : &fsec, is2digits);
2474 194376 : if (dterr)
2475 0 : return dterr;
2476 :
2477 194376 : if (fmask & dmask)
2478 0 : return DTERR_BAD_FORMAT;
2479 :
2480 194376 : fmask |= dmask;
2481 194376 : *tmask |= dmask;
2482 : }
2483 :
2484 64862 : if ((fmask & ~(DTK_M(DOY) | DTK_M(TZ))) != DTK_DATE_M)
2485 36 : return DTERR_BAD_FORMAT;
2486 :
2487 : /* validation of the field values must wait until ValidateDate() */
2488 :
2489 64826 : return 0;
2490 : }
2491 :
2492 : /* ValidateDate()
2493 : * Check valid year/month/day values, handle BC and DOY cases
2494 : * Return 0 if okay, a DTERR code if not.
2495 : */
2496 : int
2497 78404 : ValidateDate(int fmask, bool isjulian, bool is2digits, bool bc,
2498 : struct pg_tm *tm)
2499 : {
2500 78404 : if (fmask & DTK_M(YEAR))
2501 : {
2502 74506 : if (isjulian)
2503 : {
2504 : /* tm_year is correct and should not be touched */
2505 : }
2506 68476 : else if (bc)
2507 : {
2508 : /* there is no year zero in AD/BC notation */
2509 306 : if (tm->tm_year <= 0)
2510 0 : return DTERR_FIELD_OVERFLOW;
2511 : /* internally, we represent 1 BC as year zero, 2 BC as -1, etc */
2512 306 : tm->tm_year = -(tm->tm_year - 1);
2513 : }
2514 68170 : else if (is2digits)
2515 : {
2516 : /* process 1 or 2-digit input as 1970-2069 AD, allow '0' and '00' */
2517 354 : if (tm->tm_year < 0) /* just paranoia */
2518 0 : return DTERR_FIELD_OVERFLOW;
2519 354 : if (tm->tm_year < 70)
2520 174 : tm->tm_year += 2000;
2521 180 : else if (tm->tm_year < 100)
2522 180 : tm->tm_year += 1900;
2523 : }
2524 : else
2525 : {
2526 : /* there is no year zero in AD/BC notation */
2527 67816 : if (tm->tm_year <= 0)
2528 12 : return DTERR_FIELD_OVERFLOW;
2529 : }
2530 : }
2531 :
2532 : /* now that we have correct year, decode DOY */
2533 78392 : if (fmask & DTK_M(DOY))
2534 : {
2535 30 : j2date(date2j(tm->tm_year, 1, 1) + tm->tm_yday - 1,
2536 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2537 : }
2538 :
2539 : /* check for valid month */
2540 78392 : if (fmask & DTK_M(MONTH))
2541 : {
2542 74470 : if (tm->tm_mon < 1 || tm->tm_mon > MONTHS_PER_YEAR)
2543 78 : return DTERR_MD_FIELD_OVERFLOW;
2544 : }
2545 :
2546 : /* minimal check for valid day */
2547 78314 : if (fmask & DTK_M(DAY))
2548 : {
2549 74356 : if (tm->tm_mday < 1 || tm->tm_mday > 31)
2550 138 : return DTERR_MD_FIELD_OVERFLOW;
2551 : }
2552 :
2553 78176 : if ((fmask & DTK_DATE_M) == DTK_DATE_M)
2554 : {
2555 : /*
2556 : * Check for valid day of month, now that we know for sure the month
2557 : * and year. Note we don't use MD_FIELD_OVERFLOW here, since it seems
2558 : * unlikely that "Feb 29" is a YMD-order error.
2559 : */
2560 74200 : if (tm->tm_mday > day_tab[isleap(tm->tm_year)][tm->tm_mon - 1])
2561 48 : return DTERR_FIELD_OVERFLOW;
2562 : }
2563 :
2564 78128 : return 0;
2565 : }
2566 :
2567 :
2568 : /* DecodeTimeCommon()
2569 : * Decode time string which includes delimiters.
2570 : * Return 0 if okay, a DTERR code if not.
2571 : * tmask and itm are output parameters.
2572 : *
2573 : * This code is shared between the timestamp and interval cases.
2574 : * We return a struct pg_itm (of which only the tm_usec, tm_sec, tm_min,
2575 : * and tm_hour fields are used) and let the wrapper functions below
2576 : * convert and range-check as necessary.
2577 : */
2578 : static int
2579 60688 : DecodeTimeCommon(char *str, int fmask, int range,
2580 : int *tmask, struct pg_itm *itm)
2581 : {
2582 : char *cp;
2583 : int dterr;
2584 60688 : fsec_t fsec = 0;
2585 :
2586 60688 : *tmask = DTK_TIME_M;
2587 :
2588 60688 : errno = 0;
2589 60688 : itm->tm_hour = strtoi64(str, &cp, 10);
2590 60688 : if (errno == ERANGE)
2591 0 : return DTERR_FIELD_OVERFLOW;
2592 60688 : if (*cp != ':')
2593 0 : return DTERR_BAD_FORMAT;
2594 60688 : errno = 0;
2595 60688 : itm->tm_min = strtoint(cp + 1, &cp, 10);
2596 60688 : if (errno == ERANGE)
2597 0 : return DTERR_FIELD_OVERFLOW;
2598 60688 : if (*cp == '\0')
2599 : {
2600 1674 : itm->tm_sec = 0;
2601 : /* If it's a MINUTE TO SECOND interval, take 2 fields as being mm:ss */
2602 1674 : if (range == (INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND)))
2603 : {
2604 18 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2605 0 : return DTERR_FIELD_OVERFLOW;
2606 18 : itm->tm_sec = itm->tm_min;
2607 18 : itm->tm_min = (int) itm->tm_hour;
2608 18 : itm->tm_hour = 0;
2609 : }
2610 : }
2611 59014 : else if (*cp == '.')
2612 : {
2613 : /* always assume mm:ss.sss is MINUTE TO SECOND */
2614 48 : dterr = ParseFractionalSecond(cp, &fsec);
2615 48 : if (dterr)
2616 12 : return dterr;
2617 36 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2618 0 : return DTERR_FIELD_OVERFLOW;
2619 36 : itm->tm_sec = itm->tm_min;
2620 36 : itm->tm_min = (int) itm->tm_hour;
2621 36 : itm->tm_hour = 0;
2622 : }
2623 58966 : else if (*cp == ':')
2624 : {
2625 58966 : errno = 0;
2626 58966 : itm->tm_sec = strtoint(cp + 1, &cp, 10);
2627 58966 : if (errno == ERANGE)
2628 0 : return DTERR_FIELD_OVERFLOW;
2629 58966 : if (*cp == '.')
2630 : {
2631 22632 : dterr = ParseFractionalSecond(cp, &fsec);
2632 22632 : if (dterr)
2633 0 : return dterr;
2634 : }
2635 36334 : else if (*cp != '\0')
2636 0 : return DTERR_BAD_FORMAT;
2637 : }
2638 : else
2639 0 : return DTERR_BAD_FORMAT;
2640 :
2641 : /* do a sanity check; but caller must check the range of tm_hour */
2642 60676 : if (itm->tm_hour < 0 ||
2643 60676 : itm->tm_min < 0 || itm->tm_min > MINS_PER_HOUR - 1 ||
2644 60676 : itm->tm_sec < 0 || itm->tm_sec > SECS_PER_MINUTE ||
2645 60676 : fsec < 0 || fsec > USECS_PER_SEC)
2646 0 : return DTERR_FIELD_OVERFLOW;
2647 :
2648 60676 : itm->tm_usec = (int) fsec;
2649 :
2650 60676 : return 0;
2651 : }
2652 :
2653 : /* DecodeTime()
2654 : * Decode time string which includes delimiters.
2655 : * Return 0 if okay, a DTERR code if not.
2656 : *
2657 : * This version is used for timestamps. The results are returned into
2658 : * the tm_hour/tm_min/tm_sec fields of *tm, and microseconds into *fsec.
2659 : */
2660 : static int
2661 58686 : DecodeTime(char *str, int fmask, int range,
2662 : int *tmask, struct pg_tm *tm, fsec_t *fsec)
2663 : {
2664 : struct pg_itm itm;
2665 : int dterr;
2666 :
2667 58686 : dterr = DecodeTimeCommon(str, fmask, range,
2668 : tmask, &itm);
2669 58686 : if (dterr)
2670 0 : return dterr;
2671 :
2672 58686 : if (itm.tm_hour > INT_MAX)
2673 0 : return DTERR_FIELD_OVERFLOW;
2674 58686 : tm->tm_hour = (int) itm.tm_hour;
2675 58686 : tm->tm_min = itm.tm_min;
2676 58686 : tm->tm_sec = itm.tm_sec;
2677 58686 : *fsec = itm.tm_usec;
2678 :
2679 58686 : return 0;
2680 : }
2681 :
2682 : /* DecodeTimeForInterval()
2683 : * Decode time string which includes delimiters.
2684 : * Return 0 if okay, a DTERR code if not.
2685 : *
2686 : * This version is used for intervals. The results are returned into
2687 : * itm_in->tm_usec.
2688 : */
2689 : static int
2690 2002 : DecodeTimeForInterval(char *str, int fmask, int range,
2691 : int *tmask, struct pg_itm_in *itm_in)
2692 : {
2693 : struct pg_itm itm;
2694 : int dterr;
2695 :
2696 2002 : dterr = DecodeTimeCommon(str, fmask, range,
2697 : tmask, &itm);
2698 2002 : if (dterr)
2699 12 : return dterr;
2700 :
2701 1990 : itm_in->tm_usec = itm.tm_usec;
2702 1990 : if (!int64_multiply_add(itm.tm_hour, USECS_PER_HOUR, &itm_in->tm_usec) ||
2703 1990 : !int64_multiply_add(itm.tm_min, USECS_PER_MINUTE, &itm_in->tm_usec) ||
2704 1990 : !int64_multiply_add(itm.tm_sec, USECS_PER_SEC, &itm_in->tm_usec))
2705 6 : return DTERR_FIELD_OVERFLOW;
2706 :
2707 1984 : return 0;
2708 : }
2709 :
2710 :
2711 : /* DecodeNumber()
2712 : * Interpret plain numeric field as a date value in context.
2713 : * Return 0 if okay, a DTERR code if not.
2714 : */
2715 : static int
2716 200144 : DecodeNumber(int flen, char *str, bool haveTextMonth, int fmask,
2717 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2718 : {
2719 : int val;
2720 : char *cp;
2721 : int dterr;
2722 :
2723 200144 : *tmask = 0;
2724 :
2725 200144 : errno = 0;
2726 200144 : val = strtoint(str, &cp, 10);
2727 200144 : if (errno == ERANGE)
2728 0 : return DTERR_FIELD_OVERFLOW;
2729 200144 : if (cp == str)
2730 0 : return DTERR_BAD_FORMAT;
2731 :
2732 200144 : if (*cp == '.')
2733 : {
2734 : /*
2735 : * More than two digits before decimal point? Then could be a date or
2736 : * a run-together time: 2001.360 20011225 040506.789
2737 : */
2738 0 : if (cp - str > 2)
2739 : {
2740 0 : dterr = DecodeNumberField(flen, str,
2741 : (fmask | DTK_DATE_M),
2742 : tmask, tm,
2743 : fsec, is2digits);
2744 0 : if (dterr < 0)
2745 0 : return dterr;
2746 0 : return 0;
2747 : }
2748 :
2749 0 : dterr = ParseFractionalSecond(cp, fsec);
2750 0 : if (dterr)
2751 0 : return dterr;
2752 : }
2753 200144 : else if (*cp != '\0')
2754 0 : return DTERR_BAD_FORMAT;
2755 :
2756 : /* Special case for day of year */
2757 200144 : if (flen == 3 && (fmask & DTK_DATE_M) == DTK_M(YEAR) && val >= 1 &&
2758 : val <= 366)
2759 : {
2760 54 : *tmask = (DTK_M(DOY) | DTK_M(MONTH) | DTK_M(DAY));
2761 54 : tm->tm_yday = val;
2762 : /* tm_mon and tm_mday can't actually be set yet ... */
2763 54 : return 0;
2764 : }
2765 :
2766 : /* Switch based on what we have so far */
2767 200090 : switch (fmask & DTK_DATE_M)
2768 : {
2769 65052 : case 0:
2770 :
2771 : /*
2772 : * Nothing so far; make a decision about what we think the input
2773 : * is. There used to be lots of heuristics here, but the
2774 : * consensus now is to be paranoid. It *must* be either
2775 : * YYYY-MM-DD (with a more-than-two-digit year field), or the
2776 : * field order defined by DateOrder.
2777 : */
2778 65052 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2779 : {
2780 63348 : *tmask = DTK_M(YEAR);
2781 63348 : tm->tm_year = val;
2782 : }
2783 1704 : else if (DateOrder == DATEORDER_DMY)
2784 : {
2785 164 : *tmask = DTK_M(DAY);
2786 164 : tm->tm_mday = val;
2787 : }
2788 : else
2789 : {
2790 1540 : *tmask = DTK_M(MONTH);
2791 1540 : tm->tm_mon = val;
2792 : }
2793 65052 : break;
2794 :
2795 63240 : case (DTK_M(YEAR)):
2796 : /* Must be at second field of YY-MM-DD */
2797 63240 : *tmask = DTK_M(MONTH);
2798 63240 : tm->tm_mon = val;
2799 63240 : break;
2800 :
2801 4108 : case (DTK_M(MONTH)):
2802 4108 : if (haveTextMonth)
2803 : {
2804 : /*
2805 : * We are at the first numeric field of a date that included a
2806 : * textual month name. We want to support the variants
2807 : * MON-DD-YYYY, DD-MON-YYYY, and YYYY-MON-DD as unambiguous
2808 : * inputs. We will also accept MON-DD-YY or DD-MON-YY in
2809 : * either DMY or MDY modes, as well as YY-MON-DD in YMD mode.
2810 : */
2811 2624 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2812 : {
2813 72 : *tmask = DTK_M(YEAR);
2814 72 : tm->tm_year = val;
2815 : }
2816 : else
2817 : {
2818 2552 : *tmask = DTK_M(DAY);
2819 2552 : tm->tm_mday = val;
2820 : }
2821 : }
2822 : else
2823 : {
2824 : /* Must be at second field of MM-DD-YY */
2825 1484 : *tmask = DTK_M(DAY);
2826 1484 : tm->tm_mday = val;
2827 : }
2828 4108 : break;
2829 :
2830 63318 : case (DTK_M(YEAR) | DTK_M(MONTH)):
2831 63318 : if (haveTextMonth)
2832 : {
2833 : /* Need to accept DD-MON-YYYY even in YMD mode */
2834 126 : if (flen >= 3 && *is2digits)
2835 : {
2836 : /* Guess that first numeric field is day was wrong */
2837 30 : *tmask = DTK_M(DAY); /* YEAR is already set */
2838 30 : tm->tm_mday = tm->tm_year;
2839 30 : tm->tm_year = val;
2840 30 : *is2digits = false;
2841 : }
2842 : else
2843 : {
2844 96 : *tmask = DTK_M(DAY);
2845 96 : tm->tm_mday = val;
2846 : }
2847 : }
2848 : else
2849 : {
2850 : /* Must be at third field of YY-MM-DD */
2851 63192 : *tmask = DTK_M(DAY);
2852 63192 : tm->tm_mday = val;
2853 : }
2854 63318 : break;
2855 :
2856 146 : case (DTK_M(DAY)):
2857 : /* Must be at second field of DD-MM-YY */
2858 146 : *tmask = DTK_M(MONTH);
2859 146 : tm->tm_mon = val;
2860 146 : break;
2861 :
2862 4214 : case (DTK_M(MONTH) | DTK_M(DAY)):
2863 : /* Must be at third field of DD-MM-YY or MM-DD-YY */
2864 4214 : *tmask = DTK_M(YEAR);
2865 4214 : tm->tm_year = val;
2866 4214 : break;
2867 :
2868 12 : case (DTK_M(YEAR) | DTK_M(MONTH) | DTK_M(DAY)):
2869 : /* we have all the date, so it must be a time field */
2870 12 : dterr = DecodeNumberField(flen, str, fmask,
2871 : tmask, tm,
2872 : fsec, is2digits);
2873 12 : if (dterr < 0)
2874 12 : return dterr;
2875 0 : return 0;
2876 :
2877 0 : default:
2878 : /* Anything else is bogus input */
2879 0 : return DTERR_BAD_FORMAT;
2880 : }
2881 :
2882 : /*
2883 : * When processing a year field, mark it for adjustment if it's only one
2884 : * or two digits.
2885 : */
2886 200078 : if (*tmask == DTK_M(YEAR))
2887 67634 : *is2digits = (flen <= 2);
2888 :
2889 200078 : return 0;
2890 : }
2891 :
2892 :
2893 : /* DecodeNumberField()
2894 : * Interpret numeric string as a concatenated date or time field.
2895 : * Return a DTK token (>= 0) if successful, a DTERR code (< 0) if not.
2896 : *
2897 : * Use the context of previously decoded fields to help with
2898 : * the interpretation.
2899 : */
2900 : static int
2901 494 : DecodeNumberField(int len, char *str, int fmask,
2902 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2903 : {
2904 : char *cp;
2905 :
2906 : /*
2907 : * Have a decimal point? Then this is a date or something with a seconds
2908 : * field...
2909 : */
2910 494 : if ((cp = strchr(str, '.')) != NULL)
2911 : {
2912 : /*
2913 : * Can we use ParseFractionalSecond here? Not clear whether trailing
2914 : * junk should be rejected ...
2915 : */
2916 114 : if (cp[1] == '\0')
2917 : {
2918 : /* avoid assuming that strtod will accept "." */
2919 0 : *fsec = 0;
2920 : }
2921 : else
2922 : {
2923 : double frac;
2924 :
2925 114 : errno = 0;
2926 114 : frac = strtod(cp, NULL);
2927 114 : if (errno != 0)
2928 0 : return DTERR_BAD_FORMAT;
2929 114 : *fsec = rint(frac * 1000000);
2930 : }
2931 : /* Now truncate off the fraction for further processing */
2932 114 : *cp = '\0';
2933 114 : len = strlen(str);
2934 : }
2935 : /* No decimal point and no complete date yet? */
2936 380 : else if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2937 : {
2938 248 : if (len >= 6)
2939 : {
2940 248 : *tmask = DTK_DATE_M;
2941 :
2942 : /*
2943 : * Start from end and consider first 2 as Day, next 2 as Month,
2944 : * and the rest as Year.
2945 : */
2946 248 : tm->tm_mday = atoi(str + (len - 2));
2947 248 : *(str + (len - 2)) = '\0';
2948 248 : tm->tm_mon = atoi(str + (len - 4));
2949 248 : *(str + (len - 4)) = '\0';
2950 248 : tm->tm_year = atoi(str);
2951 248 : if ((len - 4) == 2)
2952 18 : *is2digits = true;
2953 :
2954 248 : return DTK_DATE;
2955 : }
2956 : }
2957 :
2958 : /* not all time fields are specified? */
2959 246 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2960 : {
2961 : /* hhmmss */
2962 246 : if (len == 6)
2963 : {
2964 234 : *tmask = DTK_TIME_M;
2965 234 : tm->tm_sec = atoi(str + 4);
2966 234 : *(str + 4) = '\0';
2967 234 : tm->tm_min = atoi(str + 2);
2968 234 : *(str + 2) = '\0';
2969 234 : tm->tm_hour = atoi(str);
2970 :
2971 234 : return DTK_TIME;
2972 : }
2973 : /* hhmm? */
2974 12 : else if (len == 4)
2975 : {
2976 0 : *tmask = DTK_TIME_M;
2977 0 : tm->tm_sec = 0;
2978 0 : tm->tm_min = atoi(str + 2);
2979 0 : *(str + 2) = '\0';
2980 0 : tm->tm_hour = atoi(str);
2981 :
2982 0 : return DTK_TIME;
2983 : }
2984 : }
2985 :
2986 12 : return DTERR_BAD_FORMAT;
2987 : }
2988 :
2989 :
2990 : /* DecodeTimezone()
2991 : * Interpret string as a numeric timezone.
2992 : *
2993 : * Return 0 if okay (and set *tzp), a DTERR code if not okay.
2994 : */
2995 : int
2996 37444 : DecodeTimezone(const char *str, int *tzp)
2997 : {
2998 : int tz;
2999 : int hr,
3000 : min,
3001 37444 : sec = 0;
3002 : char *cp;
3003 :
3004 : /* leading character must be "+" or "-" */
3005 37444 : if (*str != '+' && *str != '-')
3006 60 : return DTERR_BAD_FORMAT;
3007 :
3008 37384 : errno = 0;
3009 37384 : hr = strtoint(str + 1, &cp, 10);
3010 37384 : if (errno == ERANGE)
3011 0 : return DTERR_TZDISP_OVERFLOW;
3012 :
3013 : /* explicit delimiter? */
3014 37384 : if (*cp == ':')
3015 : {
3016 108 : errno = 0;
3017 108 : min = strtoint(cp + 1, &cp, 10);
3018 108 : if (errno == ERANGE)
3019 0 : return DTERR_TZDISP_OVERFLOW;
3020 108 : if (*cp == ':')
3021 : {
3022 24 : errno = 0;
3023 24 : sec = strtoint(cp + 1, &cp, 10);
3024 24 : if (errno == ERANGE)
3025 0 : return DTERR_TZDISP_OVERFLOW;
3026 : }
3027 : }
3028 : /* otherwise, might have run things together... */
3029 37276 : else if (*cp == '\0' && strlen(str) > 3)
3030 : {
3031 72 : min = hr % 100;
3032 72 : hr = hr / 100;
3033 : /* we could, but don't, support a run-together hhmmss format */
3034 : }
3035 : else
3036 37204 : min = 0;
3037 :
3038 : /* Range-check the values; see notes in datatype/timestamp.h */
3039 37384 : if (hr < 0 || hr > MAX_TZDISP_HOUR)
3040 12 : return DTERR_TZDISP_OVERFLOW;
3041 37372 : if (min < 0 || min >= MINS_PER_HOUR)
3042 12 : return DTERR_TZDISP_OVERFLOW;
3043 37360 : if (sec < 0 || sec >= SECS_PER_MINUTE)
3044 0 : return DTERR_TZDISP_OVERFLOW;
3045 :
3046 37360 : tz = (hr * MINS_PER_HOUR + min) * SECS_PER_MINUTE + sec;
3047 37360 : if (*str == '-')
3048 826 : tz = -tz;
3049 :
3050 37360 : *tzp = -tz;
3051 :
3052 37360 : if (*cp != '\0')
3053 0 : return DTERR_BAD_FORMAT;
3054 :
3055 37360 : return 0;
3056 : }
3057 :
3058 :
3059 : /* DecodeTimezoneAbbrev()
3060 : * Interpret string as a timezone abbreviation, if possible.
3061 : *
3062 : * Sets *ftype to an abbreviation type (TZ, DTZ, or DYNTZ), or UNKNOWN_FIELD if
3063 : * string is not any known abbreviation. On success, set *offset and *tz to
3064 : * represent the UTC offset (for TZ or DTZ) or underlying zone (for DYNTZ).
3065 : * Note that full timezone names (such as America/New_York) are not handled
3066 : * here, mostly for historical reasons.
3067 : *
3068 : * The function result is 0 or a DTERR code; in the latter case, *extra
3069 : * is filled as needed. Note that unknown-abbreviation is not considered
3070 : * an error case. Also note that many callers assume that the DTERR code
3071 : * is one that DateTimeParseError does not require "str" or "datatype"
3072 : * strings for.
3073 : *
3074 : * Given string must be lowercased already.
3075 : *
3076 : * Implement a cache lookup since it is likely that dates
3077 : * will be related in format.
3078 : */
3079 : int
3080 10376 : DecodeTimezoneAbbrev(int field, const char *lowtoken,
3081 : int *ftype, int *offset, pg_tz **tz,
3082 : DateTimeErrorExtra *extra)
3083 : {
3084 : const datetkn *tp;
3085 :
3086 10376 : tp = abbrevcache[field];
3087 : /* use strncmp so that we match truncated tokens */
3088 10376 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3089 : {
3090 7674 : if (zoneabbrevtbl)
3091 7674 : tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3092 7674 : zoneabbrevtbl->numabbrevs);
3093 : else
3094 0 : tp = NULL;
3095 : }
3096 10376 : if (tp == NULL)
3097 : {
3098 7264 : *ftype = UNKNOWN_FIELD;
3099 7264 : *offset = 0;
3100 7264 : *tz = NULL;
3101 : }
3102 : else
3103 : {
3104 3112 : abbrevcache[field] = tp;
3105 3112 : *ftype = tp->type;
3106 3112 : if (tp->type == DYNTZ)
3107 : {
3108 264 : *offset = 0;
3109 264 : *tz = FetchDynamicTimeZone(zoneabbrevtbl, tp, extra);
3110 264 : if (*tz == NULL)
3111 0 : return DTERR_BAD_ZONE_ABBREV;
3112 : }
3113 : else
3114 : {
3115 2848 : *offset = tp->value;
3116 2848 : *tz = NULL;
3117 : }
3118 : }
3119 :
3120 10376 : return 0;
3121 : }
3122 :
3123 :
3124 : /* DecodeSpecial()
3125 : * Decode text string using lookup table.
3126 : *
3127 : * Recognizes the keywords listed in datetktbl.
3128 : * Note: at one time this would also recognize timezone abbreviations,
3129 : * but no more; use DecodeTimezoneAbbrev for that.
3130 : *
3131 : * Given string must be lowercased already.
3132 : *
3133 : * Implement a cache lookup since it is likely that dates
3134 : * will be related in format.
3135 : */
3136 : int
3137 42074 : DecodeSpecial(int field, const char *lowtoken, int *val)
3138 : {
3139 : int type;
3140 : const datetkn *tp;
3141 :
3142 42074 : tp = datecache[field];
3143 : /* use strncmp so that we match truncated tokens */
3144 42074 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3145 : {
3146 9760 : tp = datebsearch(lowtoken, datetktbl, szdatetktbl);
3147 : }
3148 42074 : if (tp == NULL)
3149 : {
3150 108 : type = UNKNOWN_FIELD;
3151 108 : *val = 0;
3152 : }
3153 : else
3154 : {
3155 41966 : datecache[field] = tp;
3156 41966 : type = tp->type;
3157 41966 : *val = tp->value;
3158 : }
3159 :
3160 42074 : return type;
3161 : }
3162 :
3163 :
3164 : /* DecodeTimezoneName()
3165 : * Interpret string as a timezone abbreviation or name.
3166 : * Throw error if the name is not recognized.
3167 : *
3168 : * The return value indicates what kind of zone identifier it is:
3169 : * TZNAME_FIXED_OFFSET: fixed offset from UTC
3170 : * TZNAME_DYNTZ: dynamic timezone abbreviation
3171 : * TZNAME_ZONE: full tzdb zone name
3172 : *
3173 : * For TZNAME_FIXED_OFFSET, *offset receives the UTC offset (in seconds,
3174 : * with ISO sign convention: positive is east of Greenwich).
3175 : * For the other two cases, *tz receives the timezone struct representing
3176 : * the zone name or the abbreviation's underlying zone.
3177 : */
3178 : int
3179 786 : DecodeTimezoneName(const char *tzname, int *offset, pg_tz **tz)
3180 : {
3181 : char *lowzone;
3182 : int dterr,
3183 : type;
3184 : DateTimeErrorExtra extra;
3185 :
3186 : /*
3187 : * First we look in the timezone abbreviation table (to handle cases like
3188 : * "EST"), and if that fails, we look in the timezone database (to handle
3189 : * cases like "America/New_York"). This matches the order in which
3190 : * timestamp input checks the cases; it's important because the timezone
3191 : * database unwisely uses a few zone names that are identical to offset
3192 : * abbreviations.
3193 : */
3194 :
3195 : /* DecodeTimezoneAbbrev requires lowercase input */
3196 786 : lowzone = downcase_truncate_identifier(tzname,
3197 786 : strlen(tzname),
3198 : false);
3199 :
3200 786 : dterr = DecodeTimezoneAbbrev(0, lowzone, &type, offset, tz, &extra);
3201 786 : if (dterr)
3202 0 : DateTimeParseError(dterr, &extra, NULL, NULL, NULL);
3203 :
3204 786 : if (type == TZ || type == DTZ)
3205 : {
3206 : /* fixed-offset abbreviation, return the offset */
3207 270 : return TZNAME_FIXED_OFFSET;
3208 : }
3209 516 : else if (type == DYNTZ)
3210 : {
3211 : /* dynamic-offset abbreviation, return its referenced timezone */
3212 174 : return TZNAME_DYNTZ;
3213 : }
3214 : else
3215 : {
3216 : /* try it as a full zone name */
3217 342 : *tz = pg_tzset(tzname);
3218 342 : if (*tz == NULL)
3219 12 : ereport(ERROR,
3220 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
3221 : errmsg("time zone \"%s\" not recognized", tzname)));
3222 330 : return TZNAME_ZONE;
3223 : }
3224 : }
3225 :
3226 : /* DecodeTimezoneNameToTz()
3227 : * Interpret string as a timezone abbreviation or name.
3228 : * Throw error if the name is not recognized.
3229 : *
3230 : * This is a simple wrapper for DecodeTimezoneName that produces a pg_tz *
3231 : * result in all cases.
3232 : */
3233 : pg_tz *
3234 72 : DecodeTimezoneNameToTz(const char *tzname)
3235 : {
3236 : pg_tz *result;
3237 : int offset;
3238 :
3239 72 : if (DecodeTimezoneName(tzname, &offset, &result) == TZNAME_FIXED_OFFSET)
3240 : {
3241 : /* fixed-offset abbreviation, get a pg_tz descriptor for that */
3242 6 : result = pg_tzset_offset(-offset); /* flip to POSIX sign convention */
3243 : }
3244 72 : return result;
3245 : }
3246 :
3247 : /* DecodeTimezoneAbbrevPrefix()
3248 : * Interpret prefix of string as a timezone abbreviation, if possible.
3249 : *
3250 : * This has roughly the same functionality as DecodeTimezoneAbbrev(),
3251 : * but the API is adapted to the needs of formatting.c. Notably,
3252 : * we will match the longest possible prefix of the given string
3253 : * rather than insisting on a complete match, and downcasing is applied
3254 : * here rather than in the caller.
3255 : *
3256 : * Returns the length of the timezone abbreviation, or -1 if not recognized.
3257 : * On success, sets *offset to the GMT offset for the abbreviation if it
3258 : * is a fixed-offset abbreviation, or sets *tz to the pg_tz struct for
3259 : * a dynamic abbreviation.
3260 : */
3261 : int
3262 3534 : DecodeTimezoneAbbrevPrefix(const char *str, int *offset, pg_tz **tz)
3263 : {
3264 : char lowtoken[TOKMAXLEN + 1];
3265 : int len;
3266 :
3267 3534 : *offset = 0; /* avoid uninitialized vars on failure */
3268 3534 : *tz = NULL;
3269 :
3270 3534 : if (!zoneabbrevtbl)
3271 0 : return -1; /* no abbrevs known, so fail immediately */
3272 :
3273 : /* Downcase as much of the string as we could need */
3274 3654 : for (len = 0; len < TOKMAXLEN; len++)
3275 : {
3276 3654 : if (*str == '\0' || !isalpha((unsigned char) *str))
3277 : break;
3278 120 : lowtoken[len] = pg_tolower((unsigned char) *str++);
3279 : }
3280 3534 : lowtoken[len] = '\0';
3281 :
3282 : /*
3283 : * We could avoid doing repeated binary searches if we cared to duplicate
3284 : * datebsearch here, but it's not clear that such an optimization would be
3285 : * worth the trouble. In common cases there's probably not anything after
3286 : * the zone abbrev anyway. So just search with successively truncated
3287 : * strings.
3288 : */
3289 3576 : while (len > 0)
3290 : {
3291 72 : const datetkn *tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3292 72 : zoneabbrevtbl->numabbrevs);
3293 :
3294 72 : if (tp != NULL)
3295 : {
3296 30 : if (tp->type == DYNTZ)
3297 : {
3298 : DateTimeErrorExtra extra;
3299 6 : pg_tz *tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp,
3300 : &extra);
3301 :
3302 6 : if (tzp != NULL)
3303 : {
3304 : /* Caller must resolve the abbrev's current meaning */
3305 6 : *tz = tzp;
3306 6 : return len;
3307 : }
3308 : }
3309 : else
3310 : {
3311 : /* Fixed-offset zone abbrev, so it's easy */
3312 24 : *offset = tp->value;
3313 24 : return len;
3314 : }
3315 : }
3316 42 : lowtoken[--len] = '\0';
3317 : }
3318 :
3319 : /* Did not find a match */
3320 3504 : return -1;
3321 : }
3322 :
3323 :
3324 : /* ClearPgItmIn
3325 : *
3326 : * Zero out a pg_itm_in
3327 : */
3328 : static inline void
3329 12372 : ClearPgItmIn(struct pg_itm_in *itm_in)
3330 : {
3331 12372 : itm_in->tm_usec = 0;
3332 12372 : itm_in->tm_mday = 0;
3333 12372 : itm_in->tm_mon = 0;
3334 12372 : itm_in->tm_year = 0;
3335 12372 : }
3336 :
3337 :
3338 : /* DecodeInterval()
3339 : * Interpret previously parsed fields for general time interval.
3340 : * Returns 0 if successful, DTERR code if bogus input detected.
3341 : * dtype and itm_in are output parameters.
3342 : *
3343 : * Allow "date" field DTK_DATE since this could be just
3344 : * an unsigned floating point number. - thomas 1997-11-16
3345 : *
3346 : * Allow ISO-style time span, with implicit units on number of days
3347 : * preceding an hh:mm:ss field. - thomas 1998-04-30
3348 : *
3349 : * itm_in remains undefined for infinite interval values for which dtype alone
3350 : * suffices.
3351 : */
3352 : int
3353 11760 : DecodeInterval(char **field, int *ftype, int nf, int range,
3354 : int *dtype, struct pg_itm_in *itm_in)
3355 : {
3356 11760 : bool force_negative = false;
3357 11760 : bool is_before = false;
3358 11760 : bool parsing_unit_val = false;
3359 : char *cp;
3360 11760 : int fmask = 0,
3361 : tmask,
3362 : type,
3363 : uval;
3364 : int i;
3365 : int dterr;
3366 : int64 val;
3367 : double fval;
3368 :
3369 11760 : *dtype = DTK_DELTA;
3370 11760 : type = IGNORE_DTF;
3371 11760 : ClearPgItmIn(itm_in);
3372 :
3373 : /*----------
3374 : * The SQL standard defines the interval literal
3375 : * '-1 1:00:00'
3376 : * to mean "negative 1 days and negative 1 hours", while Postgres
3377 : * traditionally treats this as meaning "negative 1 days and positive
3378 : * 1 hours". In SQL_STANDARD intervalstyle, we apply the leading sign
3379 : * to all fields if there are no other explicit signs.
3380 : *
3381 : * We leave the signs alone if there are additional explicit signs.
3382 : * This protects us against misinterpreting postgres-style dump output,
3383 : * since the postgres-style output code has always put an explicit sign on
3384 : * all fields following a negative field. But note that SQL-spec output
3385 : * is ambiguous and can be misinterpreted on load! (So it's best practice
3386 : * to dump in postgres style, not SQL style.)
3387 : *----------
3388 : */
3389 11760 : if (IntervalStyle == INTSTYLE_SQL_STANDARD && nf > 0 && *field[0] == '-')
3390 : {
3391 38 : force_negative = true;
3392 : /* Check for additional explicit signs */
3393 244 : for (i = 1; i < nf; i++)
3394 : {
3395 224 : if (*field[i] == '-' || *field[i] == '+')
3396 : {
3397 18 : force_negative = false;
3398 18 : break;
3399 : }
3400 : }
3401 : }
3402 :
3403 : /* read through list backwards to pick up units before values */
3404 36920 : for (i = nf - 1; i >= 0; i--)
3405 : {
3406 26252 : switch (ftype[i])
3407 : {
3408 1234 : case DTK_TIME:
3409 1234 : dterr = DecodeTimeForInterval(field[i], fmask, range,
3410 : &tmask, itm_in);
3411 1234 : if (dterr)
3412 18 : return dterr;
3413 1216 : if (force_negative &&
3414 2 : itm_in->tm_usec > 0)
3415 2 : itm_in->tm_usec = -itm_in->tm_usec;
3416 1216 : type = DTK_DAY;
3417 1216 : parsing_unit_val = false;
3418 1216 : break;
3419 :
3420 2722 : case DTK_TZ:
3421 :
3422 : /*
3423 : * Timezone means a token with a leading sign character and at
3424 : * least one digit; there could be ':', '.', '-' embedded in
3425 : * it as well.
3426 : */
3427 : Assert(*field[i] == '-' || *field[i] == '+');
3428 :
3429 : /*
3430 : * Check for signed hh:mm or hh:mm:ss. If so, process exactly
3431 : * like DTK_TIME case above, plus handling the sign.
3432 : */
3433 3490 : if (strchr(field[i] + 1, ':') != NULL &&
3434 768 : DecodeTimeForInterval(field[i] + 1, fmask, range,
3435 : &tmask, itm_in) == 0)
3436 : {
3437 768 : if (*field[i] == '-')
3438 : {
3439 : /* flip the sign on time field */
3440 708 : if (itm_in->tm_usec == PG_INT64_MIN)
3441 0 : return DTERR_FIELD_OVERFLOW;
3442 708 : itm_in->tm_usec = -itm_in->tm_usec;
3443 : }
3444 :
3445 768 : if (force_negative &&
3446 0 : itm_in->tm_usec > 0)
3447 0 : itm_in->tm_usec = -itm_in->tm_usec;
3448 :
3449 : /*
3450 : * Set the next type to be a day, if units are not
3451 : * specified. This handles the case of '1 +02:03' since we
3452 : * are reading right to left.
3453 : */
3454 768 : type = DTK_DAY;
3455 768 : parsing_unit_val = false;
3456 768 : break;
3457 : }
3458 :
3459 : /*
3460 : * Otherwise, fall through to DTK_NUMBER case, which can
3461 : * handle signed float numbers and signed year-month values.
3462 : */
3463 :
3464 : /* FALLTHROUGH */
3465 :
3466 : case DTK_DATE:
3467 : case DTK_NUMBER:
3468 11982 : if (type == IGNORE_DTF)
3469 : {
3470 : /* use typmod to decide what rightmost field is */
3471 : switch (range)
3472 : {
3473 6 : case INTERVAL_MASK(YEAR):
3474 6 : type = DTK_YEAR;
3475 6 : break;
3476 30 : case INTERVAL_MASK(MONTH):
3477 : case INTERVAL_MASK(YEAR) | INTERVAL_MASK(MONTH):
3478 30 : type = DTK_MONTH;
3479 30 : break;
3480 18 : case INTERVAL_MASK(DAY):
3481 18 : type = DTK_DAY;
3482 18 : break;
3483 24 : case INTERVAL_MASK(HOUR):
3484 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR):
3485 24 : type = DTK_HOUR;
3486 24 : break;
3487 24 : case INTERVAL_MASK(MINUTE):
3488 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3489 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3490 24 : type = DTK_MINUTE;
3491 24 : break;
3492 60 : case INTERVAL_MASK(SECOND):
3493 : case INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3494 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3495 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3496 60 : type = DTK_SECOND;
3497 60 : break;
3498 438 : default:
3499 438 : type = DTK_SECOND;
3500 438 : break;
3501 : }
3502 11382 : }
3503 :
3504 11982 : errno = 0;
3505 11982 : val = strtoi64(field[i], &cp, 10);
3506 11982 : if (errno == ERANGE)
3507 12 : return DTERR_FIELD_OVERFLOW;
3508 :
3509 11970 : if (*cp == '-')
3510 : {
3511 : /* SQL "years-months" syntax */
3512 : int val2;
3513 :
3514 60 : val2 = strtoint(cp + 1, &cp, 10);
3515 60 : if (errno == ERANGE || val2 < 0 || val2 >= MONTHS_PER_YEAR)
3516 0 : return DTERR_FIELD_OVERFLOW;
3517 60 : if (*cp != '\0')
3518 0 : return DTERR_BAD_FORMAT;
3519 60 : type = DTK_MONTH;
3520 60 : if (*field[i] == '-')
3521 6 : val2 = -val2;
3522 60 : if (pg_mul_s64_overflow(val, MONTHS_PER_YEAR, &val))
3523 0 : return DTERR_FIELD_OVERFLOW;
3524 60 : if (pg_add_s64_overflow(val, val2, &val))
3525 0 : return DTERR_FIELD_OVERFLOW;
3526 60 : fval = 0;
3527 : }
3528 11910 : else if (*cp == '.')
3529 : {
3530 486 : dterr = ParseFraction(cp, &fval);
3531 486 : if (dterr)
3532 0 : return dterr;
3533 486 : if (*field[i] == '-')
3534 138 : fval = -fval;
3535 : }
3536 11424 : else if (*cp == '\0')
3537 11040 : fval = 0;
3538 : else
3539 384 : return DTERR_BAD_FORMAT;
3540 :
3541 11586 : tmask = 0; /* DTK_M(type); */
3542 :
3543 11586 : if (force_negative)
3544 : {
3545 : /* val and fval should be of same sign, but test anyway */
3546 80 : if (val > 0)
3547 60 : val = -val;
3548 80 : if (fval > 0)
3549 18 : fval = -fval;
3550 : }
3551 :
3552 : switch (type)
3553 : {
3554 318 : case DTK_MICROSEC:
3555 318 : if (!AdjustMicroseconds(val, fval, 1, itm_in))
3556 36 : return DTERR_FIELD_OVERFLOW;
3557 282 : tmask = DTK_M(MICROSECOND);
3558 282 : break;
3559 :
3560 102 : case DTK_MILLISEC:
3561 102 : if (!AdjustMicroseconds(val, fval, 1000, itm_in))
3562 12 : return DTERR_FIELD_OVERFLOW;
3563 90 : tmask = DTK_M(MILLISECOND);
3564 90 : break;
3565 :
3566 920 : case DTK_SECOND:
3567 920 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
3568 12 : return DTERR_FIELD_OVERFLOW;
3569 :
3570 : /*
3571 : * If any subseconds were specified, consider this
3572 : * microsecond and millisecond input as well.
3573 : */
3574 908 : if (fval == 0)
3575 746 : tmask = DTK_M(SECOND);
3576 : else
3577 162 : tmask = DTK_ALL_SECS_M;
3578 908 : break;
3579 :
3580 370 : case DTK_MINUTE:
3581 370 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
3582 12 : return DTERR_FIELD_OVERFLOW;
3583 358 : tmask = DTK_M(MINUTE);
3584 358 : break;
3585 :
3586 692 : case DTK_HOUR:
3587 692 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3588 12 : return DTERR_FIELD_OVERFLOW;
3589 680 : tmask = DTK_M(HOUR);
3590 680 : type = DTK_DAY; /* set for next field */
3591 680 : break;
3592 :
3593 7004 : case DTK_DAY:
3594 7004 : if (!AdjustDays(val, 1, itm_in) ||
3595 6932 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3596 90 : return DTERR_FIELD_OVERFLOW;
3597 6914 : tmask = DTK_M(DAY);
3598 6914 : break;
3599 :
3600 96 : case DTK_WEEK:
3601 96 : if (!AdjustDays(val, 7, itm_in) ||
3602 72 : !AdjustFractDays(fval, 7, itm_in))
3603 48 : return DTERR_FIELD_OVERFLOW;
3604 48 : tmask = DTK_M(WEEK);
3605 48 : break;
3606 :
3607 900 : case DTK_MONTH:
3608 900 : if (!AdjustMonths(val, itm_in) ||
3609 840 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3610 84 : return DTERR_FIELD_OVERFLOW;
3611 816 : tmask = DTK_M(MONTH);
3612 816 : break;
3613 :
3614 986 : case DTK_YEAR:
3615 986 : if (!AdjustYears(val, 1, itm_in) ||
3616 938 : !AdjustFractYears(fval, 1, itm_in))
3617 60 : return DTERR_FIELD_OVERFLOW;
3618 926 : tmask = DTK_M(YEAR);
3619 926 : break;
3620 :
3621 66 : case DTK_DECADE:
3622 66 : if (!AdjustYears(val, 10, itm_in) ||
3623 42 : !AdjustFractYears(fval, 10, itm_in))
3624 36 : return DTERR_FIELD_OVERFLOW;
3625 30 : tmask = DTK_M(DECADE);
3626 30 : break;
3627 :
3628 66 : case DTK_CENTURY:
3629 66 : if (!AdjustYears(val, 100, itm_in) ||
3630 42 : !AdjustFractYears(fval, 100, itm_in))
3631 36 : return DTERR_FIELD_OVERFLOW;
3632 30 : tmask = DTK_M(CENTURY);
3633 30 : break;
3634 :
3635 66 : case DTK_MILLENNIUM:
3636 66 : if (!AdjustYears(val, 1000, itm_in) ||
3637 42 : !AdjustFractYears(fval, 1000, itm_in))
3638 36 : return DTERR_FIELD_OVERFLOW;
3639 30 : tmask = DTK_M(MILLENNIUM);
3640 30 : break;
3641 :
3642 0 : default:
3643 0 : return DTERR_BAD_FORMAT;
3644 : }
3645 11112 : parsing_unit_val = false;
3646 11112 : break;
3647 :
3648 12268 : case DTK_STRING:
3649 : case DTK_SPECIAL:
3650 : /* reject consecutive unhandled units */
3651 12268 : if (parsing_unit_val)
3652 12 : return DTERR_BAD_FORMAT;
3653 12256 : type = DecodeUnits(i, field[i], &uval);
3654 12256 : if (type == UNKNOWN_FIELD)
3655 1062 : type = DecodeSpecial(i, field[i], &uval);
3656 12256 : if (type == IGNORE_DTF)
3657 0 : continue;
3658 :
3659 12256 : tmask = 0; /* DTK_M(type); */
3660 : switch (type)
3661 : {
3662 11092 : case UNITS:
3663 11092 : type = uval;
3664 11092 : parsing_unit_val = true;
3665 11092 : break;
3666 :
3667 102 : case AGO:
3668 :
3669 : /*
3670 : * "ago" is only allowed to appear at the end of the
3671 : * interval.
3672 : */
3673 102 : if (i != nf - 1)
3674 12 : return DTERR_BAD_FORMAT;
3675 90 : is_before = true;
3676 90 : type = uval;
3677 90 : break;
3678 :
3679 1026 : case RESERV:
3680 1026 : tmask = (DTK_DATE_M | DTK_TIME_M);
3681 :
3682 : /*
3683 : * Only reserved words corresponding to infinite
3684 : * intervals are accepted.
3685 : */
3686 1026 : if (uval != DTK_LATE && uval != DTK_EARLY)
3687 36 : return DTERR_BAD_FORMAT;
3688 :
3689 : /*
3690 : * Infinity cannot be followed by anything else. We
3691 : * could allow "ago" to reverse the sign of infinity
3692 : * but using signed infinity is more intuitive.
3693 : */
3694 990 : if (i != nf - 1)
3695 12 : return DTERR_BAD_FORMAT;
3696 :
3697 978 : *dtype = uval;
3698 978 : break;
3699 :
3700 36 : default:
3701 36 : return DTERR_BAD_FORMAT;
3702 : }
3703 12160 : break;
3704 :
3705 0 : default:
3706 0 : return DTERR_BAD_FORMAT;
3707 : }
3708 :
3709 25256 : if (tmask & fmask)
3710 96 : return DTERR_BAD_FORMAT;
3711 25160 : fmask |= tmask;
3712 : }
3713 :
3714 : /* ensure that at least one time field has been found */
3715 10668 : if (fmask == 0)
3716 6 : return DTERR_BAD_FORMAT;
3717 :
3718 : /* reject if unit appeared and was never handled */
3719 10662 : if (parsing_unit_val)
3720 6 : return DTERR_BAD_FORMAT;
3721 :
3722 : /* finally, AGO negates everything */
3723 10656 : if (is_before)
3724 : {
3725 42 : if (itm_in->tm_usec == PG_INT64_MIN ||
3726 30 : itm_in->tm_mday == INT_MIN ||
3727 24 : itm_in->tm_mon == INT_MIN ||
3728 18 : itm_in->tm_year == INT_MIN)
3729 24 : return DTERR_FIELD_OVERFLOW;
3730 :
3731 18 : itm_in->tm_usec = -itm_in->tm_usec;
3732 18 : itm_in->tm_mday = -itm_in->tm_mday;
3733 18 : itm_in->tm_mon = -itm_in->tm_mon;
3734 18 : itm_in->tm_year = -itm_in->tm_year;
3735 : }
3736 :
3737 10632 : return 0;
3738 : }
3739 :
3740 :
3741 : /*
3742 : * Helper functions to avoid duplicated code in DecodeISO8601Interval.
3743 : *
3744 : * Parse a decimal value and break it into integer and fractional parts.
3745 : * Set *endptr to end+1 of the parsed substring.
3746 : * Returns 0 or DTERR code.
3747 : */
3748 : static int
3749 954 : ParseISO8601Number(char *str, char **endptr, int64 *ipart, double *fpart)
3750 : {
3751 : double val;
3752 :
3753 : /*
3754 : * Historically this has accepted anything that strtod() would take,
3755 : * notably including "e" notation, so continue doing that. This is
3756 : * slightly annoying because the precision of double is less than that of
3757 : * int64, so we would lose accuracy for inputs larger than 2^53 or so.
3758 : * However, historically we rejected inputs outside the int32 range,
3759 : * making that concern moot. What we do now is reject abs(val) above
3760 : * 1.0e15 (a round number a bit less than 2^50), so that any accepted
3761 : * value will have an exact integer part, and thereby a fraction part with
3762 : * abs(*fpart) less than 1. In the absence of field complaints it doesn't
3763 : * seem worth working harder.
3764 : */
3765 954 : if (!(isdigit((unsigned char) *str) || *str == '-' || *str == '.'))
3766 0 : return DTERR_BAD_FORMAT;
3767 954 : errno = 0;
3768 954 : val = strtod(str, endptr);
3769 : /* did we not see anything that looks like a double? */
3770 954 : if (*endptr == str || errno != 0)
3771 6 : return DTERR_BAD_FORMAT;
3772 : /* watch out for overflow, including infinities; reject NaN too */
3773 948 : if (isnan(val) || val < -1.0e15 || val > 1.0e15)
3774 0 : return DTERR_FIELD_OVERFLOW;
3775 : /* be very sure we truncate towards zero (cf dtrunc()) */
3776 948 : if (val >= 0)
3777 732 : *ipart = (int64) floor(val);
3778 : else
3779 216 : *ipart = (int64) -floor(-val);
3780 948 : *fpart = val - *ipart;
3781 : /* Callers expect this to hold */
3782 : Assert(*fpart > -1.0 && *fpart < 1.0);
3783 948 : return 0;
3784 : }
3785 :
3786 : /*
3787 : * Determine number of integral digits in a valid ISO 8601 number field
3788 : * (we should ignore sign and any fraction part)
3789 : */
3790 : static int
3791 66 : ISO8601IntegerWidth(char *fieldstart)
3792 : {
3793 : /* We might have had a leading '-' */
3794 66 : if (*fieldstart == '-')
3795 18 : fieldstart++;
3796 66 : return strspn(fieldstart, "0123456789");
3797 : }
3798 :
3799 :
3800 : /* DecodeISO8601Interval()
3801 : * Decode an ISO 8601 time interval of the "format with designators"
3802 : * (section 4.4.3.2) or "alternative format" (section 4.4.3.3)
3803 : * Examples: P1D for 1 day
3804 : * PT1H for 1 hour
3805 : * P2Y6M7DT1H30M for 2 years, 6 months, 7 days 1 hour 30 min
3806 : * P0002-06-07T01:30:00 the same value in alternative format
3807 : *
3808 : * Returns 0 if successful, DTERR code if bogus input detected.
3809 : * Note: error code should be DTERR_BAD_FORMAT if input doesn't look like
3810 : * ISO8601, otherwise this could cause unexpected error messages.
3811 : * dtype and itm_in are output parameters.
3812 : *
3813 : * A couple exceptions from the spec:
3814 : * - a week field ('W') may coexist with other units
3815 : * - allows decimals in fields other than the least significant unit.
3816 : */
3817 : int
3818 612 : DecodeISO8601Interval(char *str,
3819 : int *dtype, struct pg_itm_in *itm_in)
3820 : {
3821 612 : bool datepart = true;
3822 612 : bool havefield = false;
3823 :
3824 612 : *dtype = DTK_DELTA;
3825 612 : ClearPgItmIn(itm_in);
3826 :
3827 612 : if (strlen(str) < 2 || str[0] != 'P')
3828 228 : return DTERR_BAD_FORMAT;
3829 :
3830 384 : str++;
3831 1122 : while (*str)
3832 : {
3833 : char *fieldstart;
3834 : int64 val;
3835 : double fval;
3836 : char unit;
3837 : int dterr;
3838 :
3839 1014 : if (*str == 'T') /* T indicates the beginning of the time part */
3840 : {
3841 198 : datepart = false;
3842 198 : havefield = false;
3843 198 : str++;
3844 240 : continue;
3845 : }
3846 :
3847 816 : fieldstart = str;
3848 816 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3849 816 : if (dterr)
3850 276 : return dterr;
3851 :
3852 : /*
3853 : * Note: we could step off the end of the string here. Code below
3854 : * *must* exit the loop if unit == '\0'.
3855 : */
3856 810 : unit = *str++;
3857 :
3858 810 : if (datepart)
3859 : {
3860 468 : switch (unit) /* before T: Y M W D */
3861 : {
3862 84 : case 'Y':
3863 84 : if (!AdjustYears(val, 1, itm_in) ||
3864 84 : !AdjustFractYears(fval, 1, itm_in))
3865 12 : return DTERR_FIELD_OVERFLOW;
3866 72 : break;
3867 108 : case 'M':
3868 108 : if (!AdjustMonths(val, itm_in) ||
3869 96 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3870 24 : return DTERR_FIELD_OVERFLOW;
3871 84 : break;
3872 54 : case 'W':
3873 54 : if (!AdjustDays(val, 7, itm_in) ||
3874 42 : !AdjustFractDays(fval, 7, itm_in))
3875 24 : return DTERR_FIELD_OVERFLOW;
3876 30 : break;
3877 132 : case 'D':
3878 132 : if (!AdjustDays(val, 1, itm_in) ||
3879 96 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3880 36 : return DTERR_FIELD_OVERFLOW;
3881 96 : break;
3882 30 : case 'T': /* ISO 8601 4.4.3.3 Alternative Format / Basic */
3883 : case '\0':
3884 30 : if (ISO8601IntegerWidth(fieldstart) == 8 && !havefield)
3885 : {
3886 6 : if (!AdjustYears(val / 10000, 1, itm_in) ||
3887 6 : !AdjustMonths((val / 100) % 100, itm_in) ||
3888 6 : !AdjustDays(val % 100, 1, itm_in) ||
3889 6 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3890 0 : return DTERR_FIELD_OVERFLOW;
3891 6 : if (unit == '\0')
3892 0 : return 0;
3893 6 : datepart = false;
3894 6 : havefield = false;
3895 6 : continue;
3896 : }
3897 : /* Else fall through to extended alternative format */
3898 : /* FALLTHROUGH */
3899 : case '-': /* ISO 8601 4.4.3.3 Alternative Format,
3900 : * Extended */
3901 84 : if (havefield)
3902 0 : return DTERR_BAD_FORMAT;
3903 :
3904 84 : if (!AdjustYears(val, 1, itm_in) ||
3905 72 : !AdjustFractYears(fval, 1, itm_in))
3906 12 : return DTERR_FIELD_OVERFLOW;
3907 72 : if (unit == '\0')
3908 6 : return 0;
3909 66 : if (unit == 'T')
3910 : {
3911 6 : datepart = false;
3912 6 : havefield = false;
3913 6 : continue;
3914 : }
3915 :
3916 60 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3917 60 : if (dterr)
3918 0 : return dterr;
3919 60 : if (!AdjustMonths(val, itm_in) ||
3920 54 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3921 6 : return DTERR_FIELD_OVERFLOW;
3922 54 : if (*str == '\0')
3923 6 : return 0;
3924 48 : if (*str == 'T')
3925 : {
3926 6 : datepart = false;
3927 6 : havefield = false;
3928 6 : continue;
3929 : }
3930 42 : if (*str != '-')
3931 0 : return DTERR_BAD_FORMAT;
3932 42 : str++;
3933 :
3934 42 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3935 42 : if (dterr)
3936 0 : return dterr;
3937 42 : if (!AdjustDays(val, 1, itm_in) ||
3938 36 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3939 6 : return DTERR_FIELD_OVERFLOW;
3940 36 : if (*str == '\0')
3941 12 : return 0;
3942 24 : if (*str == 'T')
3943 : {
3944 24 : datepart = false;
3945 24 : havefield = false;
3946 24 : continue;
3947 : }
3948 0 : return DTERR_BAD_FORMAT;
3949 0 : default:
3950 : /* not a valid date unit suffix */
3951 0 : return DTERR_BAD_FORMAT;
3952 : }
3953 : }
3954 : else
3955 : {
3956 342 : switch (unit) /* after T: H M S */
3957 : {
3958 108 : case 'H':
3959 108 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3960 36 : return DTERR_FIELD_OVERFLOW;
3961 72 : break;
3962 60 : case 'M':
3963 60 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
3964 0 : return DTERR_FIELD_OVERFLOW;
3965 60 : break;
3966 96 : case 'S':
3967 96 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
3968 12 : return DTERR_FIELD_OVERFLOW;
3969 84 : break;
3970 36 : case '\0': /* ISO 8601 4.4.3.3 Alternative Format */
3971 36 : if (ISO8601IntegerWidth(fieldstart) == 6 && !havefield)
3972 : {
3973 6 : if (!AdjustMicroseconds(val / 10000, 0, USECS_PER_HOUR, itm_in) ||
3974 6 : !AdjustMicroseconds((val / 100) % 100, 0, USECS_PER_MINUTE, itm_in) ||
3975 6 : !AdjustMicroseconds(val % 100, 0, USECS_PER_SEC, itm_in) ||
3976 6 : !AdjustFractMicroseconds(fval, 1, itm_in))
3977 0 : return DTERR_FIELD_OVERFLOW;
3978 6 : return 0;
3979 : }
3980 : /* Else fall through to extended alternative format */
3981 : /* FALLTHROUGH */
3982 : case ':': /* ISO 8601 4.4.3.3 Alternative Format,
3983 : * Extended */
3984 72 : if (havefield)
3985 0 : return DTERR_BAD_FORMAT;
3986 :
3987 72 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3988 30 : return DTERR_FIELD_OVERFLOW;
3989 42 : if (unit == '\0')
3990 18 : return 0;
3991 :
3992 24 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3993 24 : if (dterr)
3994 0 : return dterr;
3995 24 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
3996 6 : return DTERR_FIELD_OVERFLOW;
3997 18 : if (*str == '\0')
3998 6 : return 0;
3999 12 : if (*str != ':')
4000 0 : return DTERR_BAD_FORMAT;
4001 12 : str++;
4002 :
4003 12 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4004 12 : if (dterr)
4005 0 : return dterr;
4006 12 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
4007 0 : return DTERR_FIELD_OVERFLOW;
4008 12 : if (*str == '\0')
4009 12 : return 0;
4010 0 : return DTERR_BAD_FORMAT;
4011 :
4012 0 : default:
4013 : /* not a valid time unit suffix */
4014 0 : return DTERR_BAD_FORMAT;
4015 : }
4016 : }
4017 :
4018 498 : havefield = true;
4019 : }
4020 :
4021 108 : return 0;
4022 : }
4023 :
4024 :
4025 : /* DecodeUnits()
4026 : * Decode text string using lookup table.
4027 : *
4028 : * This routine recognizes keywords associated with time interval units.
4029 : *
4030 : * Given string must be lowercased already.
4031 : *
4032 : * Implement a cache lookup since it is likely that dates
4033 : * will be related in format.
4034 : */
4035 : int
4036 65258 : DecodeUnits(int field, const char *lowtoken, int *val)
4037 : {
4038 : int type;
4039 : const datetkn *tp;
4040 :
4041 65258 : tp = deltacache[field];
4042 : /* use strncmp so that we match truncated tokens */
4043 65258 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
4044 : {
4045 53300 : tp = datebsearch(lowtoken, deltatktbl, szdeltatktbl);
4046 : }
4047 65258 : if (tp == NULL)
4048 : {
4049 35008 : type = UNKNOWN_FIELD;
4050 35008 : *val = 0;
4051 : }
4052 : else
4053 : {
4054 30250 : deltacache[field] = tp;
4055 30250 : type = tp->type;
4056 30250 : *val = tp->value;
4057 : }
4058 :
4059 65258 : return type;
4060 : } /* DecodeUnits() */
4061 :
4062 : /*
4063 : * Report an error detected by one of the datetime input processing routines.
4064 : *
4065 : * dterr is the error code, and *extra contains any auxiliary info we need
4066 : * for the error report. extra can be NULL if not needed for the particular
4067 : * dterr value.
4068 : *
4069 : * str is the original input string, and datatype is the name of the datatype
4070 : * we were trying to accept. (For some DTERR codes, these are not used and
4071 : * can be NULL.)
4072 : *
4073 : * If escontext points to an ErrorSaveContext node, that is filled instead
4074 : * of throwing an error.
4075 : *
4076 : * Note: it might seem useless to distinguish DTERR_INTERVAL_OVERFLOW and
4077 : * DTERR_TZDISP_OVERFLOW from DTERR_FIELD_OVERFLOW, but SQL99 mandates three
4078 : * separate SQLSTATE codes, so ...
4079 : */
4080 : void
4081 1680 : DateTimeParseError(int dterr, DateTimeErrorExtra *extra,
4082 : const char *str, const char *datatype,
4083 : Node *escontext)
4084 : {
4085 1680 : switch (dterr)
4086 : {
4087 174 : case DTERR_FIELD_OVERFLOW:
4088 174 : errsave(escontext,
4089 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4090 : errmsg("date/time field value out of range: \"%s\"",
4091 : str)));
4092 24 : break;
4093 180 : case DTERR_MD_FIELD_OVERFLOW:
4094 : /* <nanny>same as above, but add hint about DateStyle</nanny> */
4095 180 : errsave(escontext,
4096 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4097 : errmsg("date/time field value out of range: \"%s\"",
4098 : str),
4099 : errhint("Perhaps you need a different \"datestyle\" setting.")));
4100 0 : break;
4101 720 : case DTERR_INTERVAL_OVERFLOW:
4102 720 : errsave(escontext,
4103 : (errcode(ERRCODE_INTERVAL_FIELD_OVERFLOW),
4104 : errmsg("interval field value out of range: \"%s\"",
4105 : str)));
4106 0 : break;
4107 12 : case DTERR_TZDISP_OVERFLOW:
4108 12 : errsave(escontext,
4109 : (errcode(ERRCODE_INVALID_TIME_ZONE_DISPLACEMENT_VALUE),
4110 : errmsg("time zone displacement out of range: \"%s\"",
4111 : str)));
4112 0 : break;
4113 36 : case DTERR_BAD_TIMEZONE:
4114 36 : errsave(escontext,
4115 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
4116 : errmsg("time zone \"%s\" not recognized",
4117 : extra->dtee_timezone)));
4118 24 : break;
4119 0 : case DTERR_BAD_ZONE_ABBREV:
4120 0 : errsave(escontext,
4121 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
4122 : errmsg("time zone \"%s\" not recognized",
4123 : extra->dtee_timezone),
4124 : errdetail("This time zone name appears in the configuration file for time zone abbreviation \"%s\".",
4125 : extra->dtee_abbrev)));
4126 0 : break;
4127 558 : case DTERR_BAD_FORMAT:
4128 : default:
4129 558 : errsave(escontext,
4130 : (errcode(ERRCODE_INVALID_DATETIME_FORMAT),
4131 : errmsg("invalid input syntax for type %s: \"%s\"",
4132 : datatype, str)));
4133 84 : break;
4134 : }
4135 132 : }
4136 :
4137 : /* datebsearch()
4138 : * Binary search -- from Knuth (6.2.1) Algorithm B. Special case like this
4139 : * is WAY faster than the generic bsearch().
4140 : */
4141 : static const datetkn *
4142 72624 : datebsearch(const char *key, const datetkn *base, int nel)
4143 : {
4144 72624 : if (nel > 0)
4145 : {
4146 72624 : const datetkn *last = base + nel - 1,
4147 : *position;
4148 : int result;
4149 :
4150 461432 : while (last >= base)
4151 : {
4152 417498 : position = base + ((last - base) >> 1);
4153 : /* precheck the first character for a bit of extra speed */
4154 417498 : result = (int) key[0] - (int) position->token[0];
4155 417498 : if (result == 0)
4156 : {
4157 : /* use strncmp so that we match truncated tokens */
4158 108870 : result = strncmp(key, position->token, TOKMAXLEN);
4159 108870 : if (result == 0)
4160 28690 : return position;
4161 : }
4162 388808 : if (result < 0)
4163 198294 : last = position - 1;
4164 : else
4165 190514 : base = position + 1;
4166 : }
4167 : }
4168 43934 : return NULL;
4169 : }
4170 :
4171 : /* EncodeTimezone()
4172 : * Copies representation of a numeric timezone offset to str.
4173 : *
4174 : * Returns a pointer to the new end of string. No NUL terminator is put
4175 : * there; callers are responsible for NUL terminating str themselves.
4176 : */
4177 : static char *
4178 59752 : EncodeTimezone(char *str, int tz, int style)
4179 : {
4180 : int hour,
4181 : min,
4182 : sec;
4183 :
4184 59752 : sec = abs(tz);
4185 59752 : min = sec / SECS_PER_MINUTE;
4186 59752 : sec -= min * SECS_PER_MINUTE;
4187 59752 : hour = min / MINS_PER_HOUR;
4188 59752 : min -= hour * MINS_PER_HOUR;
4189 :
4190 : /* TZ is negated compared to sign we wish to display ... */
4191 59752 : *str++ = (tz <= 0 ? '+' : '-');
4192 :
4193 59752 : if (sec != 0)
4194 : {
4195 0 : str = pg_ultostr_zeropad(str, hour, 2);
4196 0 : *str++ = ':';
4197 0 : str = pg_ultostr_zeropad(str, min, 2);
4198 0 : *str++ = ':';
4199 0 : str = pg_ultostr_zeropad(str, sec, 2);
4200 : }
4201 59752 : else if (min != 0 || style == USE_XSD_DATES)
4202 : {
4203 822 : str = pg_ultostr_zeropad(str, hour, 2);
4204 822 : *str++ = ':';
4205 822 : str = pg_ultostr_zeropad(str, min, 2);
4206 : }
4207 : else
4208 58930 : str = pg_ultostr_zeropad(str, hour, 2);
4209 59752 : return str;
4210 : }
4211 :
4212 : /* EncodeDateOnly()
4213 : * Encode date as local time.
4214 : */
4215 : void
4216 12936 : EncodeDateOnly(struct pg_tm *tm, int style, char *str)
4217 : {
4218 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4219 :
4220 12936 : switch (style)
4221 : {
4222 5686 : case USE_ISO_DATES:
4223 : case USE_XSD_DATES:
4224 : /* compatible with ISO date formats */
4225 5686 : str = pg_ultostr_zeropad(str,
4226 5686 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4227 5686 : *str++ = '-';
4228 5686 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4229 5686 : *str++ = '-';
4230 5686 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4231 5686 : break;
4232 :
4233 0 : case USE_SQL_DATES:
4234 : /* compatible with Oracle/Ingres date formats */
4235 0 : if (DateOrder == DATEORDER_DMY)
4236 : {
4237 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4238 0 : *str++ = '/';
4239 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4240 : }
4241 : else
4242 : {
4243 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4244 0 : *str++ = '/';
4245 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4246 : }
4247 0 : *str++ = '/';
4248 0 : str = pg_ultostr_zeropad(str,
4249 0 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4250 0 : break;
4251 :
4252 8 : case USE_GERMAN_DATES:
4253 : /* German-style date format */
4254 8 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4255 8 : *str++ = '.';
4256 8 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4257 8 : *str++ = '.';
4258 8 : str = pg_ultostr_zeropad(str,
4259 8 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4260 8 : break;
4261 :
4262 7242 : case USE_POSTGRES_DATES:
4263 : default:
4264 : /* traditional date-only style for Postgres */
4265 7242 : if (DateOrder == DATEORDER_DMY)
4266 : {
4267 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4268 0 : *str++ = '-';
4269 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4270 : }
4271 : else
4272 : {
4273 7242 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4274 7242 : *str++ = '-';
4275 7242 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4276 : }
4277 7242 : *str++ = '-';
4278 7242 : str = pg_ultostr_zeropad(str,
4279 7242 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4280 7242 : break;
4281 : }
4282 :
4283 12936 : if (tm->tm_year <= 0)
4284 : {
4285 80 : memcpy(str, " BC", 3); /* Don't copy NUL */
4286 80 : str += 3;
4287 : }
4288 12936 : *str = '\0';
4289 12936 : }
4290 :
4291 :
4292 : /* EncodeTimeOnly()
4293 : * Encode time fields only.
4294 : *
4295 : * tm and fsec are the value to encode, print_tz determines whether to include
4296 : * a time zone (the difference between time and timetz types), tz is the
4297 : * numeric time zone offset, style is the date style, str is where to write the
4298 : * output.
4299 : */
4300 : void
4301 12438 : EncodeTimeOnly(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, int style, char *str)
4302 : {
4303 12438 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4304 12438 : *str++ = ':';
4305 12438 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4306 12438 : *str++ = ':';
4307 12438 : str = AppendSeconds(str, tm->tm_sec, fsec, MAX_TIME_PRECISION, true);
4308 12438 : if (print_tz)
4309 6650 : str = EncodeTimezone(str, tz, style);
4310 12438 : *str = '\0';
4311 12438 : }
4312 :
4313 :
4314 : /* EncodeDateTime()
4315 : * Encode date and time interpreted as local time.
4316 : *
4317 : * tm and fsec are the value to encode, print_tz determines whether to include
4318 : * a time zone (the difference between timestamp and timestamptz types), tz is
4319 : * the numeric time zone offset, tzn is the textual time zone, which if
4320 : * specified will be used instead of tz by some styles, style is the date
4321 : * style, str is where to write the output.
4322 : *
4323 : * Supported date styles:
4324 : * Postgres - day mon hh:mm:ss yyyy tz
4325 : * SQL - mm/dd/yyyy hh:mm:ss.ss tz
4326 : * ISO - yyyy-mm-dd hh:mm:ss+/-tz
4327 : * German - dd.mm.yyyy hh:mm:ss tz
4328 : * XSD - yyyy-mm-ddThh:mm:ss.ss+/-tz
4329 : */
4330 : void
4331 113548 : EncodeDateTime(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, const char *tzn, int style, char *str)
4332 : {
4333 : int day;
4334 :
4335 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4336 :
4337 : /*
4338 : * Negative tm_isdst means we have no valid time zone translation.
4339 : */
4340 113548 : if (tm->tm_isdst < 0)
4341 41850 : print_tz = false;
4342 :
4343 113548 : switch (style)
4344 : {
4345 83830 : case USE_ISO_DATES:
4346 : case USE_XSD_DATES:
4347 : /* Compatible with ISO-8601 date formats */
4348 83830 : str = pg_ultostr_zeropad(str,
4349 83830 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4350 83830 : *str++ = '-';
4351 83830 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4352 83830 : *str++ = '-';
4353 83830 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4354 83830 : *str++ = (style == USE_ISO_DATES) ? ' ' : 'T';
4355 83830 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4356 83830 : *str++ = ':';
4357 83830 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4358 83830 : *str++ = ':';
4359 83830 : str = AppendTimestampSeconds(str, tm, fsec);
4360 83830 : if (print_tz)
4361 53102 : str = EncodeTimezone(str, tz, style);
4362 83830 : break;
4363 :
4364 780 : case USE_SQL_DATES:
4365 : /* Compatible with Oracle/Ingres date formats */
4366 780 : if (DateOrder == DATEORDER_DMY)
4367 : {
4368 384 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4369 384 : *str++ = '/';
4370 384 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4371 : }
4372 : else
4373 : {
4374 396 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4375 396 : *str++ = '/';
4376 396 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4377 : }
4378 780 : *str++ = '/';
4379 780 : str = pg_ultostr_zeropad(str,
4380 780 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4381 780 : *str++ = ' ';
4382 780 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4383 780 : *str++ = ':';
4384 780 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4385 780 : *str++ = ':';
4386 780 : str = AppendTimestampSeconds(str, tm, fsec);
4387 :
4388 : /*
4389 : * Note: the uses of %.*s in this function would be risky if the
4390 : * timezone names ever contain non-ASCII characters, since we are
4391 : * not being careful to do encoding-aware clipping. However, all
4392 : * TZ abbreviations in the IANA database are plain ASCII.
4393 : */
4394 780 : if (print_tz)
4395 : {
4396 18 : if (tzn)
4397 : {
4398 18 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4399 18 : str += strlen(str);
4400 : }
4401 : else
4402 0 : str = EncodeTimezone(str, tz, style);
4403 : }
4404 780 : break;
4405 :
4406 24 : case USE_GERMAN_DATES:
4407 : /* German variant on European style */
4408 24 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4409 24 : *str++ = '.';
4410 24 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4411 24 : *str++ = '.';
4412 24 : str = pg_ultostr_zeropad(str,
4413 24 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4414 24 : *str++ = ' ';
4415 24 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4416 24 : *str++ = ':';
4417 24 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4418 24 : *str++ = ':';
4419 24 : str = AppendTimestampSeconds(str, tm, fsec);
4420 :
4421 24 : if (print_tz)
4422 : {
4423 24 : if (tzn)
4424 : {
4425 24 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4426 24 : str += strlen(str);
4427 : }
4428 : else
4429 0 : str = EncodeTimezone(str, tz, style);
4430 : }
4431 24 : break;
4432 :
4433 28914 : case USE_POSTGRES_DATES:
4434 : default:
4435 : /* Backward-compatible with traditional Postgres abstime dates */
4436 28914 : day = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday);
4437 28914 : tm->tm_wday = j2day(day);
4438 28914 : memcpy(str, days[tm->tm_wday], 3);
4439 28914 : str += 3;
4440 28914 : *str++ = ' ';
4441 28914 : if (DateOrder == DATEORDER_DMY)
4442 : {
4443 398 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4444 398 : *str++ = ' ';
4445 398 : memcpy(str, months[tm->tm_mon - 1], 3);
4446 398 : str += 3;
4447 : }
4448 : else
4449 : {
4450 28516 : memcpy(str, months[tm->tm_mon - 1], 3);
4451 28516 : str += 3;
4452 28516 : *str++ = ' ';
4453 28516 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4454 : }
4455 28914 : *str++ = ' ';
4456 28914 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4457 28914 : *str++ = ':';
4458 28914 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4459 28914 : *str++ = ':';
4460 28914 : str = AppendTimestampSeconds(str, tm, fsec);
4461 28914 : *str++ = ' ';
4462 28914 : str = pg_ultostr_zeropad(str,
4463 28914 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4464 :
4465 28914 : if (print_tz)
4466 : {
4467 18554 : if (tzn)
4468 : {
4469 18554 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4470 18554 : str += strlen(str);
4471 : }
4472 : else
4473 : {
4474 : /*
4475 : * We have a time zone, but no string version. Use the
4476 : * numeric form, but be sure to include a leading space to
4477 : * avoid formatting something which would be rejected by
4478 : * the date/time parser later. - thomas 2001-10-19
4479 : */
4480 0 : *str++ = ' ';
4481 0 : str = EncodeTimezone(str, tz, style);
4482 : }
4483 : }
4484 28914 : break;
4485 : }
4486 :
4487 113548 : if (tm->tm_year <= 0)
4488 : {
4489 274 : memcpy(str, " BC", 3); /* Don't copy NUL */
4490 274 : str += 3;
4491 : }
4492 113548 : *str = '\0';
4493 113548 : }
4494 :
4495 :
4496 : /*
4497 : * Helper functions to avoid duplicated code in EncodeInterval.
4498 : */
4499 :
4500 : /* Append an ISO-8601-style interval field, but only if value isn't zero */
4501 : static char *
4502 210 : AddISO8601IntPart(char *cp, int64 value, char units)
4503 : {
4504 210 : if (value == 0)
4505 54 : return cp;
4506 156 : sprintf(cp, "%lld%c", (long long) value, units);
4507 156 : return cp + strlen(cp);
4508 : }
4509 :
4510 : /* Append a postgres-style interval field, but only if value isn't zero */
4511 : static char *
4512 14436 : AddPostgresIntPart(char *cp, int64 value, const char *units,
4513 : bool *is_zero, bool *is_before)
4514 : {
4515 14436 : if (value == 0)
4516 7998 : return cp;
4517 19314 : sprintf(cp, "%s%s%lld %s%s",
4518 6438 : (!*is_zero) ? " " : "",
4519 6438 : (*is_before && value > 0) ? "+" : "",
4520 : (long long) value,
4521 : units,
4522 : (value != 1) ? "s" : "");
4523 :
4524 : /*
4525 : * Each nonzero field sets is_before for (only) the next one. This is a
4526 : * tad bizarre but it's how it worked before...
4527 : */
4528 6438 : *is_before = (value < 0);
4529 6438 : *is_zero = false;
4530 6438 : return cp + strlen(cp);
4531 : }
4532 :
4533 : /* Append a verbose-style interval field, but only if value isn't zero */
4534 : static char *
4535 42190 : AddVerboseIntPart(char *cp, int64 value, const char *units,
4536 : bool *is_zero, bool *is_before)
4537 : {
4538 42190 : if (value == 0)
4539 27980 : return cp;
4540 : /* first nonzero value sets is_before */
4541 14210 : if (*is_zero)
4542 : {
4543 7820 : *is_before = (value < 0);
4544 7820 : value = i64abs(value);
4545 : }
4546 6390 : else if (*is_before)
4547 1332 : value = -value;
4548 14210 : sprintf(cp, " %lld %s%s", (long long) value, units, (value == 1) ? "" : "s");
4549 14210 : *is_zero = false;
4550 14210 : return cp + strlen(cp);
4551 : }
4552 :
4553 :
4554 : /* EncodeInterval()
4555 : * Interpret time structure as a delta time and convert to string.
4556 : *
4557 : * Support "traditional Postgres" and ISO-8601 styles.
4558 : * Actually, afaik ISO does not address time interval formatting,
4559 : * but this looks similar to the spec for absolute date/time.
4560 : * - thomas 1998-04-30
4561 : *
4562 : * Actually, afaik, ISO 8601 does specify formats for "time
4563 : * intervals...[of the]...format with time-unit designators", which
4564 : * are pretty ugly. The format looks something like
4565 : * P1Y1M1DT1H1M1.12345S
4566 : * but useful for exchanging data with computers instead of humans.
4567 : * - ron 2003-07-14
4568 : *
4569 : * And ISO's SQL 2008 standard specifies standards for
4570 : * "year-month literal"s (that look like '2-3') and
4571 : * "day-time literal"s (that look like ('4 5:6:7')
4572 : */
4573 : void
4574 13424 : EncodeInterval(struct pg_itm *itm, int style, char *str)
4575 : {
4576 13424 : char *cp = str;
4577 13424 : int year = itm->tm_year;
4578 13424 : int mon = itm->tm_mon;
4579 13424 : int64 mday = itm->tm_mday; /* tm_mday could be INT_MIN */
4580 13424 : int64 hour = itm->tm_hour;
4581 13424 : int min = itm->tm_min;
4582 13424 : int sec = itm->tm_sec;
4583 13424 : int fsec = itm->tm_usec;
4584 13424 : bool is_before = false;
4585 13424 : bool is_zero = true;
4586 :
4587 : /*
4588 : * The sign of year and month are guaranteed to match, since they are
4589 : * stored internally as "month". But we'll need to check for is_before and
4590 : * is_zero when determining the signs of day and hour/minute/seconds
4591 : * fields.
4592 : */
4593 13424 : switch (style)
4594 : {
4595 : /* SQL Standard interval format */
4596 126 : case INTSTYLE_SQL_STANDARD:
4597 : {
4598 96 : bool has_negative = year < 0 || mon < 0 ||
4599 66 : mday < 0 || hour < 0 ||
4600 222 : min < 0 || sec < 0 || fsec < 0;
4601 102 : bool has_positive = year > 0 || mon > 0 ||
4602 66 : mday > 0 || hour > 0 ||
4603 228 : min > 0 || sec > 0 || fsec > 0;
4604 126 : bool has_year_month = year != 0 || mon != 0;
4605 42 : bool has_day_time = mday != 0 || hour != 0 ||
4606 168 : min != 0 || sec != 0 || fsec != 0;
4607 126 : bool has_day = mday != 0;
4608 222 : bool sql_standard_value = !(has_negative && has_positive) &&
4609 96 : !(has_year_month && has_day_time);
4610 :
4611 : /*
4612 : * SQL Standard wants only 1 "<sign>" preceding the whole
4613 : * interval ... but can't do that if mixed signs.
4614 : */
4615 126 : if (has_negative && sql_standard_value)
4616 : {
4617 30 : *cp++ = '-';
4618 30 : year = -year;
4619 30 : mon = -mon;
4620 30 : mday = -mday;
4621 30 : hour = -hour;
4622 30 : min = -min;
4623 30 : sec = -sec;
4624 30 : fsec = -fsec;
4625 : }
4626 :
4627 126 : if (!has_negative && !has_positive)
4628 : {
4629 12 : sprintf(cp, "0");
4630 : }
4631 114 : else if (!sql_standard_value)
4632 : {
4633 : /*
4634 : * For non sql-standard interval values, force outputting
4635 : * the signs to avoid ambiguities with intervals with
4636 : * mixed sign components.
4637 : */
4638 54 : char year_sign = (year < 0 || mon < 0) ? '-' : '+';
4639 54 : char day_sign = (mday < 0) ? '-' : '+';
4640 78 : char sec_sign = (hour < 0 || min < 0 ||
4641 24 : sec < 0 || fsec < 0) ? '-' : '+';
4642 :
4643 54 : sprintf(cp, "%c%d-%d %c%lld %c%lld:%02d:",
4644 : year_sign, abs(year), abs(mon),
4645 54 : day_sign, (long long) i64abs(mday),
4646 54 : sec_sign, (long long) i64abs(hour), abs(min));
4647 54 : cp += strlen(cp);
4648 54 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4649 54 : *cp = '\0';
4650 : }
4651 60 : else if (has_year_month)
4652 : {
4653 18 : sprintf(cp, "%d-%d", year, mon);
4654 : }
4655 42 : else if (has_day)
4656 : {
4657 30 : sprintf(cp, "%lld %lld:%02d:",
4658 : (long long) mday, (long long) hour, min);
4659 30 : cp += strlen(cp);
4660 30 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4661 30 : *cp = '\0';
4662 : }
4663 : else
4664 : {
4665 12 : sprintf(cp, "%lld:%02d:", (long long) hour, min);
4666 12 : cp += strlen(cp);
4667 12 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4668 12 : *cp = '\0';
4669 : }
4670 : }
4671 126 : break;
4672 :
4673 : /* ISO 8601 "time-intervals by duration only" */
4674 48 : case INTSTYLE_ISO_8601:
4675 : /* special-case zero to avoid printing nothing */
4676 48 : if (year == 0 && mon == 0 && mday == 0 &&
4677 6 : hour == 0 && min == 0 && sec == 0 && fsec == 0)
4678 : {
4679 6 : sprintf(cp, "PT0S");
4680 6 : break;
4681 : }
4682 42 : *cp++ = 'P';
4683 42 : cp = AddISO8601IntPart(cp, year, 'Y');
4684 42 : cp = AddISO8601IntPart(cp, mon, 'M');
4685 42 : cp = AddISO8601IntPart(cp, mday, 'D');
4686 42 : if (hour != 0 || min != 0 || sec != 0 || fsec != 0)
4687 36 : *cp++ = 'T';
4688 42 : cp = AddISO8601IntPart(cp, hour, 'H');
4689 42 : cp = AddISO8601IntPart(cp, min, 'M');
4690 42 : if (sec != 0 || fsec != 0)
4691 : {
4692 36 : if (sec < 0 || fsec < 0)
4693 12 : *cp++ = '-';
4694 36 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4695 36 : *cp++ = 'S';
4696 36 : *cp++ = '\0';
4697 : }
4698 42 : break;
4699 :
4700 : /* Compatible with postgresql < 8.4 when DateStyle = 'iso' */
4701 4812 : case INTSTYLE_POSTGRES:
4702 4812 : cp = AddPostgresIntPart(cp, year, "year", &is_zero, &is_before);
4703 :
4704 : /*
4705 : * Ideally we should spell out "month" like we do for "year" and
4706 : * "day". However, for backward compatibility, we can't easily
4707 : * fix this. bjm 2011-05-24
4708 : */
4709 4812 : cp = AddPostgresIntPart(cp, mon, "mon", &is_zero, &is_before);
4710 4812 : cp = AddPostgresIntPart(cp, mday, "day", &is_zero, &is_before);
4711 4812 : if (is_zero || hour != 0 || min != 0 || sec != 0 || fsec != 0)
4712 : {
4713 3868 : bool minus = (hour < 0 || min < 0 || sec < 0 || fsec < 0);
4714 :
4715 7736 : sprintf(cp, "%s%s%02lld:%02d:",
4716 3868 : is_zero ? "" : " ",
4717 3654 : (minus ? "-" : (is_before ? "+" : "")),
4718 3868 : (long long) i64abs(hour), abs(min));
4719 3868 : cp += strlen(cp);
4720 3868 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4721 3868 : *cp = '\0';
4722 : }
4723 4812 : break;
4724 :
4725 : /* Compatible with postgresql < 8.4 when DateStyle != 'iso' */
4726 8438 : case INTSTYLE_POSTGRES_VERBOSE:
4727 : default:
4728 8438 : strcpy(cp, "@");
4729 8438 : cp++;
4730 8438 : cp = AddVerboseIntPart(cp, year, "year", &is_zero, &is_before);
4731 8438 : cp = AddVerboseIntPart(cp, mon, "mon", &is_zero, &is_before);
4732 8438 : cp = AddVerboseIntPart(cp, mday, "day", &is_zero, &is_before);
4733 8438 : cp = AddVerboseIntPart(cp, hour, "hour", &is_zero, &is_before);
4734 8438 : cp = AddVerboseIntPart(cp, min, "min", &is_zero, &is_before);
4735 8438 : if (sec != 0 || fsec != 0)
4736 : {
4737 3264 : *cp++ = ' ';
4738 3264 : if (sec < 0 || (sec == 0 && fsec < 0))
4739 : {
4740 1008 : if (is_zero)
4741 342 : is_before = true;
4742 666 : else if (!is_before)
4743 6 : *cp++ = '-';
4744 : }
4745 2256 : else if (is_before)
4746 12 : *cp++ = '-';
4747 3264 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4748 : /* We output "ago", not negatives, so use abs(). */
4749 3264 : sprintf(cp, " sec%s",
4750 3264 : (abs(sec) != 1 || fsec != 0) ? "s" : "");
4751 3264 : is_zero = false;
4752 : }
4753 : /* identically zero? then put in a unitless zero... */
4754 8438 : if (is_zero)
4755 230 : strcat(cp, " 0");
4756 8438 : if (is_before)
4757 1354 : strcat(cp, " ago");
4758 8438 : break;
4759 : }
4760 13424 : }
4761 :
4762 :
4763 : /*
4764 : * We've been burnt by stupid errors in the ordering of the datetkn tables
4765 : * once too often. Arrange to check them during postmaster start.
4766 : */
4767 : static bool
4768 2872 : CheckDateTokenTable(const char *tablename, const datetkn *base, int nel)
4769 : {
4770 2872 : bool ok = true;
4771 : int i;
4772 :
4773 193860 : for (i = 0; i < nel; i++)
4774 : {
4775 : /* check for token strings that don't fit */
4776 190988 : if (strlen(base[i].token) > TOKMAXLEN)
4777 : {
4778 : /* %.*s is safe since all our tokens are ASCII */
4779 0 : elog(LOG, "token too long in %s table: \"%.*s\"",
4780 : tablename,
4781 : TOKMAXLEN + 1, base[i].token);
4782 0 : ok = false;
4783 0 : break; /* don't risk applying strcmp */
4784 : }
4785 : /* check for out of order */
4786 190988 : if (i > 0 &&
4787 188116 : strcmp(base[i - 1].token, base[i].token) >= 0)
4788 : {
4789 0 : elog(LOG, "ordering error in %s table: \"%s\" >= \"%s\"",
4790 : tablename,
4791 : base[i - 1].token,
4792 : base[i].token);
4793 0 : ok = false;
4794 : }
4795 : }
4796 2872 : return ok;
4797 : }
4798 :
4799 : bool
4800 1436 : CheckDateTokenTables(void)
4801 : {
4802 1436 : bool ok = true;
4803 :
4804 : Assert(UNIX_EPOCH_JDATE == date2j(1970, 1, 1));
4805 : Assert(POSTGRES_EPOCH_JDATE == date2j(2000, 1, 1));
4806 :
4807 1436 : ok &= CheckDateTokenTable("datetktbl", datetktbl, szdatetktbl);
4808 1436 : ok &= CheckDateTokenTable("deltatktbl", deltatktbl, szdeltatktbl);
4809 1436 : return ok;
4810 : }
4811 :
4812 : /*
4813 : * Common code for temporal prosupport functions: simplify, if possible,
4814 : * a call to a temporal type's length-coercion function.
4815 : *
4816 : * Types time, timetz, timestamp and timestamptz each have a range of allowed
4817 : * precisions. An unspecified precision is rigorously equivalent to the
4818 : * highest specifiable precision. We can replace the function call with a
4819 : * no-op RelabelType if it is coercing to the same or higher precision as the
4820 : * input is known to have.
4821 : *
4822 : * The input Node is always a FuncExpr, but to reduce the #include footprint
4823 : * of datetime.h, we declare it as Node *.
4824 : *
4825 : * Note: timestamp_scale throws an error when the typmod is out of range, but
4826 : * we can't get there from a cast: our typmodin will have caught it already.
4827 : */
4828 : Node *
4829 24 : TemporalSimplify(int32 max_precis, Node *node)
4830 : {
4831 24 : FuncExpr *expr = castNode(FuncExpr, node);
4832 24 : Node *ret = NULL;
4833 : Node *typmod;
4834 :
4835 : Assert(list_length(expr->args) >= 2);
4836 :
4837 24 : typmod = (Node *) lsecond(expr->args);
4838 :
4839 24 : if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
4840 : {
4841 24 : Node *source = (Node *) linitial(expr->args);
4842 24 : int32 old_precis = exprTypmod(source);
4843 24 : int32 new_precis = DatumGetInt32(((Const *) typmod)->constvalue);
4844 :
4845 24 : if (new_precis < 0 || new_precis == max_precis ||
4846 0 : (old_precis >= 0 && new_precis >= old_precis))
4847 0 : ret = relabel_to_typmod(source, new_precis);
4848 : }
4849 :
4850 24 : return ret;
4851 : }
4852 :
4853 : /*
4854 : * This function gets called during timezone config file load or reload
4855 : * to create the final array of timezone tokens. The argument array
4856 : * is already sorted in name order.
4857 : *
4858 : * The result is a TimeZoneAbbrevTable (which must be a single guc_malloc'd
4859 : * chunk) or NULL on alloc failure. No other error conditions are defined.
4860 : */
4861 : TimeZoneAbbrevTable *
4862 11866 : ConvertTimeZoneAbbrevs(struct tzEntry *abbrevs, int n)
4863 : {
4864 : TimeZoneAbbrevTable *tbl;
4865 : Size tbl_size;
4866 : int i;
4867 :
4868 : /* Space for fixed fields and datetkn array */
4869 11866 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
4870 11866 : n * sizeof(datetkn);
4871 11866 : tbl_size = MAXALIGN(tbl_size);
4872 : /* Count up space for dynamic abbreviations */
4873 2325748 : for (i = 0; i < n; i++)
4874 : {
4875 2313882 : struct tzEntry *abbr = abbrevs + i;
4876 :
4877 2313882 : if (abbr->zone != NULL)
4878 : {
4879 : Size dsize;
4880 :
4881 593294 : dsize = offsetof(DynamicZoneAbbrev, zone) +
4882 593294 : strlen(abbr->zone) + 1;
4883 593294 : tbl_size += MAXALIGN(dsize);
4884 : }
4885 : }
4886 :
4887 : /* Alloc the result ... */
4888 11866 : tbl = guc_malloc(LOG, tbl_size);
4889 11866 : if (!tbl)
4890 0 : return NULL;
4891 :
4892 : /* ... and fill it in */
4893 11866 : tbl->tblsize = tbl_size;
4894 11866 : tbl->numabbrevs = n;
4895 : /* in this loop, tbl_size reprises the space calculation above */
4896 11866 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
4897 11866 : n * sizeof(datetkn);
4898 11866 : tbl_size = MAXALIGN(tbl_size);
4899 2325748 : for (i = 0; i < n; i++)
4900 : {
4901 2313882 : struct tzEntry *abbr = abbrevs + i;
4902 2313882 : datetkn *dtoken = tbl->abbrevs + i;
4903 :
4904 : /* use strlcpy to truncate name if necessary */
4905 2313882 : strlcpy(dtoken->token, abbr->abbrev, TOKMAXLEN + 1);
4906 2313882 : if (abbr->zone != NULL)
4907 : {
4908 : /* Allocate a DynamicZoneAbbrev for this abbreviation */
4909 : DynamicZoneAbbrev *dtza;
4910 : Size dsize;
4911 :
4912 593294 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tbl_size);
4913 593294 : dtza->tz = NULL;
4914 593294 : strcpy(dtza->zone, abbr->zone);
4915 :
4916 593294 : dtoken->type = DYNTZ;
4917 : /* value is offset from table start to DynamicZoneAbbrev */
4918 593294 : dtoken->value = (int32) tbl_size;
4919 :
4920 593294 : dsize = offsetof(DynamicZoneAbbrev, zone) +
4921 593294 : strlen(abbr->zone) + 1;
4922 593294 : tbl_size += MAXALIGN(dsize);
4923 : }
4924 : else
4925 : {
4926 1720588 : dtoken->type = abbr->is_dst ? DTZ : TZ;
4927 1720588 : dtoken->value = abbr->offset;
4928 : }
4929 : }
4930 :
4931 : /* Assert the two loops above agreed on size calculations */
4932 : Assert(tbl->tblsize == tbl_size);
4933 :
4934 : /* Check the ordering, if testing */
4935 : Assert(CheckDateTokenTable("timezone abbreviations", tbl->abbrevs, n));
4936 :
4937 11866 : return tbl;
4938 : }
4939 :
4940 : /*
4941 : * Install a TimeZoneAbbrevTable as the active table.
4942 : *
4943 : * Caller is responsible that the passed table doesn't go away while in use.
4944 : */
4945 : void
4946 11688 : InstallTimeZoneAbbrevs(TimeZoneAbbrevTable *tbl)
4947 : {
4948 11688 : zoneabbrevtbl = tbl;
4949 : /* reset abbrevcache, which may contain pointers into old table */
4950 11688 : memset(abbrevcache, 0, sizeof(abbrevcache));
4951 11688 : }
4952 :
4953 : /*
4954 : * Helper subroutine to locate pg_tz timezone for a dynamic abbreviation.
4955 : *
4956 : * On failure, returns NULL and fills *extra for a DTERR_BAD_ZONE_ABBREV error.
4957 : */
4958 : static pg_tz *
4959 1164 : FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp,
4960 : DateTimeErrorExtra *extra)
4961 : {
4962 : DynamicZoneAbbrev *dtza;
4963 :
4964 : /* Just some sanity checks to prevent indexing off into nowhere */
4965 : Assert(tp->type == DYNTZ);
4966 : Assert(tp->value > 0 && tp->value < tbl->tblsize);
4967 :
4968 1164 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tp->value);
4969 :
4970 : /* Look up the underlying zone if we haven't already */
4971 1164 : if (dtza->tz == NULL)
4972 : {
4973 918 : dtza->tz = pg_tzset(dtza->zone);
4974 918 : if (dtza->tz == NULL)
4975 : {
4976 : /* Ooops, bogus zone name in config file entry */
4977 0 : extra->dtee_timezone = dtza->zone;
4978 0 : extra->dtee_abbrev = tp->token;
4979 : }
4980 : }
4981 1164 : return dtza->tz;
4982 : }
4983 :
4984 :
4985 : /*
4986 : * This set-returning function reads all the available time zone abbreviations
4987 : * and returns a set of (abbrev, utc_offset, is_dst).
4988 : */
4989 : Datum
4990 3540 : pg_timezone_abbrevs(PG_FUNCTION_ARGS)
4991 : {
4992 : FuncCallContext *funcctx;
4993 : int *pindex;
4994 : Datum result;
4995 : HeapTuple tuple;
4996 : Datum values[3];
4997 3540 : bool nulls[3] = {0};
4998 : const datetkn *tp;
4999 : char buffer[TOKMAXLEN + 1];
5000 : int gmtoffset;
5001 : bool is_dst;
5002 : unsigned char *p;
5003 : struct pg_itm_in itm_in;
5004 : Interval *resInterval;
5005 :
5006 : /* stuff done only on the first call of the function */
5007 3540 : if (SRF_IS_FIRSTCALL())
5008 : {
5009 : TupleDesc tupdesc;
5010 : MemoryContext oldcontext;
5011 :
5012 : /* create a function context for cross-call persistence */
5013 18 : funcctx = SRF_FIRSTCALL_INIT();
5014 :
5015 : /*
5016 : * switch to memory context appropriate for multiple function calls
5017 : */
5018 18 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
5019 :
5020 : /* allocate memory for user context */
5021 18 : pindex = (int *) palloc(sizeof(int));
5022 18 : *pindex = 0;
5023 18 : funcctx->user_fctx = (void *) pindex;
5024 :
5025 18 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
5026 0 : elog(ERROR, "return type must be a row type");
5027 18 : funcctx->tuple_desc = tupdesc;
5028 :
5029 18 : MemoryContextSwitchTo(oldcontext);
5030 : }
5031 :
5032 : /* stuff done on every call of the function */
5033 3540 : funcctx = SRF_PERCALL_SETUP();
5034 3540 : pindex = (int *) funcctx->user_fctx;
5035 :
5036 3540 : if (zoneabbrevtbl == NULL ||
5037 3540 : *pindex >= zoneabbrevtbl->numabbrevs)
5038 18 : SRF_RETURN_DONE(funcctx);
5039 :
5040 3522 : tp = zoneabbrevtbl->abbrevs + *pindex;
5041 :
5042 3522 : switch (tp->type)
5043 : {
5044 1764 : case TZ:
5045 1764 : gmtoffset = tp->value;
5046 1764 : is_dst = false;
5047 1764 : break;
5048 864 : case DTZ:
5049 864 : gmtoffset = tp->value;
5050 864 : is_dst = true;
5051 864 : break;
5052 894 : case DYNTZ:
5053 : {
5054 : /* Determine the current meaning of the abbrev */
5055 : pg_tz *tzp;
5056 : DateTimeErrorExtra extra;
5057 : TimestampTz now;
5058 : int isdst;
5059 :
5060 894 : tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp, &extra);
5061 894 : if (tzp == NULL)
5062 0 : DateTimeParseError(DTERR_BAD_ZONE_ABBREV, &extra,
5063 : NULL, NULL, NULL);
5064 894 : now = GetCurrentTransactionStartTimestamp();
5065 1788 : gmtoffset = -DetermineTimeZoneAbbrevOffsetTS(now,
5066 894 : tp->token,
5067 : tzp,
5068 : &isdst);
5069 894 : is_dst = (bool) isdst;
5070 894 : break;
5071 : }
5072 0 : default:
5073 0 : elog(ERROR, "unrecognized timezone type %d", (int) tp->type);
5074 : gmtoffset = 0; /* keep compiler quiet */
5075 : is_dst = false;
5076 : break;
5077 : }
5078 :
5079 : /*
5080 : * Convert name to text, using upcasing conversion that is the inverse of
5081 : * what ParseDateTime() uses.
5082 : */
5083 3522 : strlcpy(buffer, tp->token, sizeof(buffer));
5084 16320 : for (p = (unsigned char *) buffer; *p; p++)
5085 12798 : *p = pg_toupper(*p);
5086 :
5087 3522 : values[0] = CStringGetTextDatum(buffer);
5088 :
5089 : /* Convert offset (in seconds) to an interval; can't overflow */
5090 14088 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5091 3522 : itm_in.tm_usec = (int64) gmtoffset * USECS_PER_SEC;
5092 3522 : resInterval = (Interval *) palloc(sizeof(Interval));
5093 3522 : (void) itmin2interval(&itm_in, resInterval);
5094 3522 : values[1] = IntervalPGetDatum(resInterval);
5095 :
5096 3522 : values[2] = BoolGetDatum(is_dst);
5097 :
5098 3522 : (*pindex)++;
5099 :
5100 3522 : tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
5101 3522 : result = HeapTupleGetDatum(tuple);
5102 :
5103 3522 : SRF_RETURN_NEXT(funcctx, result);
5104 : }
5105 :
5106 : /*
5107 : * This set-returning function reads all the available full time zones
5108 : * and returns a set of (name, abbrev, utc_offset, is_dst).
5109 : */
5110 : Datum
5111 16 : pg_timezone_names(PG_FUNCTION_ARGS)
5112 : {
5113 16 : ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
5114 : pg_tzenum *tzenum;
5115 : pg_tz *tz;
5116 : Datum values[4];
5117 16 : bool nulls[4] = {0};
5118 : int tzoff;
5119 : struct pg_tm tm;
5120 : fsec_t fsec;
5121 : const char *tzn;
5122 : Interval *resInterval;
5123 : struct pg_itm_in itm_in;
5124 :
5125 16 : InitMaterializedSRF(fcinfo, 0);
5126 :
5127 : /* initialize timezone scanning code */
5128 16 : tzenum = pg_tzenumerate_start();
5129 :
5130 : /* search for another zone to display */
5131 : for (;;)
5132 : {
5133 9568 : tz = pg_tzenumerate_next(tzenum);
5134 9568 : if (!tz)
5135 16 : break;
5136 :
5137 : /* Convert now() to local time in this zone */
5138 9552 : if (timestamp2tm(GetCurrentTransactionStartTimestamp(),
5139 : &tzoff, &tm, &fsec, &tzn, tz) != 0)
5140 0 : continue; /* ignore if conversion fails */
5141 :
5142 : /*
5143 : * IANA's rather silly "Factory" time zone used to emit ridiculously
5144 : * long "abbreviations" such as "Local time zone must be set--see zic
5145 : * manual page" or "Local time zone must be set--use tzsetup". While
5146 : * modern versions of tzdb emit the much saner "-00", it seems some
5147 : * benighted packagers are hacking the IANA data so that it continues
5148 : * to produce these strings. To prevent producing a weirdly wide
5149 : * abbrev column, reject ridiculously long abbreviations.
5150 : */
5151 9552 : if (tzn && strlen(tzn) > 31)
5152 0 : continue;
5153 :
5154 9552 : values[0] = CStringGetTextDatum(pg_get_timezone_name(tz));
5155 9552 : values[1] = CStringGetTextDatum(tzn ? tzn : "");
5156 :
5157 : /* Convert tzoff to an interval; can't overflow */
5158 38208 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5159 9552 : itm_in.tm_usec = (int64) -tzoff * USECS_PER_SEC;
5160 9552 : resInterval = (Interval *) palloc(sizeof(Interval));
5161 9552 : (void) itmin2interval(&itm_in, resInterval);
5162 9552 : values[2] = IntervalPGetDatum(resInterval);
5163 :
5164 9552 : values[3] = BoolGetDatum(tm.tm_isdst > 0);
5165 :
5166 9552 : tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls);
5167 : }
5168 :
5169 16 : pg_tzenumerate_end(tzenum);
5170 16 : return (Datum) 0;
5171 : }
|
