Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * datetime.c
4 : * Support functions for date/time types.
5 : *
6 : * Portions Copyright (c) 1996-2025, 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 : /* Cache for results of timezone abbreviation lookups */
263 :
264 : typedef struct TzAbbrevCache
265 : {
266 : char abbrev[TOKMAXLEN + 1]; /* always NUL-terminated */
267 : char ftype; /* TZ, DTZ, or DYNTZ */
268 : int offset; /* GMT offset, if fixed-offset */
269 : pg_tz *tz; /* relevant zone, if variable-offset */
270 : } TzAbbrevCache;
271 :
272 : static TzAbbrevCache tzabbrevcache[MAXDATEFIELDS];
273 :
274 :
275 : /*
276 : * Calendar time to Julian date conversions.
277 : * Julian date is commonly used in astronomical applications,
278 : * since it is numerically accurate and computationally simple.
279 : * The algorithms here will accurately convert between Julian day
280 : * and calendar date for all non-negative Julian days
281 : * (i.e. from Nov 24, -4713 on).
282 : *
283 : * Rewritten to eliminate overflow problems. This now allows the
284 : * routines to work correctly for all Julian day counts from
285 : * 0 to 2147483647 (Nov 24, -4713 to Jun 3, 5874898) assuming
286 : * a 32-bit integer. Longer types should also work to the limits
287 : * of their precision.
288 : *
289 : * Actually, date2j() will work sanely, in the sense of producing
290 : * valid negative Julian dates, significantly before Nov 24, -4713.
291 : * We rely on it to do so back to Nov 1, -4713; see IS_VALID_JULIAN()
292 : * and associated commentary in timestamp.h.
293 : */
294 :
295 : int
296 460328 : date2j(int year, int month, int day)
297 : {
298 : int julian;
299 : int century;
300 :
301 460328 : if (month > 2)
302 : {
303 299934 : month += 1;
304 299934 : year += 4800;
305 : }
306 : else
307 : {
308 160394 : month += 13;
309 160394 : year += 4799;
310 : }
311 :
312 460328 : century = year / 100;
313 460328 : julian = year * 365 - 32167;
314 460328 : julian += year / 4 - century + century / 4;
315 460328 : julian += 7834 * month / 256 + day;
316 :
317 460328 : return julian;
318 : } /* date2j() */
319 :
320 : void
321 342010 : j2date(int jd, int *year, int *month, int *day)
322 : {
323 : unsigned int julian;
324 : unsigned int quad;
325 : unsigned int extra;
326 : int y;
327 :
328 342010 : julian = jd;
329 342010 : julian += 32044;
330 342010 : quad = julian / 146097;
331 342010 : extra = (julian - quad * 146097) * 4 + 3;
332 342010 : julian += 60 + quad * 3 + extra / 146097;
333 342010 : quad = julian / 1461;
334 342010 : julian -= quad * 1461;
335 342010 : y = julian * 4 / 1461;
336 684020 : julian = ((y != 0) ? ((julian + 305) % 365) : ((julian + 306) % 366))
337 342010 : + 123;
338 342010 : y += quad * 4;
339 342010 : *year = y - 4800;
340 342010 : quad = julian * 2141 / 65536;
341 342010 : *day = julian - 7834 * quad / 256;
342 342010 : *month = (quad + 10) % MONTHS_PER_YEAR + 1;
343 342010 : } /* j2date() */
344 :
345 :
346 : /*
347 : * j2day - convert Julian date to day-of-week (0..6 == Sun..Sat)
348 : *
349 : * Note: various places use the locution j2day(date - 1) to produce a
350 : * result according to the convention 0..6 = Mon..Sun. This is a bit of
351 : * a crock, but will work as long as the computation here is just a modulo.
352 : */
353 : int
354 51890 : j2day(int date)
355 : {
356 51890 : date += 1;
357 51890 : date %= 7;
358 : /* Cope if division truncates towards zero, as it probably does */
359 51890 : if (date < 0)
360 0 : date += 7;
361 :
362 51890 : return date;
363 : } /* j2day() */
364 :
365 :
366 : /*
367 : * GetCurrentDateTime()
368 : *
369 : * Get the transaction start time ("now()") broken down as a struct pg_tm,
370 : * converted according to the session timezone setting.
371 : *
372 : * This is just a convenience wrapper for GetCurrentTimeUsec, to cover the
373 : * case where caller doesn't need either fractional seconds or tz offset.
374 : */
375 : void
376 3856 : GetCurrentDateTime(struct pg_tm *tm)
377 : {
378 : fsec_t fsec;
379 :
380 3856 : GetCurrentTimeUsec(tm, &fsec, NULL);
381 3856 : }
382 :
383 : /*
384 : * GetCurrentTimeUsec()
385 : *
386 : * Get the transaction start time ("now()") broken down as a struct pg_tm,
387 : * including fractional seconds and timezone offset. The time is converted
388 : * according to the session timezone setting.
389 : *
390 : * Callers may pass tzp = NULL if they don't need the offset, but this does
391 : * not affect the conversion behavior (unlike timestamp2tm()).
392 : *
393 : * Internally, we cache the result, since this could be called many times
394 : * in a transaction, within which now() doesn't change.
395 : */
396 : void
397 4030 : GetCurrentTimeUsec(struct pg_tm *tm, fsec_t *fsec, int *tzp)
398 : {
399 4030 : TimestampTz cur_ts = GetCurrentTransactionStartTimestamp();
400 :
401 : /*
402 : * The cache key must include both current time and current timezone. By
403 : * representing the timezone by just a pointer, we're assuming that
404 : * distinct timezone settings could never have the same pointer value.
405 : * This is true by virtue of the hashtable used inside pg_tzset();
406 : * however, it might need another look if we ever allow entries in that
407 : * hash to be recycled.
408 : */
409 : static TimestampTz cache_ts = 0;
410 : static pg_tz *cache_timezone = NULL;
411 : static struct pg_tm cache_tm;
412 : static fsec_t cache_fsec;
413 : static int cache_tz;
414 :
415 4030 : if (cur_ts != cache_ts || session_timezone != cache_timezone)
416 : {
417 : /*
418 : * Make sure cache is marked invalid in case of error after partial
419 : * update within timestamp2tm.
420 : */
421 978 : cache_timezone = NULL;
422 :
423 : /*
424 : * Perform the computation, storing results into cache. We do not
425 : * really expect any error here, since current time surely ought to be
426 : * within range, but check just for sanity's sake.
427 : */
428 978 : if (timestamp2tm(cur_ts, &cache_tz, &cache_tm, &cache_fsec,
429 : NULL, session_timezone) != 0)
430 0 : ereport(ERROR,
431 : (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
432 : errmsg("timestamp out of range")));
433 :
434 : /* OK, so mark the cache valid. */
435 978 : cache_ts = cur_ts;
436 978 : cache_timezone = session_timezone;
437 : }
438 :
439 4030 : *tm = cache_tm;
440 4030 : *fsec = cache_fsec;
441 4030 : if (tzp != NULL)
442 162 : *tzp = cache_tz;
443 4030 : }
444 :
445 :
446 : /*
447 : * Append seconds and fractional seconds (if any) at *cp.
448 : *
449 : * precision is the max number of fraction digits, fillzeros says to
450 : * pad to two integral-seconds digits.
451 : *
452 : * Returns a pointer to the new end of string. No NUL terminator is put
453 : * there; callers are responsible for NUL terminating str themselves.
454 : *
455 : * Note that any sign is stripped from the input sec and fsec values.
456 : */
457 : static char *
458 151358 : AppendSeconds(char *cp, int sec, fsec_t fsec, int precision, bool fillzeros)
459 : {
460 : Assert(precision >= 0);
461 :
462 151358 : if (fillzeros)
463 148040 : cp = pg_ultostr_zeropad(cp, abs(sec), 2);
464 : else
465 3318 : cp = pg_ultostr(cp, abs(sec));
466 :
467 : /* fsec_t is just an int32 */
468 151358 : if (fsec != 0)
469 : {
470 24404 : int32 value = abs(fsec);
471 24404 : char *end = &cp[precision + 1];
472 24404 : bool gotnonzero = false;
473 :
474 24404 : *cp++ = '.';
475 :
476 : /*
477 : * Append the fractional seconds part. Note that we don't want any
478 : * trailing zeros here, so since we're building the number in reverse
479 : * we'll skip appending zeros until we've output a non-zero digit.
480 : */
481 170828 : while (precision--)
482 : {
483 146424 : int32 oldval = value;
484 : int32 remainder;
485 :
486 146424 : value /= 10;
487 146424 : remainder = oldval - value * 10;
488 :
489 : /* check if we got a non-zero */
490 146424 : if (remainder)
491 118216 : gotnonzero = true;
492 :
493 146424 : if (gotnonzero)
494 120142 : cp[precision] = '0' + remainder;
495 : else
496 26282 : end = &cp[precision];
497 : }
498 :
499 : /*
500 : * If we still have a non-zero value then precision must have not been
501 : * enough to print the number. We punt the problem to pg_ultostr(),
502 : * which will generate a correct answer in the minimum valid width.
503 : */
504 24404 : if (value)
505 0 : return pg_ultostr(cp, abs(fsec));
506 :
507 24404 : return end;
508 : }
509 : else
510 126954 : return cp;
511 : }
512 :
513 :
514 : /*
515 : * Variant of above that's specialized to timestamp case.
516 : *
517 : * Returns a pointer to the new end of string. No NUL terminator is put
518 : * there; callers are responsible for NUL terminating str themselves.
519 : */
520 : static char *
521 122452 : AppendTimestampSeconds(char *cp, struct pg_tm *tm, fsec_t fsec)
522 : {
523 122452 : return AppendSeconds(cp, tm->tm_sec, fsec, MAX_TIMESTAMP_PRECISION, true);
524 : }
525 :
526 :
527 : /*
528 : * Add val * multiplier to *sum.
529 : * Returns true if successful, false on overflow.
530 : */
531 : static bool
532 11906 : int64_multiply_add(int64 val, int64 multiplier, int64 *sum)
533 : {
534 : int64 product;
535 :
536 23752 : if (pg_mul_s64_overflow(val, multiplier, &product) ||
537 11846 : pg_add_s64_overflow(*sum, product, sum))
538 156 : return false;
539 11750 : return true;
540 : }
541 :
542 : /*
543 : * Multiply frac by scale (to produce microseconds) and add to itm_in->tm_usec.
544 : * Returns true if successful, false if itm_in overflows.
545 : */
546 : static bool
547 10938 : AdjustFractMicroseconds(double frac, int64 scale,
548 : struct pg_itm_in *itm_in)
549 : {
550 : int64 usec;
551 :
552 : /* Fast path for common case */
553 10938 : if (frac == 0)
554 10476 : return true;
555 :
556 : /*
557 : * We assume the input frac has abs value less than 1, so overflow of frac
558 : * or usec is not an issue for interesting values of scale.
559 : */
560 462 : frac *= scale;
561 462 : usec = (int64) frac;
562 :
563 : /* Round off any fractional microsecond */
564 462 : frac -= usec;
565 462 : if (frac > 0.5)
566 24 : usec++;
567 438 : else if (frac < -0.5)
568 30 : usec--;
569 :
570 462 : return !pg_add_s64_overflow(itm_in->tm_usec, usec, &itm_in->tm_usec);
571 : }
572 :
573 : /*
574 : * Multiply frac by scale (to produce days). Add the integral part of the
575 : * result to itm_in->tm_mday, the fractional part to itm_in->tm_usec.
576 : * Returns true if successful, false if itm_in overflows.
577 : */
578 : static bool
579 1794 : AdjustFractDays(double frac, int scale,
580 : struct pg_itm_in *itm_in)
581 : {
582 : int extra_days;
583 :
584 : /* Fast path for common case */
585 1794 : if (frac == 0)
586 1608 : return true;
587 :
588 : /*
589 : * We assume the input frac has abs value less than 1, so overflow of frac
590 : * or extra_days is not an issue.
591 : */
592 186 : frac *= scale;
593 186 : extra_days = (int) frac;
594 :
595 : /* ... but this could overflow, if tm_mday is already nonzero */
596 186 : if (pg_add_s32_overflow(itm_in->tm_mday, extra_days, &itm_in->tm_mday))
597 48 : return false;
598 :
599 : /* Handle any fractional day */
600 138 : frac -= extra_days;
601 138 : return AdjustFractMicroseconds(frac, USECS_PER_DAY, itm_in);
602 : }
603 :
604 : /*
605 : * Multiply frac by scale (to produce years), then further scale up to months.
606 : * Add the integral part of the result to itm_in->tm_mon, discarding any
607 : * fractional part.
608 : * Returns true if successful, false if itm_in overflows.
609 : */
610 : static bool
611 54740 : AdjustFractYears(double frac, int scale,
612 : struct pg_itm_in *itm_in)
613 : {
614 : /*
615 : * As above, we assume abs(frac) < 1, so this can't overflow for any
616 : * interesting value of scale.
617 : */
618 54740 : int extra_months = (int) rint(frac * scale * MONTHS_PER_YEAR);
619 :
620 54740 : return !pg_add_s32_overflow(itm_in->tm_mon, extra_months, &itm_in->tm_mon);
621 : }
622 :
623 : /*
624 : * Add (val + fval) * scale to itm_in->tm_usec.
625 : * Returns true if successful, false if itm_in overflows.
626 : */
627 : static bool
628 2804 : AdjustMicroseconds(int64 val, double fval, int64 scale,
629 : struct pg_itm_in *itm_in)
630 : {
631 : /* Handle the integer part */
632 2804 : if (!int64_multiply_add(val, scale, &itm_in->tm_usec))
633 150 : return false;
634 : /* Handle the float part */
635 2654 : return AdjustFractMicroseconds(fval, scale, itm_in);
636 : }
637 :
638 : /*
639 : * Multiply val by scale (to produce days) and add to itm_in->tm_mday.
640 : * Returns true if successful, false if itm_in overflows.
641 : */
642 : static bool
643 8404 : AdjustDays(int64 val, int scale, struct pg_itm_in *itm_in)
644 : {
645 : int days;
646 :
647 8404 : if (val < INT_MIN || val > INT_MAX)
648 36 : return false;
649 16724 : return !pg_mul_s32_overflow((int32) val, scale, &days) &&
650 8356 : !pg_add_s32_overflow(itm_in->tm_mday, days, &itm_in->tm_mday);
651 : }
652 :
653 : /*
654 : * Add val to itm_in->tm_mon (no need for scale here, as val is always
655 : * in months already).
656 : * Returns true if successful, false if itm_in overflows.
657 : */
658 : static bool
659 1764 : AdjustMonths(int64 val, struct pg_itm_in *itm_in)
660 : {
661 1764 : if (val < INT_MIN || val > INT_MAX)
662 12 : return false;
663 1752 : return !pg_add_s32_overflow(itm_in->tm_mon, (int32) val, &itm_in->tm_mon);
664 : }
665 :
666 : /*
667 : * Multiply val by scale (to produce years) and add to itm_in->tm_year.
668 : * Returns true if successful, false if itm_in overflows.
669 : */
670 : static bool
671 54878 : AdjustYears(int64 val, int scale,
672 : struct pg_itm_in *itm_in)
673 : {
674 : int years;
675 :
676 54878 : if (val < INT_MIN || val > INT_MAX)
677 24 : return false;
678 109672 : return !pg_mul_s32_overflow((int32) val, scale, &years) &&
679 54818 : !pg_add_s32_overflow(itm_in->tm_year, years, &itm_in->tm_year);
680 : }
681 :
682 :
683 : /*
684 : * Parse the fractional part of a number (decimal point and optional digits,
685 : * followed by end of string). Returns the fractional value into *frac.
686 : *
687 : * Returns 0 if successful, DTERR code if bogus input detected.
688 : */
689 : static int
690 23960 : ParseFraction(char *cp, double *frac)
691 : {
692 : /* Caller should always pass the start of the fraction part */
693 : Assert(*cp == '.');
694 :
695 : /*
696 : * We want to allow just "." with no digits, but some versions of strtod
697 : * will report EINVAL for that, so special-case it.
698 : */
699 23960 : if (cp[1] == '\0')
700 : {
701 0 : *frac = 0;
702 : }
703 : else
704 : {
705 : /*
706 : * On the other hand, let's reject anything that's not digits after
707 : * the ".". strtod is happy with input like ".123e9", but that'd
708 : * break callers' expectation that the result is in 0..1. (It's quite
709 : * difficult to get here with such input, but not impossible.)
710 : */
711 23960 : if (strspn(cp + 1, "0123456789") != strlen(cp + 1))
712 12 : return DTERR_BAD_FORMAT;
713 :
714 23948 : errno = 0;
715 23948 : *frac = strtod(cp, &cp);
716 : /* check for parse failure (probably redundant given prior check) */
717 23948 : if (*cp != '\0' || errno != 0)
718 0 : return DTERR_BAD_FORMAT;
719 : }
720 23948 : return 0;
721 : }
722 :
723 : /*
724 : * Fetch a fractional-second value with suitable error checking.
725 : * Same as ParseFraction except we convert the result to integer microseconds.
726 : */
727 : static int
728 23462 : ParseFractionalSecond(char *cp, fsec_t *fsec)
729 : {
730 : double frac;
731 : int dterr;
732 :
733 23462 : dterr = ParseFraction(cp, &frac);
734 23462 : if (dterr)
735 12 : return dterr;
736 23450 : *fsec = rint(frac * 1000000);
737 23450 : return 0;
738 : }
739 :
740 :
741 : /* ParseDateTime()
742 : * Break string into tokens based on a date/time context.
743 : * Returns 0 if successful, DTERR code if bogus input detected.
744 : *
745 : * timestr - the input string
746 : * workbuf - workspace for field string storage. This must be
747 : * larger than the largest legal input for this datetime type --
748 : * some additional space will be needed to NUL terminate fields.
749 : * buflen - the size of workbuf
750 : * field[] - pointers to field strings are returned in this array
751 : * ftype[] - field type indicators are returned in this array
752 : * maxfields - dimensions of the above two arrays
753 : * *numfields - set to the actual number of fields detected
754 : *
755 : * The fields extracted from the input are stored as separate,
756 : * null-terminated strings in the workspace at workbuf. Any text is
757 : * converted to lower case.
758 : *
759 : * Several field types are assigned:
760 : * DTK_NUMBER - digits and (possibly) a decimal point
761 : * DTK_DATE - digits and two delimiters, or digits and text
762 : * DTK_TIME - digits, colon delimiters, and possibly a decimal point
763 : * DTK_STRING - text (no digits or punctuation)
764 : * DTK_SPECIAL - leading "+" or "-" followed by text
765 : * DTK_TZ - leading "+" or "-" followed by digits (also eats ':', '.', '-')
766 : *
767 : * Note that some field types can hold unexpected items:
768 : * DTK_NUMBER can hold date fields (yy.ddd)
769 : * DTK_STRING can hold months (January) and time zones (PST)
770 : * DTK_DATE can hold time zone names (America/New_York, GMT-8)
771 : */
772 : int
773 146468 : ParseDateTime(const char *timestr, char *workbuf, size_t buflen,
774 : char **field, int *ftype, int maxfields, int *numfields)
775 : {
776 146468 : int nf = 0;
777 146468 : const char *cp = timestr;
778 146468 : char *bufp = workbuf;
779 146468 : const char *bufend = workbuf + buflen;
780 :
781 : /*
782 : * Set the character pointed-to by "bufptr" to "newchar", and increment
783 : * "bufptr". "end" gives the end of the buffer -- we return an error if
784 : * there is no space left to append a character to the buffer. Note that
785 : * "bufptr" is evaluated twice.
786 : */
787 : #define APPEND_CHAR(bufptr, end, newchar) \
788 : do \
789 : { \
790 : if (((bufptr) + 1) >= (end)) \
791 : return DTERR_BAD_FORMAT; \
792 : *(bufptr)++ = newchar; \
793 : } while (0)
794 :
795 : /* outer loop through fields */
796 619774 : while (*cp != '\0')
797 : {
798 : /* Ignore spaces between fields */
799 473306 : if (isspace((unsigned char) *cp))
800 : {
801 143284 : cp++;
802 143284 : continue;
803 : }
804 :
805 : /* Record start of current field */
806 330022 : if (nf >= maxfields)
807 0 : return DTERR_BAD_FORMAT;
808 330022 : field[nf] = bufp;
809 :
810 : /* leading digit? then date or time */
811 330022 : if (isdigit((unsigned char) *cp))
812 : {
813 206188 : APPEND_CHAR(bufp, bufend, *cp++);
814 659204 : while (isdigit((unsigned char) *cp))
815 453016 : APPEND_CHAR(bufp, bufend, *cp++);
816 :
817 : /* time field? */
818 206188 : if (*cp == ':')
819 : {
820 65794 : ftype[nf] = DTK_TIME;
821 65794 : APPEND_CHAR(bufp, bufend, *cp++);
822 65794 : while (isdigit((unsigned char) *cp) ||
823 548112 : (*cp == ':') || (*cp == '.'))
824 482318 : APPEND_CHAR(bufp, bufend, *cp++);
825 : }
826 : /* date field? allow embedded text month */
827 140394 : else if (*cp == '-' || *cp == '/' || *cp == '.')
828 69706 : {
829 : /* save delimiting character to use later */
830 69706 : char delim = *cp;
831 :
832 69706 : APPEND_CHAR(bufp, bufend, *cp++);
833 : /* second field is all digits? then no embedded text month */
834 69706 : if (isdigit((unsigned char) *cp))
835 : {
836 69610 : ftype[nf] = ((delim == '.') ? DTK_NUMBER : DTK_DATE);
837 208866 : while (isdigit((unsigned char) *cp))
838 139256 : APPEND_CHAR(bufp, bufend, *cp++);
839 :
840 : /*
841 : * insist that the delimiters match to get a three-field
842 : * date.
843 : */
844 69610 : if (*cp == delim)
845 : {
846 69004 : ftype[nf] = DTK_DATE;
847 69004 : APPEND_CHAR(bufp, bufend, *cp++);
848 209890 : while (isdigit((unsigned char) *cp) || *cp == delim)
849 140886 : APPEND_CHAR(bufp, bufend, *cp++);
850 : }
851 : }
852 : else
853 : {
854 96 : ftype[nf] = DTK_DATE;
855 756 : while (isalnum((unsigned char) *cp) || *cp == delim)
856 660 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
857 : }
858 : }
859 :
860 : /*
861 : * otherwise, number only and will determine year, month, day, or
862 : * concatenated fields later...
863 : */
864 : else
865 70688 : ftype[nf] = DTK_NUMBER;
866 : }
867 : /* Leading decimal point? Then fractional seconds... */
868 123834 : else if (*cp == '.')
869 : {
870 0 : APPEND_CHAR(bufp, bufend, *cp++);
871 0 : while (isdigit((unsigned char) *cp))
872 0 : APPEND_CHAR(bufp, bufend, *cp++);
873 :
874 0 : ftype[nf] = DTK_NUMBER;
875 : }
876 :
877 : /*
878 : * text? then date string, month, day of week, special, or timezone
879 : */
880 123834 : else if (isalpha((unsigned char) *cp))
881 : {
882 : bool is_date;
883 :
884 79724 : ftype[nf] = DTK_STRING;
885 79724 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
886 371880 : while (isalpha((unsigned char) *cp))
887 292156 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
888 :
889 : /*
890 : * Dates can have embedded '-', '/', or '.' separators. It could
891 : * also be a timezone name containing embedded '/', '+', '-', '_',
892 : * or ':' (but '_' or ':' can't be the first punctuation). If the
893 : * next character is a digit or '+', we need to check whether what
894 : * we have so far is a recognized non-timezone keyword --- if so,
895 : * don't believe that this is the start of a timezone.
896 : */
897 79724 : is_date = false;
898 79724 : if (*cp == '-' || *cp == '/' || *cp == '.')
899 1510 : is_date = true;
900 78214 : else if (*cp == '+' || isdigit((unsigned char) *cp))
901 : {
902 1932 : *bufp = '\0'; /* null-terminate current field value */
903 : /* we need search only the core token table, not TZ names */
904 1932 : if (datebsearch(field[nf], datetktbl, szdatetktbl) == NULL)
905 1530 : is_date = true;
906 : }
907 79724 : if (is_date)
908 : {
909 3040 : ftype[nf] = DTK_DATE;
910 : do
911 : {
912 14326 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
913 14326 : } while (*cp == '+' || *cp == '-' ||
914 14032 : *cp == '/' || *cp == '_' ||
915 28202 : *cp == '.' || *cp == ':' ||
916 13456 : isalnum((unsigned char) *cp));
917 : }
918 : }
919 : /* sign? then special or numeric timezone */
920 44110 : else if (*cp == '+' || *cp == '-')
921 : {
922 43534 : APPEND_CHAR(bufp, bufend, *cp++);
923 : /* soak up leading whitespace */
924 43558 : while (isspace((unsigned char) *cp))
925 24 : cp++;
926 : /* numeric timezone? */
927 : /* note that "DTK_TZ" could also be a signed float or yyyy-mm */
928 43534 : if (isdigit((unsigned char) *cp))
929 : {
930 42656 : ftype[nf] = DTK_TZ;
931 42656 : APPEND_CHAR(bufp, bufend, *cp++);
932 42656 : while (isdigit((unsigned char) *cp) ||
933 99052 : *cp == ':' || *cp == '.' || *cp == '-')
934 56396 : APPEND_CHAR(bufp, bufend, *cp++);
935 : }
936 : /* special? */
937 878 : else if (isalpha((unsigned char) *cp))
938 : {
939 878 : ftype[nf] = DTK_SPECIAL;
940 878 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
941 7024 : while (isalpha((unsigned char) *cp))
942 6146 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
943 : }
944 : /* otherwise something wrong... */
945 : else
946 0 : return DTERR_BAD_FORMAT;
947 : }
948 : /* ignore other punctuation but use as delimiter */
949 576 : else if (ispunct((unsigned char) *cp))
950 : {
951 576 : cp++;
952 576 : continue;
953 : }
954 : /* otherwise, something is not right... */
955 : else
956 0 : return DTERR_BAD_FORMAT;
957 :
958 : /* force in a delimiter after each field */
959 329446 : *bufp++ = '\0';
960 329446 : nf++;
961 : }
962 :
963 146468 : *numfields = nf;
964 :
965 146468 : return 0;
966 : }
967 :
968 :
969 : /* DecodeDateTime()
970 : * Interpret previously parsed fields for general date and time.
971 : * Return 0 if full date, 1 if only time, and negative DTERR code if problems.
972 : * (Currently, all callers treat 1 as an error return too.)
973 : *
974 : * Inputs are field[] and ftype[] arrays, of length nf.
975 : * Other arguments are outputs.
976 : *
977 : * External format(s):
978 : * "<weekday> <month>-<day>-<year> <hour>:<minute>:<second>"
979 : * "Fri Feb-7-1997 15:23:27"
980 : * "Feb-7-1997 15:23:27"
981 : * "2-7-1997 15:23:27"
982 : * "1997-2-7 15:23:27"
983 : * "1997.038 15:23:27" (day of year 1-366)
984 : * Also supports input in compact time:
985 : * "970207 152327"
986 : * "97038 152327"
987 : * "20011225T040506.789-07"
988 : *
989 : * Use the system-provided functions to get the current time zone
990 : * if not specified in the input string.
991 : *
992 : * If the date is outside the range of pg_time_t (in practice that could only
993 : * happen if pg_time_t is just 32 bits), then assume UTC time zone - thomas
994 : * 1997-05-27
995 : */
996 : int
997 73832 : DecodeDateTime(char **field, int *ftype, int nf,
998 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
999 : DateTimeErrorExtra *extra)
1000 : {
1001 73832 : int fmask = 0,
1002 : tmask,
1003 : type;
1004 73832 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
1005 : int i;
1006 : int val;
1007 : int dterr;
1008 73832 : int mer = HR24;
1009 73832 : bool haveTextMonth = false;
1010 73832 : bool isjulian = false;
1011 73832 : bool is2digits = false;
1012 73832 : bool bc = false;
1013 73832 : pg_tz *namedTz = NULL;
1014 73832 : pg_tz *abbrevTz = NULL;
1015 : pg_tz *valtz;
1016 73832 : char *abbrev = NULL;
1017 : struct pg_tm cur_tm;
1018 :
1019 : /*
1020 : * We'll insist on at least all of the date fields, but initialize the
1021 : * remaining fields in case they are not set later...
1022 : */
1023 73832 : *dtype = DTK_DATE;
1024 73832 : tm->tm_hour = 0;
1025 73832 : tm->tm_min = 0;
1026 73832 : tm->tm_sec = 0;
1027 73832 : *fsec = 0;
1028 : /* don't know daylight savings time status apriori */
1029 73832 : tm->tm_isdst = -1;
1030 73832 : if (tzp != NULL)
1031 73832 : *tzp = 0;
1032 :
1033 255144 : for (i = 0; i < nf; i++)
1034 : {
1035 181642 : switch (ftype[i])
1036 : {
1037 70378 : case DTK_DATE:
1038 :
1039 : /*
1040 : * Integral julian day with attached time zone? All other
1041 : * forms with JD will be separated into distinct fields, so we
1042 : * handle just this case here.
1043 : */
1044 70378 : if (ptype == DTK_JULIAN)
1045 : {
1046 : char *cp;
1047 : int jday;
1048 :
1049 6 : if (tzp == NULL)
1050 0 : return DTERR_BAD_FORMAT;
1051 :
1052 6 : errno = 0;
1053 6 : jday = strtoint(field[i], &cp, 10);
1054 6 : if (errno == ERANGE || jday < 0)
1055 0 : return DTERR_FIELD_OVERFLOW;
1056 :
1057 6 : j2date(jday, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1058 6 : isjulian = true;
1059 :
1060 : /* Get the time zone from the end of the string */
1061 6 : dterr = DecodeTimezone(cp, tzp);
1062 6 : if (dterr)
1063 0 : return dterr;
1064 :
1065 6 : tmask = DTK_DATE_M | DTK_TIME_M | DTK_M(TZ);
1066 6 : ptype = 0;
1067 6 : break;
1068 : }
1069 :
1070 : /*
1071 : * Already have a date? Then this might be a time zone name
1072 : * with embedded punctuation (e.g. "America/New_York") or a
1073 : * run-together time with trailing time zone (e.g. hhmmss-zz).
1074 : * - thomas 2001-12-25
1075 : *
1076 : * We consider it a time zone if we already have month & day.
1077 : * This is to allow the form "mmm dd hhmmss tz year", which
1078 : * we've historically accepted.
1079 : */
1080 70372 : else if (ptype != 0 ||
1081 70348 : ((fmask & (DTK_M(MONTH) | DTK_M(DAY))) ==
1082 : (DTK_M(MONTH) | DTK_M(DAY))))
1083 : {
1084 : /* No time zone accepted? Then quit... */
1085 1434 : if (tzp == NULL)
1086 0 : return DTERR_BAD_FORMAT;
1087 :
1088 1434 : if (isdigit((unsigned char) *field[i]) || ptype != 0)
1089 18 : {
1090 : char *cp;
1091 :
1092 : /*
1093 : * Allow a preceding "t" field, but no other units.
1094 : */
1095 30 : if (ptype != 0)
1096 : {
1097 : /* Sanity check; should not fail this test */
1098 24 : if (ptype != DTK_TIME)
1099 0 : return DTERR_BAD_FORMAT;
1100 24 : ptype = 0;
1101 : }
1102 :
1103 : /*
1104 : * Starts with a digit but we already have a time
1105 : * field? Then we are in trouble with a date and time
1106 : * already...
1107 : */
1108 30 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1109 0 : return DTERR_BAD_FORMAT;
1110 :
1111 30 : if ((cp = strchr(field[i], '-')) == NULL)
1112 0 : return DTERR_BAD_FORMAT;
1113 :
1114 : /* Get the time zone from the end of the string */
1115 30 : dterr = DecodeTimezone(cp, tzp);
1116 30 : if (dterr)
1117 0 : return dterr;
1118 30 : *cp = '\0';
1119 :
1120 : /*
1121 : * Then read the rest of the field as a concatenated
1122 : * time
1123 : */
1124 30 : dterr = DecodeNumberField(strlen(field[i]), field[i],
1125 : fmask,
1126 : &tmask, tm,
1127 : fsec, &is2digits);
1128 30 : if (dterr < 0)
1129 12 : return dterr;
1130 :
1131 : /*
1132 : * modify tmask after returning from
1133 : * DecodeNumberField()
1134 : */
1135 18 : tmask |= DTK_M(TZ);
1136 : }
1137 : else
1138 : {
1139 1404 : namedTz = pg_tzset(field[i]);
1140 1404 : if (!namedTz)
1141 : {
1142 36 : extra->dtee_timezone = field[i];
1143 36 : return DTERR_BAD_TIMEZONE;
1144 : }
1145 : /* we'll apply the zone setting below */
1146 1368 : tmask = DTK_M(TZ);
1147 : }
1148 : }
1149 : else
1150 : {
1151 68938 : dterr = DecodeDate(field[i], fmask,
1152 : &tmask, &is2digits, tm);
1153 68938 : if (dterr)
1154 36 : return dterr;
1155 : }
1156 70288 : break;
1157 :
1158 57458 : case DTK_TIME:
1159 :
1160 : /*
1161 : * This might be an ISO time following a "t" field.
1162 : */
1163 57458 : if (ptype != 0)
1164 : {
1165 : /* Sanity check; should not fail this test */
1166 12 : if (ptype != DTK_TIME)
1167 0 : return DTERR_BAD_FORMAT;
1168 12 : ptype = 0;
1169 : }
1170 57458 : dterr = DecodeTime(field[i], fmask, INTERVAL_FULL_RANGE,
1171 : &tmask, tm, fsec);
1172 57458 : if (dterr)
1173 0 : return dterr;
1174 :
1175 : /* check for time overflow */
1176 57458 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec,
1177 : *fsec))
1178 0 : return DTERR_FIELD_OVERFLOW;
1179 57458 : break;
1180 :
1181 38030 : case DTK_TZ:
1182 : {
1183 : int tz;
1184 :
1185 38030 : if (tzp == NULL)
1186 12 : return DTERR_BAD_FORMAT;
1187 :
1188 38030 : dterr = DecodeTimezone(field[i], &tz);
1189 38030 : if (dterr)
1190 12 : return dterr;
1191 38018 : *tzp = tz;
1192 38018 : tmask = DTK_M(TZ);
1193 : }
1194 38018 : break;
1195 :
1196 6364 : case DTK_NUMBER:
1197 :
1198 : /*
1199 : * Deal with cases where previous field labeled this one
1200 : */
1201 6364 : if (ptype != 0)
1202 : {
1203 : char *cp;
1204 : int value;
1205 :
1206 144 : errno = 0;
1207 144 : value = strtoint(field[i], &cp, 10);
1208 144 : if (errno == ERANGE)
1209 12 : return DTERR_FIELD_OVERFLOW;
1210 144 : if (*cp != '.' && *cp != '\0')
1211 0 : return DTERR_BAD_FORMAT;
1212 :
1213 : switch (ptype)
1214 : {
1215 96 : case DTK_JULIAN:
1216 : /* previous field was a label for "julian date" */
1217 96 : if (value < 0)
1218 0 : return DTERR_FIELD_OVERFLOW;
1219 96 : tmask = DTK_DATE_M;
1220 96 : j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1221 96 : isjulian = true;
1222 :
1223 : /* fractional Julian Day? */
1224 96 : if (*cp == '.')
1225 : {
1226 : double time;
1227 :
1228 12 : dterr = ParseFraction(cp, &time);
1229 12 : if (dterr)
1230 0 : return dterr;
1231 12 : time *= USECS_PER_DAY;
1232 12 : dt2time(time,
1233 : &tm->tm_hour, &tm->tm_min,
1234 : &tm->tm_sec, fsec);
1235 12 : tmask |= DTK_TIME_M;
1236 : }
1237 96 : break;
1238 :
1239 36 : case DTK_TIME:
1240 : /* previous field was "t" for ISO time */
1241 36 : dterr = DecodeNumberField(strlen(field[i]), field[i],
1242 : (fmask | DTK_DATE_M),
1243 : &tmask, tm,
1244 : fsec, &is2digits);
1245 36 : if (dterr < 0)
1246 0 : return dterr;
1247 36 : if (tmask != DTK_TIME_M)
1248 0 : return DTERR_BAD_FORMAT;
1249 36 : break;
1250 :
1251 12 : default:
1252 12 : return DTERR_BAD_FORMAT;
1253 : break;
1254 : }
1255 :
1256 132 : ptype = 0;
1257 132 : *dtype = DTK_DATE;
1258 : }
1259 : else
1260 : {
1261 : char *cp;
1262 : int flen;
1263 :
1264 6220 : flen = strlen(field[i]);
1265 6220 : cp = strchr(field[i], '.');
1266 :
1267 : /* Embedded decimal and no date yet? */
1268 6220 : if (cp != NULL && !(fmask & DTK_DATE_M))
1269 : {
1270 30 : dterr = DecodeDate(field[i], fmask,
1271 : &tmask, &is2digits, tm);
1272 30 : if (dterr)
1273 0 : return dterr;
1274 : }
1275 : /* embedded decimal and several digits before? */
1276 6190 : else if (cp != NULL && flen - strlen(cp) > 2)
1277 : {
1278 : /*
1279 : * Interpret as a concatenated date or time Set the
1280 : * type field to allow decoding other fields later.
1281 : * Example: 20011223 or 040506
1282 : */
1283 12 : dterr = DecodeNumberField(flen, field[i], fmask,
1284 : &tmask, tm,
1285 : fsec, &is2digits);
1286 12 : if (dterr < 0)
1287 0 : return dterr;
1288 : }
1289 :
1290 : /*
1291 : * Is this a YMD or HMS specification, or a year number?
1292 : * YMD and HMS are required to be six digits or more, so
1293 : * if it is 5 digits, it is a year. If it is six or more
1294 : * digits, we assume it is YMD or HMS unless no date and
1295 : * no time values have been specified. This forces 6+
1296 : * digit years to be at the end of the string, or to use
1297 : * the ISO date specification.
1298 : */
1299 6178 : else if (flen >= 6 && (!(fmask & DTK_DATE_M) ||
1300 96 : !(fmask & DTK_TIME_M)))
1301 : {
1302 338 : dterr = DecodeNumberField(flen, field[i], fmask,
1303 : &tmask, tm,
1304 : fsec, &is2digits);
1305 338 : if (dterr < 0)
1306 0 : return dterr;
1307 : }
1308 : /* otherwise it is a single date/time field... */
1309 : else
1310 : {
1311 5840 : dterr = DecodeNumber(flen, field[i],
1312 : haveTextMonth, fmask,
1313 : &tmask, tm,
1314 : fsec, &is2digits);
1315 5840 : if (dterr)
1316 12 : return dterr;
1317 : }
1318 : }
1319 6340 : break;
1320 :
1321 9412 : case DTK_STRING:
1322 : case DTK_SPECIAL:
1323 : /* timezone abbrevs take precedence over built-in tokens */
1324 9412 : dterr = DecodeTimezoneAbbrev(i, field[i],
1325 : &type, &val, &valtz, extra);
1326 9412 : if (dterr)
1327 0 : return dterr;
1328 9412 : if (type == UNKNOWN_FIELD)
1329 6960 : type = DecodeSpecial(i, field[i], &val);
1330 9412 : if (type == IGNORE_DTF)
1331 0 : continue;
1332 :
1333 9412 : tmask = DTK_M(type);
1334 9412 : switch (type)
1335 : {
1336 1700 : case RESERV:
1337 1700 : switch (val)
1338 : {
1339 114 : case DTK_NOW:
1340 114 : tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1341 114 : *dtype = DTK_DATE;
1342 114 : GetCurrentTimeUsec(tm, fsec, tzp);
1343 114 : break;
1344 :
1345 102 : case DTK_YESTERDAY:
1346 102 : tmask = DTK_DATE_M;
1347 102 : *dtype = DTK_DATE;
1348 102 : GetCurrentDateTime(&cur_tm);
1349 102 : j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) - 1,
1350 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1351 102 : break;
1352 :
1353 144 : case DTK_TODAY:
1354 144 : tmask = DTK_DATE_M;
1355 144 : *dtype = DTK_DATE;
1356 144 : GetCurrentDateTime(&cur_tm);
1357 144 : tm->tm_year = cur_tm.tm_year;
1358 144 : tm->tm_mon = cur_tm.tm_mon;
1359 144 : tm->tm_mday = cur_tm.tm_mday;
1360 144 : break;
1361 :
1362 150 : case DTK_TOMORROW:
1363 150 : tmask = DTK_DATE_M;
1364 150 : *dtype = DTK_DATE;
1365 150 : GetCurrentDateTime(&cur_tm);
1366 150 : j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) + 1,
1367 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1368 150 : break;
1369 :
1370 6 : case DTK_ZULU:
1371 6 : tmask = (DTK_TIME_M | DTK_M(TZ));
1372 6 : *dtype = DTK_DATE;
1373 6 : tm->tm_hour = 0;
1374 6 : tm->tm_min = 0;
1375 6 : tm->tm_sec = 0;
1376 6 : if (tzp != NULL)
1377 6 : *tzp = 0;
1378 6 : break;
1379 :
1380 1184 : case DTK_EPOCH:
1381 : case DTK_LATE:
1382 : case DTK_EARLY:
1383 1184 : tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1384 1184 : *dtype = val;
1385 : /* caller ignores tm for these dtype codes */
1386 1184 : break;
1387 :
1388 0 : default:
1389 0 : elog(ERROR, "unrecognized RESERV datetime token: %d",
1390 : val);
1391 : }
1392 :
1393 1700 : break;
1394 :
1395 2680 : case MONTH:
1396 :
1397 : /*
1398 : * already have a (numeric) month? then see if we can
1399 : * substitute...
1400 : */
1401 2680 : if ((fmask & DTK_M(MONTH)) && !haveTextMonth &&
1402 92 : !(fmask & DTK_M(DAY)) && tm->tm_mon >= 1 &&
1403 80 : tm->tm_mon <= 31)
1404 : {
1405 74 : tm->tm_mday = tm->tm_mon;
1406 74 : tmask = DTK_M(DAY);
1407 : }
1408 2680 : haveTextMonth = true;
1409 2680 : tm->tm_mon = val;
1410 2680 : break;
1411 :
1412 6 : case DTZMOD:
1413 :
1414 : /*
1415 : * daylight savings time modifier (solves "MET DST"
1416 : * syntax)
1417 : */
1418 6 : tmask |= DTK_M(DTZ);
1419 6 : tm->tm_isdst = 1;
1420 6 : if (tzp == NULL)
1421 0 : return DTERR_BAD_FORMAT;
1422 6 : *tzp -= val;
1423 6 : break;
1424 :
1425 1778 : case DTZ:
1426 :
1427 : /*
1428 : * set mask for TZ here _or_ check for DTZ later when
1429 : * getting default timezone
1430 : */
1431 1778 : tmask |= DTK_M(TZ);
1432 1778 : tm->tm_isdst = 1;
1433 1778 : if (tzp == NULL)
1434 0 : return DTERR_BAD_FORMAT;
1435 1778 : *tzp = -val;
1436 1778 : break;
1437 :
1438 590 : case TZ:
1439 590 : tm->tm_isdst = 0;
1440 590 : if (tzp == NULL)
1441 0 : return DTERR_BAD_FORMAT;
1442 590 : *tzp = -val;
1443 590 : break;
1444 :
1445 84 : case DYNTZ:
1446 84 : tmask |= DTK_M(TZ);
1447 84 : if (tzp == NULL)
1448 0 : return DTERR_BAD_FORMAT;
1449 : /* we'll determine the actual offset later */
1450 84 : abbrevTz = valtz;
1451 84 : abbrev = field[i];
1452 84 : break;
1453 :
1454 36 : case AMPM:
1455 36 : mer = val;
1456 36 : break;
1457 :
1458 316 : case ADBC:
1459 316 : bc = (val == BC);
1460 316 : break;
1461 :
1462 1958 : case DOW:
1463 1958 : tm->tm_wday = val;
1464 1958 : break;
1465 :
1466 138 : case UNITS:
1467 138 : tmask = 0;
1468 : /* reject consecutive unhandled units */
1469 138 : if (ptype != 0)
1470 12 : return DTERR_BAD_FORMAT;
1471 126 : ptype = val;
1472 126 : break;
1473 :
1474 72 : case ISOTIME:
1475 :
1476 : /*
1477 : * This is a filler field "t" indicating that the next
1478 : * field is time. Try to verify that this is sensible.
1479 : */
1480 72 : tmask = 0;
1481 :
1482 : /* No preceding date? Then quit... */
1483 72 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1484 0 : return DTERR_BAD_FORMAT;
1485 :
1486 : /* reject consecutive unhandled units */
1487 72 : if (ptype != 0)
1488 0 : return DTERR_BAD_FORMAT;
1489 72 : ptype = val;
1490 72 : break;
1491 :
1492 54 : case UNKNOWN_FIELD:
1493 :
1494 : /*
1495 : * Before giving up and declaring error, check to see
1496 : * if it is an all-alpha timezone name.
1497 : */
1498 54 : namedTz = pg_tzset(field[i]);
1499 54 : if (!namedTz)
1500 54 : return DTERR_BAD_FORMAT;
1501 : /* we'll apply the zone setting below */
1502 0 : tmask = DTK_M(TZ);
1503 0 : break;
1504 :
1505 0 : default:
1506 0 : return DTERR_BAD_FORMAT;
1507 : }
1508 9346 : break;
1509 :
1510 0 : default:
1511 0 : return DTERR_BAD_FORMAT;
1512 : }
1513 :
1514 181456 : if (tmask & fmask)
1515 144 : return DTERR_BAD_FORMAT;
1516 181312 : fmask |= tmask;
1517 : } /* end loop over fields */
1518 :
1519 : /* reject if prefix type appeared and was never handled */
1520 73502 : if (ptype != 0)
1521 0 : return DTERR_BAD_FORMAT;
1522 :
1523 : /* do additional checking for normal date specs (but not "infinity" etc) */
1524 73502 : if (*dtype == DTK_DATE)
1525 : {
1526 : /* do final checking/adjustment of Y/M/D fields */
1527 72462 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
1528 72462 : if (dterr)
1529 198 : return dterr;
1530 :
1531 : /* handle AM/PM */
1532 72264 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
1533 0 : return DTERR_FIELD_OVERFLOW;
1534 72264 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
1535 0 : tm->tm_hour = 0;
1536 72264 : else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
1537 36 : tm->tm_hour += HOURS_PER_DAY / 2;
1538 :
1539 : /* check for incomplete input */
1540 72264 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1541 : {
1542 6 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1543 0 : return 1;
1544 6 : return DTERR_BAD_FORMAT;
1545 : }
1546 :
1547 : /*
1548 : * If we had a full timezone spec, compute the offset (we could not do
1549 : * it before, because we need the date to resolve DST status).
1550 : */
1551 72258 : if (namedTz != NULL)
1552 : {
1553 : /* daylight savings time modifier disallowed with full TZ */
1554 1368 : if (fmask & DTK_M(DTZMOD))
1555 0 : return DTERR_BAD_FORMAT;
1556 :
1557 1368 : *tzp = DetermineTimeZoneOffset(tm, namedTz);
1558 : }
1559 :
1560 : /*
1561 : * Likewise, if we had a dynamic timezone abbreviation, resolve it
1562 : * now.
1563 : */
1564 72258 : if (abbrevTz != NULL)
1565 : {
1566 : /* daylight savings time modifier disallowed with dynamic TZ */
1567 84 : if (fmask & DTK_M(DTZMOD))
1568 0 : return DTERR_BAD_FORMAT;
1569 :
1570 84 : *tzp = DetermineTimeZoneAbbrevOffset(tm, abbrev, abbrevTz);
1571 : }
1572 :
1573 : /* timezone not specified? then use session timezone */
1574 72258 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
1575 : {
1576 : /*
1577 : * daylight savings time modifier but no standard timezone? then
1578 : * error
1579 : */
1580 30300 : if (fmask & DTK_M(DTZMOD))
1581 0 : return DTERR_BAD_FORMAT;
1582 :
1583 30300 : *tzp = DetermineTimeZoneOffset(tm, session_timezone);
1584 : }
1585 : }
1586 :
1587 73298 : return 0;
1588 : }
1589 :
1590 :
1591 : /* DetermineTimeZoneOffset()
1592 : *
1593 : * Given a struct pg_tm in which tm_year, tm_mon, tm_mday, tm_hour, tm_min,
1594 : * and tm_sec fields are set, and a zic-style time zone definition, determine
1595 : * the applicable GMT offset and daylight-savings status at that time.
1596 : * Set the struct pg_tm's tm_isdst field accordingly, and return the GMT
1597 : * offset as the function result.
1598 : *
1599 : * Note: if the date is out of the range we can deal with, we return zero
1600 : * as the GMT offset and set tm_isdst = 0. We don't throw an error here,
1601 : * though probably some higher-level code will.
1602 : */
1603 : int
1604 114574 : DetermineTimeZoneOffset(struct pg_tm *tm, pg_tz *tzp)
1605 : {
1606 : pg_time_t t;
1607 :
1608 114574 : return DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1609 : }
1610 :
1611 :
1612 : /* DetermineTimeZoneOffsetInternal()
1613 : *
1614 : * As above, but also return the actual UTC time imputed to the date/time
1615 : * into *tp.
1616 : *
1617 : * In event of an out-of-range date, we punt by returning zero into *tp.
1618 : * This is okay for the immediate callers but is a good reason for not
1619 : * exposing this worker function globally.
1620 : *
1621 : * Note: it might seem that we should use mktime() for this, but bitter
1622 : * experience teaches otherwise. This code is much faster than most versions
1623 : * of mktime(), anyway.
1624 : */
1625 : static int
1626 114760 : DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp, pg_time_t *tp)
1627 : {
1628 : int date,
1629 : sec;
1630 : pg_time_t day,
1631 : mytime,
1632 : prevtime,
1633 : boundary,
1634 : beforetime,
1635 : aftertime;
1636 : long int before_gmtoff,
1637 : after_gmtoff;
1638 : int before_isdst,
1639 : after_isdst;
1640 : int res;
1641 :
1642 : /*
1643 : * First, generate the pg_time_t value corresponding to the given
1644 : * y/m/d/h/m/s taken as GMT time. If this overflows, punt and decide the
1645 : * timezone is GMT. (For a valid Julian date, integer overflow should be
1646 : * impossible with 64-bit pg_time_t, but let's check for safety.)
1647 : */
1648 114760 : if (!IS_VALID_JULIAN(tm->tm_year, tm->tm_mon, tm->tm_mday))
1649 24 : goto overflow;
1650 114736 : date = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday) - UNIX_EPOCH_JDATE;
1651 :
1652 114736 : day = ((pg_time_t) date) * SECS_PER_DAY;
1653 114736 : if (day / SECS_PER_DAY != date)
1654 0 : goto overflow;
1655 114736 : sec = tm->tm_sec + (tm->tm_min + tm->tm_hour * MINS_PER_HOUR) * SECS_PER_MINUTE;
1656 114736 : mytime = day + sec;
1657 : /* since sec >= 0, overflow could only be from +day to -mytime */
1658 114736 : if (mytime < 0 && day > 0)
1659 0 : goto overflow;
1660 :
1661 : /*
1662 : * Find the DST time boundary just before or following the target time. We
1663 : * assume that all zones have GMT offsets less than 24 hours, and that DST
1664 : * boundaries can't be closer together than 48 hours, so backing up 24
1665 : * hours and finding the "next" boundary will work.
1666 : */
1667 114736 : prevtime = mytime - SECS_PER_DAY;
1668 114736 : if (mytime < 0 && prevtime > 0)
1669 0 : goto overflow;
1670 :
1671 114736 : res = pg_next_dst_boundary(&prevtime,
1672 : &before_gmtoff, &before_isdst,
1673 : &boundary,
1674 : &after_gmtoff, &after_isdst,
1675 : tzp);
1676 114736 : if (res < 0)
1677 0 : goto overflow; /* failure? */
1678 :
1679 114736 : if (res == 0)
1680 : {
1681 : /* Non-DST zone, life is simple */
1682 7916 : tm->tm_isdst = before_isdst;
1683 7916 : *tp = mytime - before_gmtoff;
1684 7916 : return -(int) before_gmtoff;
1685 : }
1686 :
1687 : /*
1688 : * Form the candidate pg_time_t values with local-time adjustment
1689 : */
1690 106820 : beforetime = mytime - before_gmtoff;
1691 106820 : if ((before_gmtoff > 0 &&
1692 12 : mytime < 0 && beforetime > 0) ||
1693 106820 : (before_gmtoff <= 0 &&
1694 93990 : mytime > 0 && beforetime < 0))
1695 0 : goto overflow;
1696 106820 : aftertime = mytime - after_gmtoff;
1697 106820 : if ((after_gmtoff > 0 &&
1698 12 : mytime < 0 && aftertime > 0) ||
1699 106820 : (after_gmtoff <= 0 &&
1700 93990 : mytime > 0 && aftertime < 0))
1701 0 : goto overflow;
1702 :
1703 : /*
1704 : * If both before or both after the boundary time, we know what to do. The
1705 : * boundary time itself is considered to be after the transition, which
1706 : * means we can accept aftertime == boundary in the second case.
1707 : */
1708 106820 : if (beforetime < boundary && aftertime < boundary)
1709 : {
1710 105942 : tm->tm_isdst = before_isdst;
1711 105942 : *tp = beforetime;
1712 105942 : return -(int) before_gmtoff;
1713 : }
1714 878 : if (beforetime > boundary && aftertime >= boundary)
1715 : {
1716 728 : tm->tm_isdst = after_isdst;
1717 728 : *tp = aftertime;
1718 728 : return -(int) after_gmtoff;
1719 : }
1720 :
1721 : /*
1722 : * It's an invalid or ambiguous time due to timezone transition. In a
1723 : * spring-forward transition, prefer the "before" interpretation; in a
1724 : * fall-back transition, prefer "after". (We used to define and implement
1725 : * this test as "prefer the standard-time interpretation", but that rule
1726 : * does not help to resolve the behavior when both times are reported as
1727 : * standard time; which does happen, eg Europe/Moscow in Oct 2014. Also,
1728 : * in some zones such as Europe/Dublin, there is widespread confusion
1729 : * about which time offset is "standard" time, so it's fortunate that our
1730 : * behavior doesn't depend on that.)
1731 : */
1732 150 : if (beforetime > aftertime)
1733 : {
1734 72 : tm->tm_isdst = before_isdst;
1735 72 : *tp = beforetime;
1736 72 : return -(int) before_gmtoff;
1737 : }
1738 78 : tm->tm_isdst = after_isdst;
1739 78 : *tp = aftertime;
1740 78 : return -(int) after_gmtoff;
1741 :
1742 24 : overflow:
1743 : /* Given date is out of range, so assume UTC */
1744 24 : tm->tm_isdst = 0;
1745 24 : *tp = 0;
1746 24 : return 0;
1747 : }
1748 :
1749 :
1750 : /* DetermineTimeZoneAbbrevOffset()
1751 : *
1752 : * Determine the GMT offset and DST flag to be attributed to a dynamic
1753 : * time zone abbreviation, that is one whose meaning has changed over time.
1754 : * *tm contains the local time at which the meaning should be determined,
1755 : * and tm->tm_isdst receives the DST flag.
1756 : *
1757 : * This differs from the behavior of DetermineTimeZoneOffset() in that a
1758 : * standard-time or daylight-time abbreviation forces use of the corresponding
1759 : * GMT offset even when the zone was then in DS or standard time respectively.
1760 : * (However, that happens only if we can match the given abbreviation to some
1761 : * abbreviation that appears in the IANA timezone data. Otherwise, we fall
1762 : * back to doing DetermineTimeZoneOffset().)
1763 : */
1764 : int
1765 186 : DetermineTimeZoneAbbrevOffset(struct pg_tm *tm, const char *abbr, pg_tz *tzp)
1766 : {
1767 : pg_time_t t;
1768 : int zone_offset;
1769 : int abbr_offset;
1770 : int abbr_isdst;
1771 :
1772 : /*
1773 : * Compute the UTC time we want to probe at. (In event of overflow, we'll
1774 : * probe at the epoch, which is a bit random but probably doesn't matter.)
1775 : */
1776 186 : zone_offset = DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1777 :
1778 : /*
1779 : * Try to match the abbreviation to something in the zone definition.
1780 : */
1781 186 : if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1782 : &abbr_offset, &abbr_isdst))
1783 : {
1784 : /* Success, so use the abbrev-specific answers. */
1785 186 : tm->tm_isdst = abbr_isdst;
1786 186 : return abbr_offset;
1787 : }
1788 :
1789 : /*
1790 : * No match, so use the answers we already got from
1791 : * DetermineTimeZoneOffsetInternal.
1792 : */
1793 0 : return zone_offset;
1794 : }
1795 :
1796 :
1797 : /* DetermineTimeZoneAbbrevOffsetTS()
1798 : *
1799 : * As above but the probe time is specified as a TimestampTz (hence, UTC time),
1800 : * and DST status is returned into *isdst rather than into tm->tm_isdst.
1801 : */
1802 : int
1803 1266 : DetermineTimeZoneAbbrevOffsetTS(TimestampTz ts, const char *abbr,
1804 : pg_tz *tzp, int *isdst)
1805 : {
1806 1266 : pg_time_t t = timestamptz_to_time_t(ts);
1807 : int zone_offset;
1808 : int abbr_offset;
1809 : int tz;
1810 : struct pg_tm tm;
1811 : fsec_t fsec;
1812 :
1813 : /*
1814 : * If the abbrev matches anything in the zone data, this is pretty easy.
1815 : */
1816 1266 : if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1817 : &abbr_offset, isdst))
1818 96 : return abbr_offset;
1819 :
1820 : /*
1821 : * Else, break down the timestamp so we can use DetermineTimeZoneOffset.
1822 : */
1823 1170 : if (timestamp2tm(ts, &tz, &tm, &fsec, NULL, tzp) != 0)
1824 0 : ereport(ERROR,
1825 : (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
1826 : errmsg("timestamp out of range")));
1827 :
1828 1170 : zone_offset = DetermineTimeZoneOffset(&tm, tzp);
1829 1170 : *isdst = tm.tm_isdst;
1830 1170 : return zone_offset;
1831 : }
1832 :
1833 :
1834 : /* DetermineTimeZoneAbbrevOffsetInternal()
1835 : *
1836 : * Workhorse for above two functions: work from a pg_time_t probe instant.
1837 : * On success, return GMT offset and DST status into *offset and *isdst.
1838 : */
1839 : static bool
1840 1452 : DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t, const char *abbr, pg_tz *tzp,
1841 : int *offset, int *isdst)
1842 : {
1843 : char upabbr[TZ_STRLEN_MAX + 1];
1844 : unsigned char *p;
1845 : long int gmtoff;
1846 :
1847 : /* We need to force the abbrev to upper case */
1848 1452 : strlcpy(upabbr, abbr, sizeof(upabbr));
1849 6834 : for (p = (unsigned char *) upabbr; *p; p++)
1850 5382 : *p = pg_toupper(*p);
1851 :
1852 : /* Look up the abbrev's meaning at this time in this zone */
1853 1452 : if (pg_interpret_timezone_abbrev(upabbr,
1854 : &t,
1855 : &gmtoff,
1856 : isdst,
1857 : tzp))
1858 : {
1859 : /* Change sign to agree with DetermineTimeZoneOffset() */
1860 282 : *offset = (int) -gmtoff;
1861 282 : return true;
1862 : }
1863 1170 : return false;
1864 : }
1865 :
1866 :
1867 : /* TimeZoneAbbrevIsKnown()
1868 : *
1869 : * Detect whether the given string is a time zone abbreviation that's known
1870 : * in the specified TZDB timezone, and if so whether it's fixed or varying
1871 : * meaning. The match is not case-sensitive.
1872 : */
1873 : static bool
1874 8042 : TimeZoneAbbrevIsKnown(const char *abbr, pg_tz *tzp,
1875 : bool *isfixed, int *offset, int *isdst)
1876 : {
1877 : char upabbr[TZ_STRLEN_MAX + 1];
1878 : unsigned char *p;
1879 : long int gmtoff;
1880 :
1881 : /* We need to force the abbrev to upper case */
1882 8042 : strlcpy(upabbr, abbr, sizeof(upabbr));
1883 42748 : for (p = (unsigned char *) upabbr; *p; p++)
1884 34706 : *p = pg_toupper(*p);
1885 :
1886 : /* Look up the abbrev's meaning in this zone */
1887 8042 : if (pg_timezone_abbrev_is_known(upabbr,
1888 : isfixed,
1889 : &gmtoff,
1890 : isdst,
1891 : tzp))
1892 : {
1893 : /* Change sign to agree with DetermineTimeZoneOffset() */
1894 240 : *offset = (int) -gmtoff;
1895 240 : return true;
1896 : }
1897 7802 : return false;
1898 : }
1899 :
1900 :
1901 : /* DecodeTimeOnly()
1902 : * Interpret parsed string as time fields only.
1903 : * Returns 0 if successful, DTERR code if bogus input detected.
1904 : *
1905 : * Inputs are field[] and ftype[] arrays, of length nf.
1906 : * Other arguments are outputs.
1907 : *
1908 : * Note that support for time zone is here for
1909 : * SQL TIME WITH TIME ZONE, but it reveals
1910 : * bogosity with SQL date/time standards, since
1911 : * we must infer a time zone from current time.
1912 : * - thomas 2000-03-10
1913 : * Allow specifying date to get a better time zone,
1914 : * if time zones are allowed. - thomas 2001-12-26
1915 : */
1916 : int
1917 6234 : DecodeTimeOnly(char **field, int *ftype, int nf,
1918 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
1919 : DateTimeErrorExtra *extra)
1920 : {
1921 6234 : int fmask = 0,
1922 : tmask,
1923 : type;
1924 6234 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
1925 : int i;
1926 : int val;
1927 : int dterr;
1928 6234 : bool isjulian = false;
1929 6234 : bool is2digits = false;
1930 6234 : bool bc = false;
1931 6234 : int mer = HR24;
1932 6234 : pg_tz *namedTz = NULL;
1933 6234 : pg_tz *abbrevTz = NULL;
1934 6234 : char *abbrev = NULL;
1935 : pg_tz *valtz;
1936 :
1937 6234 : *dtype = DTK_TIME;
1938 6234 : tm->tm_hour = 0;
1939 6234 : tm->tm_min = 0;
1940 6234 : tm->tm_sec = 0;
1941 6234 : *fsec = 0;
1942 : /* don't know daylight savings time status apriori */
1943 6234 : tm->tm_isdst = -1;
1944 :
1945 6234 : if (tzp != NULL)
1946 6234 : *tzp = 0;
1947 :
1948 15844 : for (i = 0; i < nf; i++)
1949 : {
1950 9634 : switch (ftype[i])
1951 : {
1952 1402 : case DTK_DATE:
1953 :
1954 : /*
1955 : * Time zone not allowed? Then should not accept dates or time
1956 : * zones no matter what else!
1957 : */
1958 1402 : if (tzp == NULL)
1959 0 : return DTERR_BAD_FORMAT;
1960 :
1961 : /* Under limited circumstances, we will accept a date... */
1962 1402 : if (i == 0 && nf >= 2 &&
1963 210 : (ftype[nf - 1] == DTK_DATE || ftype[1] == DTK_TIME))
1964 : {
1965 210 : dterr = DecodeDate(field[i], fmask,
1966 : &tmask, &is2digits, tm);
1967 210 : if (dterr)
1968 0 : return dterr;
1969 : }
1970 : /* otherwise, this is a time and/or time zone */
1971 : else
1972 : {
1973 1192 : if (isdigit((unsigned char) *field[i]))
1974 : {
1975 : char *cp;
1976 :
1977 : /*
1978 : * Starts with a digit but we already have a time
1979 : * field? Then we are in trouble with time already...
1980 : */
1981 0 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1982 0 : return DTERR_BAD_FORMAT;
1983 :
1984 : /*
1985 : * Should not get here and fail. Sanity check only...
1986 : */
1987 0 : if ((cp = strchr(field[i], '-')) == NULL)
1988 0 : return DTERR_BAD_FORMAT;
1989 :
1990 : /* Get the time zone from the end of the string */
1991 0 : dterr = DecodeTimezone(cp, tzp);
1992 0 : if (dterr)
1993 0 : return dterr;
1994 0 : *cp = '\0';
1995 :
1996 : /*
1997 : * Then read the rest of the field as a concatenated
1998 : * time
1999 : */
2000 0 : dterr = DecodeNumberField(strlen(field[i]), field[i],
2001 : (fmask | DTK_DATE_M),
2002 : &tmask, tm,
2003 : fsec, &is2digits);
2004 0 : if (dterr < 0)
2005 0 : return dterr;
2006 0 : ftype[i] = dterr;
2007 :
2008 0 : tmask |= DTK_M(TZ);
2009 : }
2010 : else
2011 : {
2012 1192 : namedTz = pg_tzset(field[i]);
2013 1192 : if (!namedTz)
2014 : {
2015 0 : extra->dtee_timezone = field[i];
2016 0 : return DTERR_BAD_TIMEZONE;
2017 : }
2018 : /* we'll apply the zone setting below */
2019 1192 : ftype[i] = DTK_TZ;
2020 1192 : tmask = DTK_M(TZ);
2021 : }
2022 : }
2023 1402 : break;
2024 :
2025 6060 : case DTK_TIME:
2026 :
2027 : /*
2028 : * This might be an ISO time following a "t" field.
2029 : */
2030 6060 : if (ptype != 0)
2031 : {
2032 36 : if (ptype != DTK_TIME)
2033 0 : return DTERR_BAD_FORMAT;
2034 36 : ptype = 0;
2035 : }
2036 :
2037 6060 : dterr = DecodeTime(field[i], (fmask | DTK_DATE_M),
2038 : INTERVAL_FULL_RANGE,
2039 : &tmask, tm, fsec);
2040 6060 : if (dterr)
2041 0 : return dterr;
2042 6060 : break;
2043 :
2044 1548 : case DTK_TZ:
2045 : {
2046 : int tz;
2047 :
2048 1548 : if (tzp == NULL)
2049 0 : return DTERR_BAD_FORMAT;
2050 :
2051 1548 : dterr = DecodeTimezone(field[i], &tz);
2052 1548 : if (dterr)
2053 0 : return dterr;
2054 1548 : *tzp = tz;
2055 1548 : tmask = DTK_M(TZ);
2056 : }
2057 1548 : break;
2058 :
2059 180 : case DTK_NUMBER:
2060 :
2061 : /*
2062 : * Deal with cases where previous field labeled this one
2063 : */
2064 180 : if (ptype != 0)
2065 : {
2066 : char *cp;
2067 : int value;
2068 :
2069 120 : errno = 0;
2070 120 : value = strtoint(field[i], &cp, 10);
2071 120 : if (errno == ERANGE)
2072 24 : return DTERR_FIELD_OVERFLOW;
2073 120 : if (*cp != '.' && *cp != '\0')
2074 0 : return DTERR_BAD_FORMAT;
2075 :
2076 : switch (ptype)
2077 : {
2078 6 : case DTK_JULIAN:
2079 : /* previous field was a label for "julian date" */
2080 6 : if (tzp == NULL)
2081 0 : return DTERR_BAD_FORMAT;
2082 6 : if (value < 0)
2083 0 : return DTERR_FIELD_OVERFLOW;
2084 6 : tmask = DTK_DATE_M;
2085 6 : j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2086 6 : isjulian = true;
2087 :
2088 6 : if (*cp == '.')
2089 : {
2090 : double time;
2091 :
2092 0 : dterr = ParseFraction(cp, &time);
2093 0 : if (dterr)
2094 0 : return dterr;
2095 0 : time *= USECS_PER_DAY;
2096 0 : dt2time(time,
2097 : &tm->tm_hour, &tm->tm_min,
2098 : &tm->tm_sec, fsec);
2099 0 : tmask |= DTK_TIME_M;
2100 : }
2101 6 : break;
2102 :
2103 102 : case DTK_TIME:
2104 : /* previous field was "t" for ISO time */
2105 102 : dterr = DecodeNumberField(strlen(field[i]), field[i],
2106 : (fmask | DTK_DATE_M),
2107 : &tmask, tm,
2108 : fsec, &is2digits);
2109 102 : if (dterr < 0)
2110 12 : return dterr;
2111 90 : ftype[i] = dterr;
2112 :
2113 90 : if (tmask != DTK_TIME_M)
2114 0 : return DTERR_BAD_FORMAT;
2115 90 : break;
2116 :
2117 12 : default:
2118 12 : return DTERR_BAD_FORMAT;
2119 : break;
2120 : }
2121 :
2122 96 : ptype = 0;
2123 96 : *dtype = DTK_DATE;
2124 : }
2125 : else
2126 : {
2127 : char *cp;
2128 : int flen;
2129 :
2130 60 : flen = strlen(field[i]);
2131 60 : cp = strchr(field[i], '.');
2132 :
2133 : /* Embedded decimal? */
2134 60 : if (cp != NULL)
2135 : {
2136 : /*
2137 : * Under limited circumstances, we will accept a
2138 : * date...
2139 : */
2140 36 : if (i == 0 && nf >= 2 && ftype[nf - 1] == DTK_DATE)
2141 : {
2142 0 : dterr = DecodeDate(field[i], fmask,
2143 : &tmask, &is2digits, tm);
2144 0 : if (dterr)
2145 0 : return dterr;
2146 : }
2147 : /* embedded decimal and several digits before? */
2148 36 : else if (flen - strlen(cp) > 2)
2149 : {
2150 : /*
2151 : * Interpret as a concatenated date or time Set
2152 : * the type field to allow decoding other fields
2153 : * later. Example: 20011223 or 040506
2154 : */
2155 36 : dterr = DecodeNumberField(flen, field[i],
2156 : (fmask | DTK_DATE_M),
2157 : &tmask, tm,
2158 : fsec, &is2digits);
2159 36 : if (dterr < 0)
2160 0 : return dterr;
2161 36 : ftype[i] = dterr;
2162 : }
2163 : else
2164 0 : return DTERR_BAD_FORMAT;
2165 : }
2166 24 : else if (flen > 4)
2167 : {
2168 12 : dterr = DecodeNumberField(flen, field[i],
2169 : (fmask | DTK_DATE_M),
2170 : &tmask, tm,
2171 : fsec, &is2digits);
2172 12 : if (dterr < 0)
2173 0 : return dterr;
2174 12 : ftype[i] = dterr;
2175 : }
2176 : /* otherwise it is a single date/time field... */
2177 : else
2178 : {
2179 12 : dterr = DecodeNumber(flen, field[i],
2180 : false,
2181 : (fmask | DTK_DATE_M),
2182 : &tmask, tm,
2183 : fsec, &is2digits);
2184 12 : if (dterr)
2185 0 : return dterr;
2186 : }
2187 : }
2188 156 : break;
2189 :
2190 444 : case DTK_STRING:
2191 : case DTK_SPECIAL:
2192 : /* timezone abbrevs take precedence over built-in tokens */
2193 444 : dterr = DecodeTimezoneAbbrev(i, field[i],
2194 : &type, &val, &valtz, extra);
2195 444 : if (dterr)
2196 0 : return dterr;
2197 444 : if (type == UNKNOWN_FIELD)
2198 180 : type = DecodeSpecial(i, field[i], &val);
2199 444 : if (type == IGNORE_DTF)
2200 0 : continue;
2201 :
2202 444 : tmask = DTK_M(type);
2203 444 : switch (type)
2204 : {
2205 12 : case RESERV:
2206 12 : switch (val)
2207 : {
2208 12 : case DTK_NOW:
2209 12 : tmask = DTK_TIME_M;
2210 12 : *dtype = DTK_TIME;
2211 12 : GetCurrentTimeUsec(tm, fsec, NULL);
2212 12 : break;
2213 :
2214 0 : case DTK_ZULU:
2215 0 : tmask = (DTK_TIME_M | DTK_M(TZ));
2216 0 : *dtype = DTK_TIME;
2217 0 : tm->tm_hour = 0;
2218 0 : tm->tm_min = 0;
2219 0 : tm->tm_sec = 0;
2220 0 : tm->tm_isdst = 0;
2221 0 : break;
2222 :
2223 0 : default:
2224 0 : return DTERR_BAD_FORMAT;
2225 : }
2226 :
2227 12 : break;
2228 :
2229 0 : case DTZMOD:
2230 :
2231 : /*
2232 : * daylight savings time modifier (solves "MET DST"
2233 : * syntax)
2234 : */
2235 0 : tmask |= DTK_M(DTZ);
2236 0 : tm->tm_isdst = 1;
2237 0 : if (tzp == NULL)
2238 0 : return DTERR_BAD_FORMAT;
2239 0 : *tzp -= val;
2240 0 : break;
2241 :
2242 198 : case DTZ:
2243 :
2244 : /*
2245 : * set mask for TZ here _or_ check for DTZ later when
2246 : * getting default timezone
2247 : */
2248 198 : tmask |= DTK_M(TZ);
2249 198 : tm->tm_isdst = 1;
2250 198 : if (tzp == NULL)
2251 0 : return DTERR_BAD_FORMAT;
2252 198 : *tzp = -val;
2253 198 : ftype[i] = DTK_TZ;
2254 198 : break;
2255 :
2256 60 : case TZ:
2257 60 : tm->tm_isdst = 0;
2258 60 : if (tzp == NULL)
2259 0 : return DTERR_BAD_FORMAT;
2260 60 : *tzp = -val;
2261 60 : ftype[i] = DTK_TZ;
2262 60 : break;
2263 :
2264 6 : case DYNTZ:
2265 6 : tmask |= DTK_M(TZ);
2266 6 : if (tzp == NULL)
2267 0 : return DTERR_BAD_FORMAT;
2268 : /* we'll determine the actual offset later */
2269 6 : abbrevTz = valtz;
2270 6 : abbrev = field[i];
2271 6 : ftype[i] = DTK_TZ;
2272 6 : break;
2273 :
2274 12 : case AMPM:
2275 12 : mer = val;
2276 12 : break;
2277 :
2278 0 : case ADBC:
2279 0 : bc = (val == BC);
2280 0 : break;
2281 :
2282 18 : case UNITS:
2283 18 : tmask = 0;
2284 : /* reject consecutive unhandled units */
2285 18 : if (ptype != 0)
2286 0 : return DTERR_BAD_FORMAT;
2287 18 : ptype = val;
2288 18 : break;
2289 :
2290 138 : case ISOTIME:
2291 138 : tmask = 0;
2292 : /* reject consecutive unhandled units */
2293 138 : if (ptype != 0)
2294 0 : return DTERR_BAD_FORMAT;
2295 138 : ptype = val;
2296 138 : break;
2297 :
2298 0 : case UNKNOWN_FIELD:
2299 :
2300 : /*
2301 : * Before giving up and declaring error, check to see
2302 : * if it is an all-alpha timezone name.
2303 : */
2304 0 : namedTz = pg_tzset(field[i]);
2305 0 : if (!namedTz)
2306 0 : return DTERR_BAD_FORMAT;
2307 : /* we'll apply the zone setting below */
2308 0 : tmask = DTK_M(TZ);
2309 0 : break;
2310 :
2311 0 : default:
2312 0 : return DTERR_BAD_FORMAT;
2313 : }
2314 444 : break;
2315 :
2316 0 : default:
2317 0 : return DTERR_BAD_FORMAT;
2318 : }
2319 :
2320 9610 : if (tmask & fmask)
2321 0 : return DTERR_BAD_FORMAT;
2322 9610 : fmask |= tmask;
2323 : } /* end loop over fields */
2324 :
2325 : /* reject if prefix type appeared and was never handled */
2326 6210 : if (ptype != 0)
2327 0 : return DTERR_BAD_FORMAT;
2328 :
2329 : /* do final checking/adjustment of Y/M/D fields */
2330 6210 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
2331 6210 : if (dterr)
2332 0 : return dterr;
2333 :
2334 : /* handle AM/PM */
2335 6210 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
2336 0 : return DTERR_FIELD_OVERFLOW;
2337 6210 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
2338 0 : tm->tm_hour = 0;
2339 6210 : else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
2340 12 : tm->tm_hour += HOURS_PER_DAY / 2;
2341 :
2342 : /* check for time overflow */
2343 6210 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec, *fsec))
2344 72 : return DTERR_FIELD_OVERFLOW;
2345 :
2346 6138 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2347 0 : return DTERR_BAD_FORMAT;
2348 :
2349 : /*
2350 : * If we had a full timezone spec, compute the offset (we could not do it
2351 : * before, because we may need the date to resolve DST status).
2352 : */
2353 6138 : if (namedTz != NULL)
2354 : {
2355 : long int gmtoff;
2356 :
2357 : /* daylight savings time modifier disallowed with full TZ */
2358 1192 : if (fmask & DTK_M(DTZMOD))
2359 42 : return DTERR_BAD_FORMAT;
2360 :
2361 : /* if non-DST zone, we do not need to know the date */
2362 1192 : if (pg_get_timezone_offset(namedTz, &gmtoff))
2363 : {
2364 1114 : *tzp = -(int) gmtoff;
2365 : }
2366 : else
2367 : {
2368 : /* a date has to be specified */
2369 78 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2370 42 : return DTERR_BAD_FORMAT;
2371 36 : *tzp = DetermineTimeZoneOffset(tm, namedTz);
2372 : }
2373 : }
2374 :
2375 : /*
2376 : * Likewise, if we had a dynamic timezone abbreviation, resolve it now.
2377 : */
2378 6096 : if (abbrevTz != NULL)
2379 : {
2380 : struct pg_tm tt,
2381 0 : *tmp = &tt;
2382 :
2383 : /*
2384 : * daylight savings time modifier but no standard timezone? then error
2385 : */
2386 0 : if (fmask & DTK_M(DTZMOD))
2387 0 : return DTERR_BAD_FORMAT;
2388 :
2389 0 : if ((fmask & DTK_DATE_M) == 0)
2390 0 : GetCurrentDateTime(tmp);
2391 : else
2392 : {
2393 : /* a date has to be specified */
2394 0 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2395 0 : return DTERR_BAD_FORMAT;
2396 0 : tmp->tm_year = tm->tm_year;
2397 0 : tmp->tm_mon = tm->tm_mon;
2398 0 : tmp->tm_mday = tm->tm_mday;
2399 : }
2400 0 : tmp->tm_hour = tm->tm_hour;
2401 0 : tmp->tm_min = tm->tm_min;
2402 0 : tmp->tm_sec = tm->tm_sec;
2403 0 : *tzp = DetermineTimeZoneAbbrevOffset(tmp, abbrev, abbrevTz);
2404 0 : tm->tm_isdst = tmp->tm_isdst;
2405 : }
2406 :
2407 : /* timezone not specified? then use session timezone */
2408 6096 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
2409 : {
2410 : struct pg_tm tt,
2411 3176 : *tmp = &tt;
2412 :
2413 : /*
2414 : * daylight savings time modifier but no standard timezone? then error
2415 : */
2416 3176 : if (fmask & DTK_M(DTZMOD))
2417 0 : return DTERR_BAD_FORMAT;
2418 :
2419 3176 : if ((fmask & DTK_DATE_M) == 0)
2420 3098 : GetCurrentDateTime(tmp);
2421 : else
2422 : {
2423 : /* a date has to be specified */
2424 78 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2425 0 : return DTERR_BAD_FORMAT;
2426 78 : tmp->tm_year = tm->tm_year;
2427 78 : tmp->tm_mon = tm->tm_mon;
2428 78 : tmp->tm_mday = tm->tm_mday;
2429 : }
2430 3176 : tmp->tm_hour = tm->tm_hour;
2431 3176 : tmp->tm_min = tm->tm_min;
2432 3176 : tmp->tm_sec = tm->tm_sec;
2433 3176 : *tzp = DetermineTimeZoneOffset(tmp, session_timezone);
2434 3176 : tm->tm_isdst = tmp->tm_isdst;
2435 : }
2436 :
2437 6096 : return 0;
2438 : }
2439 :
2440 : /* DecodeDate()
2441 : * Decode date string which includes delimiters.
2442 : * Return 0 if okay, a DTERR code if not.
2443 : *
2444 : * str: field to be parsed
2445 : * fmask: bitmask for field types already seen
2446 : * *tmask: receives bitmask for fields found here
2447 : * *is2digits: set to true if we find 2-digit year
2448 : * *tm: field values are stored into appropriate members of this struct
2449 : */
2450 : static int
2451 69178 : DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
2452 : struct pg_tm *tm)
2453 : {
2454 : fsec_t fsec;
2455 69178 : int nf = 0;
2456 : int i,
2457 : len;
2458 : int dterr;
2459 69178 : bool haveTextMonth = false;
2460 : int type,
2461 : val,
2462 69178 : dmask = 0;
2463 : char *field[MAXDATEFIELDS];
2464 :
2465 69178 : *tmask = 0;
2466 :
2467 : /* parse this string... */
2468 276646 : while (*str != '\0' && nf < MAXDATEFIELDS)
2469 : {
2470 : /* skip field separators */
2471 207468 : while (*str != '\0' && !isalnum((unsigned char) *str))
2472 0 : str++;
2473 :
2474 207468 : if (*str == '\0')
2475 0 : return DTERR_BAD_FORMAT; /* end of string after separator */
2476 :
2477 207468 : field[nf] = str;
2478 207468 : if (isdigit((unsigned char) *str))
2479 : {
2480 760240 : while (isdigit((unsigned char) *str))
2481 552916 : str++;
2482 : }
2483 144 : else if (isalpha((unsigned char) *str))
2484 : {
2485 576 : while (isalpha((unsigned char) *str))
2486 432 : str++;
2487 : }
2488 :
2489 : /* Just get rid of any non-digit, non-alpha characters... */
2490 207468 : if (*str != '\0')
2491 138326 : *str++ = '\0';
2492 207468 : nf++;
2493 : }
2494 :
2495 : /* look first for text fields, since that will be unambiguous month */
2496 276646 : for (i = 0; i < nf; i++)
2497 : {
2498 207468 : if (isalpha((unsigned char) *field[i]))
2499 : {
2500 144 : type = DecodeSpecial(i, field[i], &val);
2501 144 : if (type == IGNORE_DTF)
2502 0 : continue;
2503 :
2504 144 : dmask = DTK_M(type);
2505 144 : switch (type)
2506 : {
2507 144 : case MONTH:
2508 144 : tm->tm_mon = val;
2509 144 : haveTextMonth = true;
2510 144 : break;
2511 :
2512 0 : default:
2513 0 : return DTERR_BAD_FORMAT;
2514 : }
2515 144 : if (fmask & dmask)
2516 0 : return DTERR_BAD_FORMAT;
2517 :
2518 144 : fmask |= dmask;
2519 144 : *tmask |= dmask;
2520 :
2521 : /* mark this field as being completed */
2522 144 : field[i] = NULL;
2523 : }
2524 : }
2525 :
2526 : /* now pick up remaining numeric fields */
2527 276646 : for (i = 0; i < nf; i++)
2528 : {
2529 207468 : if (field[i] == NULL)
2530 144 : continue;
2531 :
2532 207324 : if ((len = strlen(field[i])) <= 0)
2533 0 : return DTERR_BAD_FORMAT;
2534 :
2535 207324 : dterr = DecodeNumber(len, field[i], haveTextMonth, fmask,
2536 : &dmask, tm,
2537 : &fsec, is2digits);
2538 207324 : if (dterr)
2539 0 : return dterr;
2540 :
2541 207324 : if (fmask & dmask)
2542 0 : return DTERR_BAD_FORMAT;
2543 :
2544 207324 : fmask |= dmask;
2545 207324 : *tmask |= dmask;
2546 : }
2547 :
2548 69178 : if ((fmask & ~(DTK_M(DOY) | DTK_M(TZ))) != DTK_DATE_M)
2549 36 : return DTERR_BAD_FORMAT;
2550 :
2551 : /* validation of the field values must wait until ValidateDate() */
2552 :
2553 69142 : return 0;
2554 : }
2555 :
2556 : /* ValidateDate()
2557 : * Check valid year/month/day values, handle BC and DOY cases
2558 : * Return 0 if okay, a DTERR code if not.
2559 : */
2560 : int
2561 84966 : ValidateDate(int fmask, bool isjulian, bool is2digits, bool bc,
2562 : struct pg_tm *tm)
2563 : {
2564 84966 : if (fmask & DTK_M(YEAR))
2565 : {
2566 78960 : if (isjulian)
2567 : {
2568 : /* tm_year is correct and should not be touched */
2569 : }
2570 72846 : else if (bc)
2571 : {
2572 : /* there is no year zero in AD/BC notation */
2573 316 : if (tm->tm_year <= 0)
2574 0 : return DTERR_FIELD_OVERFLOW;
2575 : /* internally, we represent 1 BC as year zero, 2 BC as -1, etc */
2576 316 : tm->tm_year = -(tm->tm_year - 1);
2577 : }
2578 72530 : else if (is2digits)
2579 : {
2580 : /* process 1 or 2-digit input as 1970-2069 AD, allow '0' and '00' */
2581 354 : if (tm->tm_year < 0) /* just paranoia */
2582 0 : return DTERR_FIELD_OVERFLOW;
2583 354 : if (tm->tm_year < 70)
2584 174 : tm->tm_year += 2000;
2585 180 : else if (tm->tm_year < 100)
2586 180 : tm->tm_year += 1900;
2587 : }
2588 : else
2589 : {
2590 : /* there is no year zero in AD/BC notation */
2591 72176 : if (tm->tm_year <= 0)
2592 12 : return DTERR_FIELD_OVERFLOW;
2593 : }
2594 : }
2595 :
2596 : /* now that we have correct year, decode DOY */
2597 84954 : if (fmask & DTK_M(DOY))
2598 : {
2599 30 : j2date(date2j(tm->tm_year, 1, 1) + tm->tm_yday - 1,
2600 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2601 : }
2602 :
2603 : /* check for valid month */
2604 84954 : if (fmask & DTK_M(MONTH))
2605 : {
2606 78924 : if (tm->tm_mon < 1 || tm->tm_mon > MONTHS_PER_YEAR)
2607 78 : return DTERR_MD_FIELD_OVERFLOW;
2608 : }
2609 :
2610 : /* minimal check for valid day */
2611 84876 : if (fmask & DTK_M(DAY))
2612 : {
2613 78810 : if (tm->tm_mday < 1 || tm->tm_mday > 31)
2614 138 : return DTERR_MD_FIELD_OVERFLOW;
2615 : }
2616 :
2617 84738 : if ((fmask & DTK_DATE_M) == DTK_DATE_M)
2618 : {
2619 : /*
2620 : * Check for valid day of month, now that we know for sure the month
2621 : * and year. Note we don't use MD_FIELD_OVERFLOW here, since it seems
2622 : * unlikely that "Feb 29" is a YMD-order error.
2623 : */
2624 78654 : if (tm->tm_mday > day_tab[isleap(tm->tm_year)][tm->tm_mon - 1])
2625 48 : return DTERR_FIELD_OVERFLOW;
2626 : }
2627 :
2628 84690 : return 0;
2629 : }
2630 :
2631 :
2632 : /* DecodeTimeCommon()
2633 : * Decode time string which includes delimiters.
2634 : * Return 0 if okay, a DTERR code if not.
2635 : * tmask and itm are output parameters.
2636 : *
2637 : * This code is shared between the timestamp and interval cases.
2638 : * We return a struct pg_itm (of which only the tm_usec, tm_sec, tm_min,
2639 : * and tm_hour fields are used) and let the wrapper functions below
2640 : * convert and range-check as necessary.
2641 : */
2642 : static int
2643 66564 : DecodeTimeCommon(char *str, int fmask, int range,
2644 : int *tmask, struct pg_itm *itm)
2645 : {
2646 : char *cp;
2647 : int dterr;
2648 66564 : fsec_t fsec = 0;
2649 :
2650 66564 : *tmask = DTK_TIME_M;
2651 :
2652 66564 : errno = 0;
2653 66564 : itm->tm_hour = strtoi64(str, &cp, 10);
2654 66564 : if (errno == ERANGE)
2655 0 : return DTERR_FIELD_OVERFLOW;
2656 66564 : if (*cp != ':')
2657 0 : return DTERR_BAD_FORMAT;
2658 66564 : errno = 0;
2659 66564 : itm->tm_min = strtoint(cp + 1, &cp, 10);
2660 66564 : if (errno == ERANGE)
2661 0 : return DTERR_FIELD_OVERFLOW;
2662 66564 : if (*cp == '\0')
2663 : {
2664 1722 : itm->tm_sec = 0;
2665 : /* If it's a MINUTE TO SECOND interval, take 2 fields as being mm:ss */
2666 1722 : if (range == (INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND)))
2667 : {
2668 18 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2669 0 : return DTERR_FIELD_OVERFLOW;
2670 18 : itm->tm_sec = itm->tm_min;
2671 18 : itm->tm_min = (int) itm->tm_hour;
2672 18 : itm->tm_hour = 0;
2673 : }
2674 : }
2675 64842 : else if (*cp == '.')
2676 : {
2677 : /* always assume mm:ss.sss is MINUTE TO SECOND */
2678 48 : dterr = ParseFractionalSecond(cp, &fsec);
2679 48 : if (dterr)
2680 12 : return dterr;
2681 36 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2682 0 : return DTERR_FIELD_OVERFLOW;
2683 36 : itm->tm_sec = itm->tm_min;
2684 36 : itm->tm_min = (int) itm->tm_hour;
2685 36 : itm->tm_hour = 0;
2686 : }
2687 64794 : else if (*cp == ':')
2688 : {
2689 64794 : errno = 0;
2690 64794 : itm->tm_sec = strtoint(cp + 1, &cp, 10);
2691 64794 : if (errno == ERANGE)
2692 0 : return DTERR_FIELD_OVERFLOW;
2693 64794 : if (*cp == '.')
2694 : {
2695 23276 : dterr = ParseFractionalSecond(cp, &fsec);
2696 23276 : if (dterr)
2697 0 : return dterr;
2698 : }
2699 41518 : else if (*cp != '\0')
2700 0 : return DTERR_BAD_FORMAT;
2701 : }
2702 : else
2703 0 : return DTERR_BAD_FORMAT;
2704 :
2705 : /* do a sanity check; but caller must check the range of tm_hour */
2706 66552 : if (itm->tm_hour < 0 ||
2707 66552 : itm->tm_min < 0 || itm->tm_min > MINS_PER_HOUR - 1 ||
2708 66552 : itm->tm_sec < 0 || itm->tm_sec > SECS_PER_MINUTE ||
2709 66552 : fsec < 0 || fsec > USECS_PER_SEC)
2710 0 : return DTERR_FIELD_OVERFLOW;
2711 :
2712 66552 : itm->tm_usec = (int) fsec;
2713 :
2714 66552 : return 0;
2715 : }
2716 :
2717 : /* DecodeTime()
2718 : * Decode time string which includes delimiters.
2719 : * Return 0 if okay, a DTERR code if not.
2720 : *
2721 : * This version is used for timestamps. The results are returned into
2722 : * the tm_hour/tm_min/tm_sec fields of *tm, and microseconds into *fsec.
2723 : */
2724 : static int
2725 63518 : DecodeTime(char *str, int fmask, int range,
2726 : int *tmask, struct pg_tm *tm, fsec_t *fsec)
2727 : {
2728 : struct pg_itm itm;
2729 : int dterr;
2730 :
2731 63518 : dterr = DecodeTimeCommon(str, fmask, range,
2732 : tmask, &itm);
2733 63518 : if (dterr)
2734 0 : return dterr;
2735 :
2736 63518 : if (itm.tm_hour > INT_MAX)
2737 0 : return DTERR_FIELD_OVERFLOW;
2738 63518 : tm->tm_hour = (int) itm.tm_hour;
2739 63518 : tm->tm_min = itm.tm_min;
2740 63518 : tm->tm_sec = itm.tm_sec;
2741 63518 : *fsec = itm.tm_usec;
2742 :
2743 63518 : return 0;
2744 : }
2745 :
2746 : /* DecodeTimeForInterval()
2747 : * Decode time string which includes delimiters.
2748 : * Return 0 if okay, a DTERR code if not.
2749 : *
2750 : * This version is used for intervals. The results are returned into
2751 : * itm_in->tm_usec.
2752 : */
2753 : static int
2754 3046 : DecodeTimeForInterval(char *str, int fmask, int range,
2755 : int *tmask, struct pg_itm_in *itm_in)
2756 : {
2757 : struct pg_itm itm;
2758 : int dterr;
2759 :
2760 3046 : dterr = DecodeTimeCommon(str, fmask, range,
2761 : tmask, &itm);
2762 3046 : if (dterr)
2763 12 : return dterr;
2764 :
2765 3034 : itm_in->tm_usec = itm.tm_usec;
2766 3034 : if (!int64_multiply_add(itm.tm_hour, USECS_PER_HOUR, &itm_in->tm_usec) ||
2767 3034 : !int64_multiply_add(itm.tm_min, USECS_PER_MINUTE, &itm_in->tm_usec) ||
2768 3034 : !int64_multiply_add(itm.tm_sec, USECS_PER_SEC, &itm_in->tm_usec))
2769 6 : return DTERR_FIELD_OVERFLOW;
2770 :
2771 3028 : return 0;
2772 : }
2773 :
2774 :
2775 : /* DecodeNumber()
2776 : * Interpret plain numeric field as a date value in context.
2777 : * Return 0 if okay, a DTERR code if not.
2778 : */
2779 : static int
2780 213176 : DecodeNumber(int flen, char *str, bool haveTextMonth, int fmask,
2781 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2782 : {
2783 : int val;
2784 : char *cp;
2785 : int dterr;
2786 :
2787 213176 : *tmask = 0;
2788 :
2789 213176 : errno = 0;
2790 213176 : val = strtoint(str, &cp, 10);
2791 213176 : if (errno == ERANGE)
2792 0 : return DTERR_FIELD_OVERFLOW;
2793 213176 : if (cp == str)
2794 0 : return DTERR_BAD_FORMAT;
2795 :
2796 213176 : if (*cp == '.')
2797 : {
2798 : /*
2799 : * More than two digits before decimal point? Then could be a date or
2800 : * a run-together time: 2001.360 20011225 040506.789
2801 : */
2802 0 : if (cp - str > 2)
2803 : {
2804 0 : dterr = DecodeNumberField(flen, str,
2805 : (fmask | DTK_DATE_M),
2806 : tmask, tm,
2807 : fsec, is2digits);
2808 0 : if (dterr < 0)
2809 0 : return dterr;
2810 0 : return 0;
2811 : }
2812 :
2813 0 : dterr = ParseFractionalSecond(cp, fsec);
2814 0 : if (dterr)
2815 0 : return dterr;
2816 : }
2817 213176 : else if (*cp != '\0')
2818 0 : return DTERR_BAD_FORMAT;
2819 :
2820 : /* Special case for day of year */
2821 213176 : if (flen == 3 && (fmask & DTK_DATE_M) == DTK_M(YEAR) && val >= 1 &&
2822 : val <= 366)
2823 : {
2824 54 : *tmask = (DTK_M(DOY) | DTK_M(MONTH) | DTK_M(DAY));
2825 54 : tm->tm_yday = val;
2826 : /* tm_mon and tm_mday can't actually be set yet ... */
2827 54 : return 0;
2828 : }
2829 :
2830 : /* Switch based on what we have so far */
2831 213122 : switch (fmask & DTK_DATE_M)
2832 : {
2833 69368 : case 0:
2834 :
2835 : /*
2836 : * Nothing so far; make a decision about what we think the input
2837 : * is. There used to be lots of heuristics here, but the
2838 : * consensus now is to be paranoid. It *must* be either
2839 : * YYYY-MM-DD (with a more-than-two-digit year field), or the
2840 : * field order defined by DateOrder.
2841 : */
2842 69368 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2843 : {
2844 67616 : *tmask = DTK_M(YEAR);
2845 67616 : tm->tm_year = val;
2846 : }
2847 1752 : else if (DateOrder == DATEORDER_DMY)
2848 : {
2849 164 : *tmask = DTK_M(DAY);
2850 164 : tm->tm_mday = val;
2851 : }
2852 : else
2853 : {
2854 1588 : *tmask = DTK_M(MONTH);
2855 1588 : tm->tm_mon = val;
2856 : }
2857 69368 : break;
2858 :
2859 67508 : case (DTK_M(YEAR)):
2860 : /* Must be at second field of YY-MM-DD */
2861 67508 : *tmask = DTK_M(MONTH);
2862 67508 : tm->tm_mon = val;
2863 67508 : break;
2864 :
2865 4192 : case (DTK_M(MONTH)):
2866 4192 : if (haveTextMonth)
2867 : {
2868 : /*
2869 : * We are at the first numeric field of a date that included a
2870 : * textual month name. We want to support the variants
2871 : * MON-DD-YYYY, DD-MON-YYYY, and YYYY-MON-DD as unambiguous
2872 : * inputs. We will also accept MON-DD-YY or DD-MON-YY in
2873 : * either DMY or MDY modes, as well as YY-MON-DD in YMD mode.
2874 : */
2875 2660 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2876 : {
2877 72 : *tmask = DTK_M(YEAR);
2878 72 : tm->tm_year = val;
2879 : }
2880 : else
2881 : {
2882 2588 : *tmask = DTK_M(DAY);
2883 2588 : tm->tm_mday = val;
2884 : }
2885 : }
2886 : else
2887 : {
2888 : /* Must be at second field of MM-DD-YY */
2889 1532 : *tmask = DTK_M(DAY);
2890 1532 : tm->tm_mday = val;
2891 : }
2892 4192 : break;
2893 :
2894 67586 : case (DTK_M(YEAR) | DTK_M(MONTH)):
2895 67586 : if (haveTextMonth)
2896 : {
2897 : /* Need to accept DD-MON-YYYY even in YMD mode */
2898 126 : if (flen >= 3 && *is2digits)
2899 : {
2900 : /* Guess that first numeric field is day was wrong */
2901 30 : *tmask = DTK_M(DAY); /* YEAR is already set */
2902 30 : tm->tm_mday = tm->tm_year;
2903 30 : tm->tm_year = val;
2904 30 : *is2digits = false;
2905 : }
2906 : else
2907 : {
2908 96 : *tmask = DTK_M(DAY);
2909 96 : tm->tm_mday = val;
2910 : }
2911 : }
2912 : else
2913 : {
2914 : /* Must be at third field of YY-MM-DD */
2915 67460 : *tmask = DTK_M(DAY);
2916 67460 : tm->tm_mday = val;
2917 : }
2918 67586 : break;
2919 :
2920 146 : case (DTK_M(DAY)):
2921 : /* Must be at second field of DD-MM-YY */
2922 146 : *tmask = DTK_M(MONTH);
2923 146 : tm->tm_mon = val;
2924 146 : break;
2925 :
2926 4298 : case (DTK_M(MONTH) | DTK_M(DAY)):
2927 : /* Must be at third field of DD-MM-YY or MM-DD-YY */
2928 4298 : *tmask = DTK_M(YEAR);
2929 4298 : tm->tm_year = val;
2930 4298 : break;
2931 :
2932 24 : case (DTK_M(YEAR) | DTK_M(MONTH) | DTK_M(DAY)):
2933 : /* we have all the date, so it must be a time field */
2934 24 : dterr = DecodeNumberField(flen, str, fmask,
2935 : tmask, tm,
2936 : fsec, is2digits);
2937 24 : if (dterr < 0)
2938 12 : return dterr;
2939 12 : return 0;
2940 :
2941 0 : default:
2942 : /* Anything else is bogus input */
2943 0 : return DTERR_BAD_FORMAT;
2944 : }
2945 :
2946 : /*
2947 : * When processing a year field, mark it for adjustment if it's only one
2948 : * or two digits.
2949 : */
2950 213098 : if (*tmask == DTK_M(YEAR))
2951 71986 : *is2digits = (flen <= 2);
2952 :
2953 213098 : return 0;
2954 : }
2955 :
2956 :
2957 : /* DecodeNumberField()
2958 : * Interpret numeric string as a concatenated date or time field.
2959 : * Return a DTK token (>= 0) if successful, a DTERR code (< 0) if not.
2960 : *
2961 : * Use the context of previously decoded fields to help with
2962 : * the interpretation.
2963 : */
2964 : static int
2965 590 : DecodeNumberField(int len, char *str, int fmask,
2966 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2967 : {
2968 : char *cp;
2969 :
2970 : /*
2971 : * This function was originally meant to cope only with DTK_NUMBER fields,
2972 : * but we now sometimes abuse it to parse (parts of) DTK_DATE fields,
2973 : * which can contain letters and other punctuation. Reject if it's not a
2974 : * valid DTK_NUMBER, that is digits and decimal point(s). (ParseFraction
2975 : * will reject if there's more than one decimal point.)
2976 : */
2977 590 : if (strspn(str, "0123456789.") != len)
2978 12 : return DTERR_BAD_FORMAT;
2979 :
2980 : /*
2981 : * Have a decimal point? Then this is a date or something with a seconds
2982 : * field...
2983 : */
2984 578 : if ((cp = strchr(str, '.')) != NULL)
2985 : {
2986 : int dterr;
2987 :
2988 : /* Convert the fraction and store at *fsec */
2989 138 : dterr = ParseFractionalSecond(cp, fsec);
2990 138 : if (dterr)
2991 0 : return dterr;
2992 : /* Now truncate off the fraction for further processing */
2993 138 : *cp = '\0';
2994 138 : len = strlen(str);
2995 : }
2996 : /* No decimal point and no complete date yet? */
2997 440 : else if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2998 : {
2999 248 : if (len >= 6)
3000 : {
3001 248 : *tmask = DTK_DATE_M;
3002 :
3003 : /*
3004 : * Start from end and consider first 2 as Day, next 2 as Month,
3005 : * and the rest as Year.
3006 : */
3007 248 : tm->tm_mday = atoi(str + (len - 2));
3008 248 : *(str + (len - 2)) = '\0';
3009 248 : tm->tm_mon = atoi(str + (len - 4));
3010 248 : *(str + (len - 4)) = '\0';
3011 248 : tm->tm_year = atoi(str);
3012 248 : if ((len - 4) == 2)
3013 18 : *is2digits = true;
3014 :
3015 248 : return DTK_DATE;
3016 : }
3017 : }
3018 :
3019 : /* not all time fields are specified? */
3020 330 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
3021 : {
3022 : /* hhmmss */
3023 330 : if (len == 6)
3024 : {
3025 282 : *tmask = DTK_TIME_M;
3026 282 : tm->tm_sec = atoi(str + 4);
3027 282 : *(str + 4) = '\0';
3028 282 : tm->tm_min = atoi(str + 2);
3029 282 : *(str + 2) = '\0';
3030 282 : tm->tm_hour = atoi(str);
3031 :
3032 282 : return DTK_TIME;
3033 : }
3034 : /* hhmm? */
3035 48 : else if (len == 4)
3036 : {
3037 24 : *tmask = DTK_TIME_M;
3038 24 : tm->tm_sec = 0;
3039 24 : tm->tm_min = atoi(str + 2);
3040 24 : *(str + 2) = '\0';
3041 24 : tm->tm_hour = atoi(str);
3042 :
3043 24 : return DTK_TIME;
3044 : }
3045 : }
3046 :
3047 24 : return DTERR_BAD_FORMAT;
3048 : }
3049 :
3050 :
3051 : /* DecodeTimezone()
3052 : * Interpret string as a numeric timezone.
3053 : *
3054 : * Return 0 if okay (and set *tzp), a DTERR code if not okay.
3055 : */
3056 : int
3057 39806 : DecodeTimezone(const char *str, int *tzp)
3058 : {
3059 : int tz;
3060 : int hr,
3061 : min,
3062 39806 : sec = 0;
3063 : char *cp;
3064 :
3065 : /* leading character must be "+" or "-" */
3066 39806 : if (*str != '+' && *str != '-')
3067 72 : return DTERR_BAD_FORMAT;
3068 :
3069 39734 : errno = 0;
3070 39734 : hr = strtoint(str + 1, &cp, 10);
3071 39734 : if (errno == ERANGE)
3072 0 : return DTERR_TZDISP_OVERFLOW;
3073 :
3074 : /* explicit delimiter? */
3075 39734 : if (*cp == ':')
3076 : {
3077 108 : errno = 0;
3078 108 : min = strtoint(cp + 1, &cp, 10);
3079 108 : if (errno == ERANGE)
3080 0 : return DTERR_TZDISP_OVERFLOW;
3081 108 : if (*cp == ':')
3082 : {
3083 24 : errno = 0;
3084 24 : sec = strtoint(cp + 1, &cp, 10);
3085 24 : if (errno == ERANGE)
3086 0 : return DTERR_TZDISP_OVERFLOW;
3087 : }
3088 : }
3089 : /* otherwise, might have run things together... */
3090 39626 : else if (*cp == '\0' && strlen(str) > 3)
3091 : {
3092 72 : min = hr % 100;
3093 72 : hr = hr / 100;
3094 : /* we could, but don't, support a run-together hhmmss format */
3095 : }
3096 : else
3097 39554 : min = 0;
3098 :
3099 : /* Range-check the values; see notes in datatype/timestamp.h */
3100 39734 : if (hr < 0 || hr > MAX_TZDISP_HOUR)
3101 12 : return DTERR_TZDISP_OVERFLOW;
3102 39722 : if (min < 0 || min >= MINS_PER_HOUR)
3103 12 : return DTERR_TZDISP_OVERFLOW;
3104 39710 : if (sec < 0 || sec >= SECS_PER_MINUTE)
3105 0 : return DTERR_TZDISP_OVERFLOW;
3106 :
3107 39710 : tz = (hr * MINS_PER_HOUR + min) * SECS_PER_MINUTE + sec;
3108 39710 : if (*str == '-')
3109 914 : tz = -tz;
3110 :
3111 39710 : *tzp = -tz;
3112 :
3113 39710 : if (*cp != '\0')
3114 0 : return DTERR_BAD_FORMAT;
3115 :
3116 39710 : return 0;
3117 : }
3118 :
3119 :
3120 : /* DecodeTimezoneAbbrev()
3121 : * Interpret string as a timezone abbreviation, if possible.
3122 : *
3123 : * Sets *ftype to an abbreviation type (TZ, DTZ, or DYNTZ), or UNKNOWN_FIELD if
3124 : * string is not any known abbreviation. On success, set *offset and *tz to
3125 : * represent the UTC offset (for TZ or DTZ) or underlying zone (for DYNTZ).
3126 : * Note that full timezone names (such as America/New_York) are not handled
3127 : * here, mostly for historical reasons.
3128 : *
3129 : * The function result is 0 or a DTERR code; in the latter case, *extra
3130 : * is filled as needed. Note that unknown-abbreviation is not considered
3131 : * an error case. Also note that many callers assume that the DTERR code
3132 : * is one that DateTimeParseError does not require "str" or "datatype"
3133 : * strings for.
3134 : *
3135 : * Given string must be lowercased already.
3136 : *
3137 : * Implement a cache lookup since it is likely that dates
3138 : * will be related in format.
3139 : */
3140 : int
3141 10678 : DecodeTimezoneAbbrev(int field, const char *lowtoken,
3142 : int *ftype, int *offset, pg_tz **tz,
3143 : DateTimeErrorExtra *extra)
3144 : {
3145 10678 : TzAbbrevCache *tzc = &tzabbrevcache[field];
3146 : bool isfixed;
3147 : int isdst;
3148 : const datetkn *tp;
3149 :
3150 : /*
3151 : * Do we have a cached result? Use strncmp so that we match truncated
3152 : * names, although we shouldn't really see that happen with normal
3153 : * abbreviations.
3154 : */
3155 10678 : if (strncmp(lowtoken, tzc->abbrev, TOKMAXLEN) == 0)
3156 : {
3157 2714 : *ftype = tzc->ftype;
3158 2714 : *offset = tzc->offset;
3159 2714 : *tz = tzc->tz;
3160 2714 : return 0;
3161 : }
3162 :
3163 : /*
3164 : * See if the current session_timezone recognizes it. Checking this
3165 : * before zoneabbrevtbl allows us to correctly handle abbreviations whose
3166 : * meaning varies across zones, such as "LMT".
3167 : */
3168 15928 : if (session_timezone &&
3169 7964 : TimeZoneAbbrevIsKnown(lowtoken, session_timezone,
3170 : &isfixed, offset, &isdst))
3171 : {
3172 234 : *ftype = (isfixed ? (isdst ? DTZ : TZ) : DYNTZ);
3173 234 : *tz = (isfixed ? NULL : session_timezone);
3174 : /* flip sign to agree with the convention used in zoneabbrevtbl */
3175 234 : *offset = -(*offset);
3176 : /* cache result; use strlcpy to truncate name if necessary */
3177 234 : strlcpy(tzc->abbrev, lowtoken, TOKMAXLEN + 1);
3178 234 : tzc->ftype = *ftype;
3179 234 : tzc->offset = *offset;
3180 234 : tzc->tz = *tz;
3181 234 : return 0;
3182 : }
3183 :
3184 : /* Nope, so look in zoneabbrevtbl */
3185 7730 : if (zoneabbrevtbl)
3186 7730 : tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3187 7730 : zoneabbrevtbl->numabbrevs);
3188 : else
3189 0 : tp = NULL;
3190 7730 : if (tp == NULL)
3191 : {
3192 7494 : *ftype = UNKNOWN_FIELD;
3193 7494 : *offset = 0;
3194 7494 : *tz = NULL;
3195 : /* failure results are not cached */
3196 : }
3197 : else
3198 : {
3199 236 : *ftype = tp->type;
3200 236 : if (tp->type == DYNTZ)
3201 : {
3202 30 : *offset = 0;
3203 30 : *tz = FetchDynamicTimeZone(zoneabbrevtbl, tp, extra);
3204 30 : if (*tz == NULL)
3205 0 : return DTERR_BAD_ZONE_ABBREV;
3206 : }
3207 : else
3208 : {
3209 206 : *offset = tp->value;
3210 206 : *tz = NULL;
3211 : }
3212 :
3213 : /* cache result; use strlcpy to truncate name if necessary */
3214 236 : strlcpy(tzc->abbrev, lowtoken, TOKMAXLEN + 1);
3215 236 : tzc->ftype = *ftype;
3216 236 : tzc->offset = *offset;
3217 236 : tzc->tz = *tz;
3218 : }
3219 :
3220 7730 : return 0;
3221 : }
3222 :
3223 : /*
3224 : * Reset tzabbrevcache after a change in session_timezone.
3225 : */
3226 : void
3227 15434 : ClearTimeZoneAbbrevCache(void)
3228 : {
3229 15434 : memset(tzabbrevcache, 0, sizeof(tzabbrevcache));
3230 15434 : }
3231 :
3232 :
3233 : /* DecodeSpecial()
3234 : * Decode text string using lookup table.
3235 : *
3236 : * Recognizes the keywords listed in datetktbl.
3237 : * Note: at one time this would also recognize timezone abbreviations,
3238 : * but no more; use DecodeTimezoneAbbrev for that.
3239 : *
3240 : * Given string must be lowercased already.
3241 : *
3242 : * Implement a cache lookup since it is likely that dates
3243 : * will be related in format.
3244 : */
3245 : int
3246 42372 : DecodeSpecial(int field, const char *lowtoken, int *val)
3247 : {
3248 : int type;
3249 : const datetkn *tp;
3250 :
3251 42372 : tp = datecache[field];
3252 : /* use strncmp so that we match truncated tokens */
3253 42372 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3254 : {
3255 9852 : tp = datebsearch(lowtoken, datetktbl, szdatetktbl);
3256 : }
3257 42372 : if (tp == NULL)
3258 : {
3259 114 : type = UNKNOWN_FIELD;
3260 114 : *val = 0;
3261 : }
3262 : else
3263 : {
3264 42258 : datecache[field] = tp;
3265 42258 : type = tp->type;
3266 42258 : *val = tp->value;
3267 : }
3268 :
3269 42372 : return type;
3270 : }
3271 :
3272 :
3273 : /* DecodeTimezoneName()
3274 : * Interpret string as a timezone abbreviation or name.
3275 : * Throw error if the name is not recognized.
3276 : *
3277 : * The return value indicates what kind of zone identifier it is:
3278 : * TZNAME_FIXED_OFFSET: fixed offset from UTC
3279 : * TZNAME_DYNTZ: dynamic timezone abbreviation
3280 : * TZNAME_ZONE: full tzdb zone name
3281 : *
3282 : * For TZNAME_FIXED_OFFSET, *offset receives the UTC offset (in seconds,
3283 : * with ISO sign convention: positive is east of Greenwich).
3284 : * For the other two cases, *tz receives the timezone struct representing
3285 : * the zone name or the abbreviation's underlying zone.
3286 : */
3287 : int
3288 822 : DecodeTimezoneName(const char *tzname, int *offset, pg_tz **tz)
3289 : {
3290 : char *lowzone;
3291 : int dterr,
3292 : type;
3293 : DateTimeErrorExtra extra;
3294 :
3295 : /*
3296 : * First we look in the timezone abbreviation table (to handle cases like
3297 : * "EST"), and if that fails, we look in the timezone database (to handle
3298 : * cases like "America/New_York"). This matches the order in which
3299 : * timestamp input checks the cases; it's important because the timezone
3300 : * database unwisely uses a few zone names that are identical to offset
3301 : * abbreviations.
3302 : */
3303 :
3304 : /* DecodeTimezoneAbbrev requires lowercase input */
3305 822 : lowzone = downcase_truncate_identifier(tzname,
3306 822 : strlen(tzname),
3307 : false);
3308 :
3309 822 : dterr = DecodeTimezoneAbbrev(0, lowzone, &type, offset, tz, &extra);
3310 822 : if (dterr)
3311 0 : DateTimeParseError(dterr, &extra, NULL, NULL, NULL);
3312 :
3313 822 : if (type == TZ || type == DTZ)
3314 : {
3315 : /* fixed-offset abbreviation, return the offset */
3316 294 : return TZNAME_FIXED_OFFSET;
3317 : }
3318 528 : else if (type == DYNTZ)
3319 : {
3320 : /* dynamic-offset abbreviation, return its referenced timezone */
3321 174 : return TZNAME_DYNTZ;
3322 : }
3323 : else
3324 : {
3325 : /* try it as a full zone name */
3326 354 : *tz = pg_tzset(tzname);
3327 354 : if (*tz == NULL)
3328 12 : ereport(ERROR,
3329 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
3330 : errmsg("time zone \"%s\" not recognized", tzname)));
3331 342 : return TZNAME_ZONE;
3332 : }
3333 : }
3334 :
3335 : /* DecodeTimezoneNameToTz()
3336 : * Interpret string as a timezone abbreviation or name.
3337 : * Throw error if the name is not recognized.
3338 : *
3339 : * This is a simple wrapper for DecodeTimezoneName that produces a pg_tz *
3340 : * result in all cases.
3341 : */
3342 : pg_tz *
3343 84 : DecodeTimezoneNameToTz(const char *tzname)
3344 : {
3345 : pg_tz *result;
3346 : int offset;
3347 :
3348 84 : if (DecodeTimezoneName(tzname, &offset, &result) == TZNAME_FIXED_OFFSET)
3349 : {
3350 : /* fixed-offset abbreviation, get a pg_tz descriptor for that */
3351 18 : result = pg_tzset_offset(-offset); /* flip to POSIX sign convention */
3352 : }
3353 84 : return result;
3354 : }
3355 :
3356 : /* DecodeTimezoneAbbrevPrefix()
3357 : * Interpret prefix of string as a timezone abbreviation, if possible.
3358 : *
3359 : * This has roughly the same functionality as DecodeTimezoneAbbrev(),
3360 : * but the API is adapted to the needs of formatting.c. Notably,
3361 : * we will match the longest possible prefix of the given string
3362 : * rather than insisting on a complete match, and downcasing is applied
3363 : * here rather than in the caller.
3364 : *
3365 : * Returns the length of the timezone abbreviation, or -1 if not recognized.
3366 : * On success, sets *offset to the GMT offset for the abbreviation if it
3367 : * is a fixed-offset abbreviation, or sets *tz to the pg_tz struct for
3368 : * a dynamic abbreviation.
3369 : */
3370 : int
3371 3558 : DecodeTimezoneAbbrevPrefix(const char *str, int *offset, pg_tz **tz)
3372 : {
3373 : char lowtoken[TOKMAXLEN + 1];
3374 : int len;
3375 :
3376 3558 : *offset = 0; /* avoid uninitialized vars on failure */
3377 3558 : *tz = NULL;
3378 :
3379 : /* Downcase as much of the string as we could need */
3380 3696 : for (len = 0; len < TOKMAXLEN; len++)
3381 : {
3382 3696 : if (*str == '\0' || !isalpha((unsigned char) *str))
3383 : break;
3384 138 : lowtoken[len] = pg_tolower((unsigned char) *str++);
3385 : }
3386 3558 : lowtoken[len] = '\0';
3387 :
3388 : /*
3389 : * We could avoid doing repeated binary searches if we cared to duplicate
3390 : * datebsearch here, but it's not clear that such an optimization would be
3391 : * worth the trouble. In common cases there's probably not anything after
3392 : * the zone abbrev anyway. So just search with successively truncated
3393 : * strings.
3394 : */
3395 3600 : while (len > 0)
3396 : {
3397 : bool isfixed;
3398 : int isdst;
3399 : const datetkn *tp;
3400 :
3401 : /* See if the current session_timezone recognizes it. */
3402 156 : if (session_timezone &&
3403 78 : TimeZoneAbbrevIsKnown(lowtoken, session_timezone,
3404 : &isfixed, offset, &isdst))
3405 : {
3406 6 : if (isfixed)
3407 : {
3408 : /* flip sign to agree with the convention in zoneabbrevtbl */
3409 6 : *offset = -(*offset);
3410 : }
3411 : else
3412 : {
3413 : /* Caller must resolve the abbrev's current meaning */
3414 0 : *tz = session_timezone;
3415 : }
3416 36 : return len;
3417 : }
3418 :
3419 : /* Known in zoneabbrevtbl? */
3420 72 : if (zoneabbrevtbl)
3421 72 : tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3422 72 : zoneabbrevtbl->numabbrevs);
3423 : else
3424 0 : tp = NULL;
3425 72 : if (tp != NULL)
3426 : {
3427 30 : if (tp->type == DYNTZ)
3428 : {
3429 : DateTimeErrorExtra extra;
3430 6 : pg_tz *tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp,
3431 : &extra);
3432 :
3433 6 : if (tzp != NULL)
3434 : {
3435 : /* Caller must resolve the abbrev's current meaning */
3436 6 : *tz = tzp;
3437 6 : return len;
3438 : }
3439 : }
3440 : else
3441 : {
3442 : /* Fixed-offset zone abbrev, so it's easy */
3443 24 : *offset = tp->value;
3444 24 : return len;
3445 : }
3446 : }
3447 :
3448 : /* Nope, try the next shorter string. */
3449 42 : lowtoken[--len] = '\0';
3450 : }
3451 :
3452 : /* Did not find a match */
3453 3522 : return -1;
3454 : }
3455 :
3456 :
3457 : /* ClearPgItmIn
3458 : *
3459 : * Zero out a pg_itm_in
3460 : */
3461 : static inline void
3462 67014 : ClearPgItmIn(struct pg_itm_in *itm_in)
3463 : {
3464 67014 : itm_in->tm_usec = 0;
3465 67014 : itm_in->tm_mday = 0;
3466 67014 : itm_in->tm_mon = 0;
3467 67014 : itm_in->tm_year = 0;
3468 67014 : }
3469 :
3470 :
3471 : /* DecodeInterval()
3472 : * Interpret previously parsed fields for general time interval.
3473 : * Returns 0 if successful, DTERR code if bogus input detected.
3474 : * dtype and itm_in are output parameters.
3475 : *
3476 : * Allow "date" field DTK_DATE since this could be just
3477 : * an unsigned floating point number. - thomas 1997-11-16
3478 : *
3479 : * Allow ISO-style time span, with implicit units on number of days
3480 : * preceding an hh:mm:ss field. - thomas 1998-04-30
3481 : *
3482 : * itm_in remains undefined for infinite interval values for which dtype alone
3483 : * suffices.
3484 : */
3485 : int
3486 66402 : DecodeInterval(char **field, int *ftype, int nf, int range,
3487 : int *dtype, struct pg_itm_in *itm_in)
3488 : {
3489 66402 : bool force_negative = false;
3490 66402 : bool is_before = false;
3491 66402 : bool parsing_unit_val = false;
3492 : char *cp;
3493 66402 : int fmask = 0,
3494 : tmask,
3495 : type,
3496 : uval;
3497 : int i;
3498 : int dterr;
3499 : int64 val;
3500 : double fval;
3501 :
3502 66402 : *dtype = DTK_DELTA;
3503 66402 : type = IGNORE_DTF;
3504 66402 : ClearPgItmIn(itm_in);
3505 :
3506 : /*----------
3507 : * The SQL standard defines the interval literal
3508 : * '-1 1:00:00'
3509 : * to mean "negative 1 days and negative 1 hours", while Postgres
3510 : * traditionally treats this as meaning "negative 1 days and positive
3511 : * 1 hours". In SQL_STANDARD intervalstyle, we apply the leading sign
3512 : * to all fields if there are no other explicit signs.
3513 : *
3514 : * We leave the signs alone if there are additional explicit signs.
3515 : * This protects us against misinterpreting postgres-style dump output,
3516 : * since the postgres-style output code has always put an explicit sign on
3517 : * all fields following a negative field. But note that SQL-spec output
3518 : * is ambiguous and can be misinterpreted on load! (So it's best practice
3519 : * to dump in postgres style, not SQL style.)
3520 : *----------
3521 : */
3522 66402 : if (IntervalStyle == INTSTYLE_SQL_STANDARD && nf > 0 && *field[0] == '-')
3523 : {
3524 38 : force_negative = true;
3525 : /* Check for additional explicit signs */
3526 244 : for (i = 1; i < nf; i++)
3527 : {
3528 224 : if (*field[i] == '-' || *field[i] == '+')
3529 : {
3530 18 : force_negative = false;
3531 18 : break;
3532 : }
3533 : }
3534 : }
3535 :
3536 : /* read through list backwards to pick up units before values */
3537 203198 : for (i = nf - 1; i >= 0; i--)
3538 : {
3539 137888 : switch (ftype[i])
3540 : {
3541 2276 : case DTK_TIME:
3542 2276 : dterr = DecodeTimeForInterval(field[i], fmask, range,
3543 : &tmask, itm_in);
3544 2276 : if (dterr)
3545 18 : return dterr;
3546 2258 : if (force_negative &&
3547 2 : itm_in->tm_usec > 0)
3548 2 : itm_in->tm_usec = -itm_in->tm_usec;
3549 2258 : type = DTK_DAY;
3550 2258 : parsing_unit_val = false;
3551 2258 : break;
3552 :
3553 3060 : case DTK_TZ:
3554 :
3555 : /*
3556 : * Timezone means a token with a leading sign character and at
3557 : * least one digit; there could be ':', '.', '-' embedded in
3558 : * it as well.
3559 : */
3560 : Assert(*field[i] == '-' || *field[i] == '+');
3561 :
3562 : /*
3563 : * Check for signed hh:mm or hh:mm:ss. If so, process exactly
3564 : * like DTK_TIME case above, plus handling the sign.
3565 : */
3566 3830 : if (strchr(field[i] + 1, ':') != NULL &&
3567 770 : DecodeTimeForInterval(field[i] + 1, fmask, range,
3568 : &tmask, itm_in) == 0)
3569 : {
3570 770 : if (*field[i] == '-')
3571 : {
3572 : /* flip the sign on time field */
3573 710 : if (itm_in->tm_usec == PG_INT64_MIN)
3574 0 : return DTERR_FIELD_OVERFLOW;
3575 710 : itm_in->tm_usec = -itm_in->tm_usec;
3576 : }
3577 :
3578 770 : if (force_negative &&
3579 0 : itm_in->tm_usec > 0)
3580 0 : itm_in->tm_usec = -itm_in->tm_usec;
3581 :
3582 : /*
3583 : * Set the next type to be a day, if units are not
3584 : * specified. This handles the case of '1 +02:03' since we
3585 : * are reading right to left.
3586 : */
3587 770 : type = DTK_DAY;
3588 770 : parsing_unit_val = false;
3589 770 : break;
3590 : }
3591 :
3592 : /*
3593 : * Otherwise, fall through to DTK_NUMBER case, which can
3594 : * handle signed float numbers and signed year-month values.
3595 : */
3596 :
3597 : /* FALLTHROUGH */
3598 :
3599 : case DTK_DATE:
3600 : case DTK_NUMBER:
3601 67274 : if (type == IGNORE_DTF)
3602 : {
3603 : /* use typmod to decide what rightmost field is */
3604 : switch (range)
3605 : {
3606 6 : case INTERVAL_MASK(YEAR):
3607 6 : type = DTK_YEAR;
3608 6 : break;
3609 30 : case INTERVAL_MASK(MONTH):
3610 : case INTERVAL_MASK(YEAR) | INTERVAL_MASK(MONTH):
3611 30 : type = DTK_MONTH;
3612 30 : break;
3613 18 : case INTERVAL_MASK(DAY):
3614 18 : type = DTK_DAY;
3615 18 : break;
3616 24 : case INTERVAL_MASK(HOUR):
3617 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR):
3618 24 : type = DTK_HOUR;
3619 24 : break;
3620 24 : case INTERVAL_MASK(MINUTE):
3621 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3622 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3623 24 : type = DTK_MINUTE;
3624 24 : break;
3625 60 : case INTERVAL_MASK(SECOND):
3626 : case INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3627 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3628 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3629 60 : type = DTK_SECOND;
3630 60 : break;
3631 438 : default:
3632 438 : type = DTK_SECOND;
3633 438 : break;
3634 : }
3635 : }
3636 :
3637 67274 : errno = 0;
3638 67274 : val = strtoi64(field[i], &cp, 10);
3639 67274 : if (errno == ERANGE)
3640 12 : return DTERR_FIELD_OVERFLOW;
3641 :
3642 67262 : if (*cp == '-')
3643 : {
3644 : /* SQL "years-months" syntax */
3645 : int val2;
3646 :
3647 60 : val2 = strtoint(cp + 1, &cp, 10);
3648 60 : if (errno == ERANGE || val2 < 0 || val2 >= MONTHS_PER_YEAR)
3649 0 : return DTERR_FIELD_OVERFLOW;
3650 60 : if (*cp != '\0')
3651 0 : return DTERR_BAD_FORMAT;
3652 60 : type = DTK_MONTH;
3653 60 : if (*field[i] == '-')
3654 6 : val2 = -val2;
3655 60 : if (pg_mul_s64_overflow(val, MONTHS_PER_YEAR, &val))
3656 0 : return DTERR_FIELD_OVERFLOW;
3657 60 : if (pg_add_s64_overflow(val, val2, &val))
3658 0 : return DTERR_FIELD_OVERFLOW;
3659 60 : fval = 0;
3660 : }
3661 67202 : else if (*cp == '.')
3662 : {
3663 486 : dterr = ParseFraction(cp, &fval);
3664 486 : if (dterr)
3665 0 : return dterr;
3666 486 : if (*field[i] == '-')
3667 138 : fval = -fval;
3668 : }
3669 66716 : else if (*cp == '\0')
3670 66332 : fval = 0;
3671 : else
3672 384 : return DTERR_BAD_FORMAT;
3673 :
3674 66878 : tmask = 0; /* DTK_M(type); */
3675 :
3676 66878 : if (force_negative)
3677 : {
3678 : /* val and fval should be of same sign, but test anyway */
3679 80 : if (val > 0)
3680 60 : val = -val;
3681 80 : if (fval > 0)
3682 18 : fval = -fval;
3683 : }
3684 :
3685 : switch (type)
3686 : {
3687 330 : case DTK_MICROSEC:
3688 330 : if (!AdjustMicroseconds(val, fval, 1, itm_in))
3689 36 : return DTERR_FIELD_OVERFLOW;
3690 294 : tmask = DTK_M(MICROSECOND);
3691 294 : break;
3692 :
3693 102 : case DTK_MILLISEC:
3694 102 : if (!AdjustMicroseconds(val, fval, 1000, itm_in))
3695 12 : return DTERR_FIELD_OVERFLOW;
3696 90 : tmask = DTK_M(MILLISECOND);
3697 90 : break;
3698 :
3699 920 : case DTK_SECOND:
3700 920 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
3701 12 : return DTERR_FIELD_OVERFLOW;
3702 :
3703 : /*
3704 : * If any subseconds were specified, consider this
3705 : * microsecond and millisecond input as well.
3706 : */
3707 908 : if (fval == 0)
3708 746 : tmask = DTK_M(SECOND);
3709 : else
3710 162 : tmask = DTK_ALL_SECS_M;
3711 908 : break;
3712 :
3713 370 : case DTK_MINUTE:
3714 370 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
3715 12 : return DTERR_FIELD_OVERFLOW;
3716 358 : tmask = DTK_M(MINUTE);
3717 358 : break;
3718 :
3719 692 : case DTK_HOUR:
3720 692 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3721 12 : return DTERR_FIELD_OVERFLOW;
3722 680 : tmask = DTK_M(HOUR);
3723 680 : type = DTK_DAY; /* set for next field */
3724 680 : break;
3725 :
3726 8074 : case DTK_DAY:
3727 8074 : if (!AdjustDays(val, 1, itm_in) ||
3728 8002 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3729 90 : return DTERR_FIELD_OVERFLOW;
3730 7984 : tmask = DTK_M(DAY);
3731 7984 : break;
3732 :
3733 96 : case DTK_WEEK:
3734 96 : if (!AdjustDays(val, 7, itm_in) ||
3735 72 : !AdjustFractDays(fval, 7, itm_in))
3736 48 : return DTERR_FIELD_OVERFLOW;
3737 48 : tmask = DTK_M(WEEK);
3738 48 : break;
3739 :
3740 1590 : case DTK_MONTH:
3741 1590 : if (!AdjustMonths(val, itm_in) ||
3742 1530 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3743 84 : return DTERR_FIELD_OVERFLOW;
3744 1506 : tmask = DTK_M(MONTH);
3745 1506 : break;
3746 :
3747 54506 : case DTK_YEAR:
3748 54506 : if (!AdjustYears(val, 1, itm_in) ||
3749 54458 : !AdjustFractYears(fval, 1, itm_in))
3750 60 : return DTERR_FIELD_OVERFLOW;
3751 54446 : tmask = DTK_M(YEAR);
3752 54446 : break;
3753 :
3754 66 : case DTK_DECADE:
3755 66 : if (!AdjustYears(val, 10, itm_in) ||
3756 42 : !AdjustFractYears(fval, 10, itm_in))
3757 36 : return DTERR_FIELD_OVERFLOW;
3758 30 : tmask = DTK_M(DECADE);
3759 30 : break;
3760 :
3761 66 : case DTK_CENTURY:
3762 66 : if (!AdjustYears(val, 100, itm_in) ||
3763 42 : !AdjustFractYears(fval, 100, itm_in))
3764 36 : return DTERR_FIELD_OVERFLOW;
3765 30 : tmask = DTK_M(CENTURY);
3766 30 : break;
3767 :
3768 66 : case DTK_MILLENNIUM:
3769 66 : if (!AdjustYears(val, 1000, itm_in) ||
3770 42 : !AdjustFractYears(fval, 1000, itm_in))
3771 36 : return DTERR_FIELD_OVERFLOW;
3772 30 : tmask = DTK_M(MILLENNIUM);
3773 30 : break;
3774 :
3775 0 : default:
3776 0 : return DTERR_BAD_FORMAT;
3777 : }
3778 66404 : parsing_unit_val = false;
3779 66404 : break;
3780 :
3781 67568 : case DTK_STRING:
3782 : case DTK_SPECIAL:
3783 : /* reject consecutive unhandled units */
3784 67568 : if (parsing_unit_val)
3785 12 : return DTERR_BAD_FORMAT;
3786 67556 : type = DecodeUnits(i, field[i], &uval);
3787 67556 : if (type == UNKNOWN_FIELD)
3788 1070 : type = DecodeSpecial(i, field[i], &uval);
3789 67556 : if (type == IGNORE_DTF)
3790 0 : continue;
3791 :
3792 67556 : tmask = 0; /* DTK_M(type); */
3793 : switch (type)
3794 : {
3795 66384 : case UNITS:
3796 66384 : type = uval;
3797 66384 : parsing_unit_val = true;
3798 66384 : break;
3799 :
3800 102 : case AGO:
3801 :
3802 : /*
3803 : * "ago" is only allowed to appear at the end of the
3804 : * interval.
3805 : */
3806 102 : if (i != nf - 1)
3807 12 : return DTERR_BAD_FORMAT;
3808 90 : is_before = true;
3809 90 : type = uval;
3810 90 : break;
3811 :
3812 1034 : case RESERV:
3813 1034 : tmask = (DTK_DATE_M | DTK_TIME_M);
3814 :
3815 : /*
3816 : * Only reserved words corresponding to infinite
3817 : * intervals are accepted.
3818 : */
3819 1034 : if (uval != DTK_LATE && uval != DTK_EARLY)
3820 36 : return DTERR_BAD_FORMAT;
3821 :
3822 : /*
3823 : * Infinity cannot be followed by anything else. We
3824 : * could allow "ago" to reverse the sign of infinity
3825 : * but using signed infinity is more intuitive.
3826 : */
3827 998 : if (i != nf - 1)
3828 12 : return DTERR_BAD_FORMAT;
3829 :
3830 986 : *dtype = uval;
3831 986 : break;
3832 :
3833 36 : default:
3834 36 : return DTERR_BAD_FORMAT;
3835 : }
3836 67460 : break;
3837 :
3838 0 : default:
3839 0 : return DTERR_BAD_FORMAT;
3840 : }
3841 :
3842 136892 : if (tmask & fmask)
3843 96 : return DTERR_BAD_FORMAT;
3844 136796 : fmask |= tmask;
3845 : }
3846 :
3847 : /* ensure that at least one time field has been found */
3848 65310 : if (fmask == 0)
3849 6 : return DTERR_BAD_FORMAT;
3850 :
3851 : /* reject if unit appeared and was never handled */
3852 65304 : if (parsing_unit_val)
3853 6 : return DTERR_BAD_FORMAT;
3854 :
3855 : /* finally, AGO negates everything */
3856 65298 : if (is_before)
3857 : {
3858 42 : if (itm_in->tm_usec == PG_INT64_MIN ||
3859 30 : itm_in->tm_mday == INT_MIN ||
3860 24 : itm_in->tm_mon == INT_MIN ||
3861 18 : itm_in->tm_year == INT_MIN)
3862 24 : return DTERR_FIELD_OVERFLOW;
3863 :
3864 18 : itm_in->tm_usec = -itm_in->tm_usec;
3865 18 : itm_in->tm_mday = -itm_in->tm_mday;
3866 18 : itm_in->tm_mon = -itm_in->tm_mon;
3867 18 : itm_in->tm_year = -itm_in->tm_year;
3868 : }
3869 :
3870 65274 : return 0;
3871 : }
3872 :
3873 :
3874 : /*
3875 : * Helper functions to avoid duplicated code in DecodeISO8601Interval.
3876 : *
3877 : * Parse a decimal value and break it into integer and fractional parts.
3878 : * Set *endptr to end+1 of the parsed substring.
3879 : * Returns 0 or DTERR code.
3880 : */
3881 : static int
3882 954 : ParseISO8601Number(char *str, char **endptr, int64 *ipart, double *fpart)
3883 : {
3884 : double val;
3885 :
3886 : /*
3887 : * Historically this has accepted anything that strtod() would take,
3888 : * notably including "e" notation, so continue doing that. This is
3889 : * slightly annoying because the precision of double is less than that of
3890 : * int64, so we would lose accuracy for inputs larger than 2^53 or so.
3891 : * However, historically we rejected inputs outside the int32 range,
3892 : * making that concern moot. What we do now is reject abs(val) above
3893 : * 1.0e15 (a round number a bit less than 2^50), so that any accepted
3894 : * value will have an exact integer part, and thereby a fraction part with
3895 : * abs(*fpart) less than 1. In the absence of field complaints it doesn't
3896 : * seem worth working harder.
3897 : */
3898 954 : if (!(isdigit((unsigned char) *str) || *str == '-' || *str == '.'))
3899 0 : return DTERR_BAD_FORMAT;
3900 954 : errno = 0;
3901 954 : val = strtod(str, endptr);
3902 : /* did we not see anything that looks like a double? */
3903 954 : if (*endptr == str || errno != 0)
3904 6 : return DTERR_BAD_FORMAT;
3905 : /* watch out for overflow, including infinities; reject NaN too */
3906 948 : if (isnan(val) || val < -1.0e15 || val > 1.0e15)
3907 0 : return DTERR_FIELD_OVERFLOW;
3908 : /* be very sure we truncate towards zero (cf dtrunc()) */
3909 948 : if (val >= 0)
3910 732 : *ipart = (int64) floor(val);
3911 : else
3912 216 : *ipart = (int64) -floor(-val);
3913 948 : *fpart = val - *ipart;
3914 : /* Callers expect this to hold */
3915 : Assert(*fpart > -1.0 && *fpart < 1.0);
3916 948 : return 0;
3917 : }
3918 :
3919 : /*
3920 : * Determine number of integral digits in a valid ISO 8601 number field
3921 : * (we should ignore sign and any fraction part)
3922 : */
3923 : static int
3924 66 : ISO8601IntegerWidth(char *fieldstart)
3925 : {
3926 : /* We might have had a leading '-' */
3927 66 : if (*fieldstart == '-')
3928 18 : fieldstart++;
3929 66 : return strspn(fieldstart, "0123456789");
3930 : }
3931 :
3932 :
3933 : /* DecodeISO8601Interval()
3934 : * Decode an ISO 8601 time interval of the "format with designators"
3935 : * (section 4.4.3.2) or "alternative format" (section 4.4.3.3)
3936 : * Examples: P1D for 1 day
3937 : * PT1H for 1 hour
3938 : * P2Y6M7DT1H30M for 2 years, 6 months, 7 days 1 hour 30 min
3939 : * P0002-06-07T01:30:00 the same value in alternative format
3940 : *
3941 : * Returns 0 if successful, DTERR code if bogus input detected.
3942 : * Note: error code should be DTERR_BAD_FORMAT if input doesn't look like
3943 : * ISO8601, otherwise this could cause unexpected error messages.
3944 : * dtype and itm_in are output parameters.
3945 : *
3946 : * A couple exceptions from the spec:
3947 : * - a week field ('W') may coexist with other units
3948 : * - allows decimals in fields other than the least significant unit.
3949 : */
3950 : int
3951 612 : DecodeISO8601Interval(char *str,
3952 : int *dtype, struct pg_itm_in *itm_in)
3953 : {
3954 612 : bool datepart = true;
3955 612 : bool havefield = false;
3956 :
3957 612 : *dtype = DTK_DELTA;
3958 612 : ClearPgItmIn(itm_in);
3959 :
3960 612 : if (strlen(str) < 2 || str[0] != 'P')
3961 228 : return DTERR_BAD_FORMAT;
3962 :
3963 384 : str++;
3964 1122 : while (*str)
3965 : {
3966 : char *fieldstart;
3967 : int64 val;
3968 : double fval;
3969 : char unit;
3970 : int dterr;
3971 :
3972 1014 : if (*str == 'T') /* T indicates the beginning of the time part */
3973 : {
3974 198 : datepart = false;
3975 198 : havefield = false;
3976 198 : str++;
3977 240 : continue;
3978 : }
3979 :
3980 816 : fieldstart = str;
3981 816 : dterr = ParseISO8601Number(str, &str, &val, &fval);
3982 816 : if (dterr)
3983 276 : return dterr;
3984 :
3985 : /*
3986 : * Note: we could step off the end of the string here. Code below
3987 : * *must* exit the loop if unit == '\0'.
3988 : */
3989 810 : unit = *str++;
3990 :
3991 810 : if (datepart)
3992 : {
3993 468 : switch (unit) /* before T: Y M W D */
3994 : {
3995 84 : case 'Y':
3996 84 : if (!AdjustYears(val, 1, itm_in) ||
3997 84 : !AdjustFractYears(fval, 1, itm_in))
3998 12 : return DTERR_FIELD_OVERFLOW;
3999 72 : break;
4000 108 : case 'M':
4001 108 : if (!AdjustMonths(val, itm_in) ||
4002 96 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
4003 24 : return DTERR_FIELD_OVERFLOW;
4004 84 : break;
4005 54 : case 'W':
4006 54 : if (!AdjustDays(val, 7, itm_in) ||
4007 42 : !AdjustFractDays(fval, 7, itm_in))
4008 24 : return DTERR_FIELD_OVERFLOW;
4009 30 : break;
4010 132 : case 'D':
4011 132 : if (!AdjustDays(val, 1, itm_in) ||
4012 96 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
4013 36 : return DTERR_FIELD_OVERFLOW;
4014 96 : break;
4015 30 : case 'T': /* ISO 8601 4.4.3.3 Alternative Format / Basic */
4016 : case '\0':
4017 30 : if (ISO8601IntegerWidth(fieldstart) == 8 && !havefield)
4018 : {
4019 6 : if (!AdjustYears(val / 10000, 1, itm_in) ||
4020 6 : !AdjustMonths((val / 100) % 100, itm_in) ||
4021 6 : !AdjustDays(val % 100, 1, itm_in) ||
4022 6 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
4023 0 : return DTERR_FIELD_OVERFLOW;
4024 6 : if (unit == '\0')
4025 0 : return 0;
4026 6 : datepart = false;
4027 6 : havefield = false;
4028 6 : continue;
4029 : }
4030 : /* Else fall through to extended alternative format */
4031 : /* FALLTHROUGH */
4032 : case '-': /* ISO 8601 4.4.3.3 Alternative Format,
4033 : * Extended */
4034 84 : if (havefield)
4035 0 : return DTERR_BAD_FORMAT;
4036 :
4037 84 : if (!AdjustYears(val, 1, itm_in) ||
4038 72 : !AdjustFractYears(fval, 1, itm_in))
4039 12 : return DTERR_FIELD_OVERFLOW;
4040 72 : if (unit == '\0')
4041 6 : return 0;
4042 66 : if (unit == 'T')
4043 : {
4044 6 : datepart = false;
4045 6 : havefield = false;
4046 6 : continue;
4047 : }
4048 :
4049 60 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4050 60 : if (dterr)
4051 0 : return dterr;
4052 60 : if (!AdjustMonths(val, itm_in) ||
4053 54 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
4054 6 : return DTERR_FIELD_OVERFLOW;
4055 54 : if (*str == '\0')
4056 6 : return 0;
4057 48 : if (*str == 'T')
4058 : {
4059 6 : datepart = false;
4060 6 : havefield = false;
4061 6 : continue;
4062 : }
4063 42 : if (*str != '-')
4064 0 : return DTERR_BAD_FORMAT;
4065 42 : str++;
4066 :
4067 42 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4068 42 : if (dterr)
4069 0 : return dterr;
4070 42 : if (!AdjustDays(val, 1, itm_in) ||
4071 36 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
4072 6 : return DTERR_FIELD_OVERFLOW;
4073 36 : if (*str == '\0')
4074 12 : return 0;
4075 24 : if (*str == 'T')
4076 : {
4077 24 : datepart = false;
4078 24 : havefield = false;
4079 24 : continue;
4080 : }
4081 0 : return DTERR_BAD_FORMAT;
4082 0 : default:
4083 : /* not a valid date unit suffix */
4084 0 : return DTERR_BAD_FORMAT;
4085 : }
4086 : }
4087 : else
4088 : {
4089 342 : switch (unit) /* after T: H M S */
4090 : {
4091 108 : case 'H':
4092 108 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
4093 36 : return DTERR_FIELD_OVERFLOW;
4094 72 : break;
4095 60 : case 'M':
4096 60 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
4097 0 : return DTERR_FIELD_OVERFLOW;
4098 60 : break;
4099 96 : case 'S':
4100 96 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
4101 12 : return DTERR_FIELD_OVERFLOW;
4102 84 : break;
4103 36 : case '\0': /* ISO 8601 4.4.3.3 Alternative Format */
4104 36 : if (ISO8601IntegerWidth(fieldstart) == 6 && !havefield)
4105 : {
4106 6 : if (!AdjustMicroseconds(val / 10000, 0, USECS_PER_HOUR, itm_in) ||
4107 6 : !AdjustMicroseconds((val / 100) % 100, 0, USECS_PER_MINUTE, itm_in) ||
4108 6 : !AdjustMicroseconds(val % 100, 0, USECS_PER_SEC, itm_in) ||
4109 6 : !AdjustFractMicroseconds(fval, 1, itm_in))
4110 0 : return DTERR_FIELD_OVERFLOW;
4111 6 : return 0;
4112 : }
4113 : /* Else fall through to extended alternative format */
4114 : /* FALLTHROUGH */
4115 : case ':': /* ISO 8601 4.4.3.3 Alternative Format,
4116 : * Extended */
4117 72 : if (havefield)
4118 0 : return DTERR_BAD_FORMAT;
4119 :
4120 72 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
4121 30 : return DTERR_FIELD_OVERFLOW;
4122 42 : if (unit == '\0')
4123 18 : return 0;
4124 :
4125 24 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4126 24 : if (dterr)
4127 0 : return dterr;
4128 24 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
4129 6 : return DTERR_FIELD_OVERFLOW;
4130 18 : if (*str == '\0')
4131 6 : return 0;
4132 12 : if (*str != ':')
4133 0 : return DTERR_BAD_FORMAT;
4134 12 : str++;
4135 :
4136 12 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4137 12 : if (dterr)
4138 0 : return dterr;
4139 12 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
4140 0 : return DTERR_FIELD_OVERFLOW;
4141 12 : if (*str == '\0')
4142 12 : return 0;
4143 0 : return DTERR_BAD_FORMAT;
4144 :
4145 0 : default:
4146 : /* not a valid time unit suffix */
4147 0 : return DTERR_BAD_FORMAT;
4148 : }
4149 : }
4150 :
4151 498 : havefield = true;
4152 : }
4153 :
4154 108 : return 0;
4155 : }
4156 :
4157 :
4158 : /* DecodeUnits()
4159 : * Decode text string using lookup table.
4160 : *
4161 : * This routine recognizes keywords associated with time interval units.
4162 : *
4163 : * Given string must be lowercased already.
4164 : *
4165 : * Implement a cache lookup since it is likely that dates
4166 : * will be related in format.
4167 : */
4168 : int
4169 121704 : DecodeUnits(int field, const char *lowtoken, int *val)
4170 : {
4171 : int type;
4172 : const datetkn *tp;
4173 :
4174 121704 : tp = deltacache[field];
4175 : /* use strncmp so that we match truncated tokens */
4176 121704 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
4177 : {
4178 54912 : tp = datebsearch(lowtoken, deltatktbl, szdeltatktbl);
4179 : }
4180 121704 : if (tp == NULL)
4181 : {
4182 35088 : type = UNKNOWN_FIELD;
4183 35088 : *val = 0;
4184 : }
4185 : else
4186 : {
4187 86616 : deltacache[field] = tp;
4188 86616 : type = tp->type;
4189 86616 : *val = tp->value;
4190 : }
4191 :
4192 121704 : return type;
4193 : } /* DecodeUnits() */
4194 :
4195 : /*
4196 : * Report an error detected by one of the datetime input processing routines.
4197 : *
4198 : * dterr is the error code, and *extra contains any auxiliary info we need
4199 : * for the error report. extra can be NULL if not needed for the particular
4200 : * dterr value.
4201 : *
4202 : * str is the original input string, and datatype is the name of the datatype
4203 : * we were trying to accept. (For some DTERR codes, these are not used and
4204 : * can be NULL.)
4205 : *
4206 : * If escontext points to an ErrorSaveContext node, that is filled instead
4207 : * of throwing an error.
4208 : *
4209 : * Note: it might seem useless to distinguish DTERR_INTERVAL_OVERFLOW and
4210 : * DTERR_TZDISP_OVERFLOW from DTERR_FIELD_OVERFLOW, but SQL99 mandates three
4211 : * separate SQLSTATE codes, so ...
4212 : */
4213 : void
4214 1752 : DateTimeParseError(int dterr, DateTimeErrorExtra *extra,
4215 : const char *str, const char *datatype,
4216 : Node *escontext)
4217 : {
4218 1752 : switch (dterr)
4219 : {
4220 216 : case DTERR_FIELD_OVERFLOW:
4221 216 : errsave(escontext,
4222 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4223 : errmsg("date/time field value out of range: \"%s\"",
4224 : str)));
4225 24 : break;
4226 180 : case DTERR_MD_FIELD_OVERFLOW:
4227 : /* <nanny>same as above, but add hint about DateStyle</nanny> */
4228 180 : errsave(escontext,
4229 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4230 : errmsg("date/time field value out of range: \"%s\"",
4231 : str),
4232 : errhint("Perhaps you need a different \"DateStyle\" setting.")));
4233 0 : break;
4234 720 : case DTERR_INTERVAL_OVERFLOW:
4235 720 : errsave(escontext,
4236 : (errcode(ERRCODE_INTERVAL_FIELD_OVERFLOW),
4237 : errmsg("interval field value out of range: \"%s\"",
4238 : str)));
4239 0 : break;
4240 12 : case DTERR_TZDISP_OVERFLOW:
4241 12 : errsave(escontext,
4242 : (errcode(ERRCODE_INVALID_TIME_ZONE_DISPLACEMENT_VALUE),
4243 : errmsg("time zone displacement out of range: \"%s\"",
4244 : str)));
4245 0 : break;
4246 36 : case DTERR_BAD_TIMEZONE:
4247 36 : errsave(escontext,
4248 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
4249 : errmsg("time zone \"%s\" not recognized",
4250 : extra->dtee_timezone)));
4251 24 : break;
4252 0 : case DTERR_BAD_ZONE_ABBREV:
4253 0 : errsave(escontext,
4254 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
4255 : errmsg("time zone \"%s\" not recognized",
4256 : extra->dtee_timezone),
4257 : errdetail("This time zone name appears in the configuration file for time zone abbreviation \"%s\".",
4258 : extra->dtee_abbrev)));
4259 0 : break;
4260 588 : case DTERR_BAD_FORMAT:
4261 : default:
4262 588 : errsave(escontext,
4263 : (errcode(ERRCODE_INVALID_DATETIME_FORMAT),
4264 : errmsg("invalid input syntax for type %s: \"%s\"",
4265 : datatype, str)));
4266 84 : break;
4267 : }
4268 132 : }
4269 :
4270 : /* datebsearch()
4271 : * Binary search -- from Knuth (6.2.1) Algorithm B. Special case like this
4272 : * is WAY faster than the generic bsearch().
4273 : */
4274 : static const datetkn *
4275 74498 : datebsearch(const char *key, const datetkn *base, int nel)
4276 : {
4277 74498 : if (nel > 0)
4278 : {
4279 74498 : const datetkn *last = base + nel - 1,
4280 : *position;
4281 : int result;
4282 :
4283 471456 : while (last >= base)
4284 : {
4285 427188 : position = base + ((last - base) >> 1);
4286 : /* precheck the first character for a bit of extra speed */
4287 427188 : result = (int) key[0] - (int) position->token[0];
4288 427188 : if (result == 0)
4289 : {
4290 : /* use strncmp so that we match truncated tokens */
4291 106848 : result = strncmp(key, position->token, TOKMAXLEN);
4292 106848 : if (result == 0)
4293 30230 : return position;
4294 : }
4295 396958 : if (result < 0)
4296 197082 : last = position - 1;
4297 : else
4298 199876 : base = position + 1;
4299 : }
4300 : }
4301 44268 : return NULL;
4302 : }
4303 :
4304 : /* EncodeTimezone()
4305 : * Copies representation of a numeric timezone offset to str.
4306 : *
4307 : * Returns a pointer to the new end of string. No NUL terminator is put
4308 : * there; callers are responsible for NUL terminating str themselves.
4309 : */
4310 : static char *
4311 67180 : EncodeTimezone(char *str, int tz, int style)
4312 : {
4313 : int hour,
4314 : min,
4315 : sec;
4316 :
4317 67180 : sec = abs(tz);
4318 67180 : min = sec / SECS_PER_MINUTE;
4319 67180 : sec -= min * SECS_PER_MINUTE;
4320 67180 : hour = min / MINS_PER_HOUR;
4321 67180 : min -= hour * MINS_PER_HOUR;
4322 :
4323 : /* TZ is negated compared to sign we wish to display ... */
4324 67180 : *str++ = (tz <= 0 ? '+' : '-');
4325 :
4326 67180 : if (sec != 0)
4327 : {
4328 0 : str = pg_ultostr_zeropad(str, hour, 2);
4329 0 : *str++ = ':';
4330 0 : str = pg_ultostr_zeropad(str, min, 2);
4331 0 : *str++ = ':';
4332 0 : str = pg_ultostr_zeropad(str, sec, 2);
4333 : }
4334 67180 : else if (min != 0 || style == USE_XSD_DATES)
4335 : {
4336 852 : str = pg_ultostr_zeropad(str, hour, 2);
4337 852 : *str++ = ':';
4338 852 : str = pg_ultostr_zeropad(str, min, 2);
4339 : }
4340 : else
4341 66328 : str = pg_ultostr_zeropad(str, hour, 2);
4342 67180 : return str;
4343 : }
4344 :
4345 : /* EncodeDateOnly()
4346 : * Encode date as local time.
4347 : */
4348 : void
4349 16738 : EncodeDateOnly(struct pg_tm *tm, int style, char *str)
4350 : {
4351 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4352 :
4353 16738 : switch (style)
4354 : {
4355 11232 : case USE_ISO_DATES:
4356 : case USE_XSD_DATES:
4357 : /* compatible with ISO date formats */
4358 11232 : str = pg_ultostr_zeropad(str,
4359 11232 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4360 11232 : *str++ = '-';
4361 11232 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4362 11232 : *str++ = '-';
4363 11232 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4364 11232 : break;
4365 :
4366 0 : case USE_SQL_DATES:
4367 : /* compatible with Oracle/Ingres date formats */
4368 0 : if (DateOrder == DATEORDER_DMY)
4369 : {
4370 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4371 0 : *str++ = '/';
4372 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4373 : }
4374 : else
4375 : {
4376 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4377 0 : *str++ = '/';
4378 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4379 : }
4380 0 : *str++ = '/';
4381 0 : str = pg_ultostr_zeropad(str,
4382 0 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4383 0 : break;
4384 :
4385 8 : case USE_GERMAN_DATES:
4386 : /* German-style date format */
4387 8 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4388 8 : *str++ = '.';
4389 8 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4390 8 : *str++ = '.';
4391 8 : str = pg_ultostr_zeropad(str,
4392 8 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4393 8 : break;
4394 :
4395 5498 : case USE_POSTGRES_DATES:
4396 : default:
4397 : /* traditional date-only style for Postgres */
4398 5498 : if (DateOrder == DATEORDER_DMY)
4399 : {
4400 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4401 0 : *str++ = '-';
4402 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4403 : }
4404 : else
4405 : {
4406 5498 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4407 5498 : *str++ = '-';
4408 5498 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4409 : }
4410 5498 : *str++ = '-';
4411 5498 : str = pg_ultostr_zeropad(str,
4412 5498 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4413 5498 : break;
4414 : }
4415 :
4416 16738 : if (tm->tm_year <= 0)
4417 : {
4418 86 : memcpy(str, " BC", 3); /* Don't copy NUL */
4419 86 : str += 3;
4420 : }
4421 16738 : *str = '\0';
4422 16738 : }
4423 :
4424 :
4425 : /* EncodeTimeOnly()
4426 : * Encode time fields only.
4427 : *
4428 : * tm and fsec are the value to encode, print_tz determines whether to include
4429 : * a time zone (the difference between time and timetz types), tz is the
4430 : * numeric time zone offset, style is the date style, str is where to write the
4431 : * output.
4432 : */
4433 : void
4434 18492 : EncodeTimeOnly(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, int style, char *str)
4435 : {
4436 18492 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4437 18492 : *str++ = ':';
4438 18492 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4439 18492 : *str++ = ':';
4440 18492 : str = AppendSeconds(str, tm->tm_sec, fsec, MAX_TIME_PRECISION, true);
4441 18492 : if (print_tz)
4442 9488 : str = EncodeTimezone(str, tz, style);
4443 18492 : *str = '\0';
4444 18492 : }
4445 :
4446 :
4447 : /* EncodeDateTime()
4448 : * Encode date and time interpreted as local time.
4449 : *
4450 : * tm and fsec are the value to encode, print_tz determines whether to include
4451 : * a time zone (the difference between timestamp and timestamptz types), tz is
4452 : * the numeric time zone offset, tzn is the textual time zone, which if
4453 : * specified will be used instead of tz by some styles, style is the date
4454 : * style, str is where to write the output.
4455 : *
4456 : * Supported date styles:
4457 : * Postgres - day mon hh:mm:ss yyyy tz
4458 : * SQL - mm/dd/yyyy hh:mm:ss.ss tz
4459 : * ISO - yyyy-mm-dd hh:mm:ss+/-tz
4460 : * German - dd.mm.yyyy hh:mm:ss tz
4461 : * XSD - yyyy-mm-ddThh:mm:ss.ss+/-tz
4462 : */
4463 : void
4464 122452 : EncodeDateTime(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, const char *tzn, int style, char *str)
4465 : {
4466 : int day;
4467 :
4468 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4469 :
4470 : /*
4471 : * Negative tm_isdst means we have no valid time zone translation.
4472 : */
4473 122452 : if (tm->tm_isdst < 0)
4474 45982 : print_tz = false;
4475 :
4476 122452 : switch (style)
4477 : {
4478 92402 : case USE_ISO_DATES:
4479 : case USE_XSD_DATES:
4480 : /* Compatible with ISO-8601 date formats */
4481 92402 : str = pg_ultostr_zeropad(str,
4482 92402 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4483 92402 : *str++ = '-';
4484 92402 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4485 92402 : *str++ = '-';
4486 92402 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4487 92402 : *str++ = (style == USE_ISO_DATES) ? ' ' : 'T';
4488 92402 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4489 92402 : *str++ = ':';
4490 92402 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4491 92402 : *str++ = ':';
4492 92402 : str = AppendTimestampSeconds(str, tm, fsec);
4493 92402 : if (print_tz)
4494 57692 : str = EncodeTimezone(str, tz, style);
4495 92402 : break;
4496 :
4497 780 : case USE_SQL_DATES:
4498 : /* Compatible with Oracle/Ingres date formats */
4499 780 : if (DateOrder == DATEORDER_DMY)
4500 : {
4501 384 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4502 384 : *str++ = '/';
4503 384 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4504 : }
4505 : else
4506 : {
4507 396 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4508 396 : *str++ = '/';
4509 396 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4510 : }
4511 780 : *str++ = '/';
4512 780 : str = pg_ultostr_zeropad(str,
4513 780 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4514 780 : *str++ = ' ';
4515 780 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4516 780 : *str++ = ':';
4517 780 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4518 780 : *str++ = ':';
4519 780 : str = AppendTimestampSeconds(str, tm, fsec);
4520 :
4521 : /*
4522 : * Note: the uses of %.*s in this function would be risky if the
4523 : * timezone names ever contain non-ASCII characters, since we are
4524 : * not being careful to do encoding-aware clipping. However, all
4525 : * TZ abbreviations in the IANA database are plain ASCII.
4526 : */
4527 780 : if (print_tz)
4528 : {
4529 18 : if (tzn)
4530 : {
4531 18 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4532 18 : str += strlen(str);
4533 : }
4534 : else
4535 0 : str = EncodeTimezone(str, tz, style);
4536 : }
4537 780 : break;
4538 :
4539 24 : case USE_GERMAN_DATES:
4540 : /* German variant on European style */
4541 24 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4542 24 : *str++ = '.';
4543 24 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4544 24 : *str++ = '.';
4545 24 : str = pg_ultostr_zeropad(str,
4546 24 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4547 24 : *str++ = ' ';
4548 24 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4549 24 : *str++ = ':';
4550 24 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4551 24 : *str++ = ':';
4552 24 : str = AppendTimestampSeconds(str, tm, fsec);
4553 :
4554 24 : if (print_tz)
4555 : {
4556 24 : if (tzn)
4557 : {
4558 24 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4559 24 : str += strlen(str);
4560 : }
4561 : else
4562 0 : str = EncodeTimezone(str, tz, style);
4563 : }
4564 24 : break;
4565 :
4566 29246 : case USE_POSTGRES_DATES:
4567 : default:
4568 : /* Backward-compatible with traditional Postgres abstime dates */
4569 29246 : day = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday);
4570 29246 : tm->tm_wday = j2day(day);
4571 29246 : memcpy(str, days[tm->tm_wday], 3);
4572 29246 : str += 3;
4573 29246 : *str++ = ' ';
4574 29246 : if (DateOrder == DATEORDER_DMY)
4575 : {
4576 398 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4577 398 : *str++ = ' ';
4578 398 : memcpy(str, months[tm->tm_mon - 1], 3);
4579 398 : str += 3;
4580 : }
4581 : else
4582 : {
4583 28848 : memcpy(str, months[tm->tm_mon - 1], 3);
4584 28848 : str += 3;
4585 28848 : *str++ = ' ';
4586 28848 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4587 : }
4588 29246 : *str++ = ' ';
4589 29246 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4590 29246 : *str++ = ':';
4591 29246 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4592 29246 : *str++ = ':';
4593 29246 : str = AppendTimestampSeconds(str, tm, fsec);
4594 29246 : *str++ = ' ';
4595 29246 : str = pg_ultostr_zeropad(str,
4596 29246 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4597 :
4598 29246 : if (print_tz)
4599 : {
4600 18736 : if (tzn)
4601 : {
4602 18736 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4603 18736 : str += strlen(str);
4604 : }
4605 : else
4606 : {
4607 : /*
4608 : * We have a time zone, but no string version. Use the
4609 : * numeric form, but be sure to include a leading space to
4610 : * avoid formatting something which would be rejected by
4611 : * the date/time parser later. - thomas 2001-10-19
4612 : */
4613 0 : *str++ = ' ';
4614 0 : str = EncodeTimezone(str, tz, style);
4615 : }
4616 : }
4617 29246 : break;
4618 : }
4619 :
4620 122452 : if (tm->tm_year <= 0)
4621 : {
4622 298 : memcpy(str, " BC", 3); /* Don't copy NUL */
4623 298 : str += 3;
4624 : }
4625 122452 : *str = '\0';
4626 122452 : }
4627 :
4628 :
4629 : /*
4630 : * Helper functions to avoid duplicated code in EncodeInterval.
4631 : */
4632 :
4633 : /* Append an ISO-8601-style interval field, but only if value isn't zero */
4634 : static char *
4635 210 : AddISO8601IntPart(char *cp, int64 value, char units)
4636 : {
4637 210 : if (value == 0)
4638 54 : return cp;
4639 156 : sprintf(cp, "%" PRId64 "%c", value, units);
4640 156 : return cp + strlen(cp);
4641 : }
4642 :
4643 : /* Append a postgres-style interval field, but only if value isn't zero */
4644 : static char *
4645 23922 : AddPostgresIntPart(char *cp, int64 value, const char *units,
4646 : bool *is_zero, bool *is_before)
4647 : {
4648 23922 : if (value == 0)
4649 12348 : return cp;
4650 34722 : sprintf(cp, "%s%s%" PRId64 " %s%s",
4651 11574 : (!*is_zero) ? " " : "",
4652 11574 : (*is_before && value > 0) ? "+" : "",
4653 : value,
4654 : units,
4655 : (value != 1) ? "s" : "");
4656 :
4657 : /*
4658 : * Each nonzero field sets is_before for (only) the next one. This is a
4659 : * tad bizarre but it's how it worked before...
4660 : */
4661 11574 : *is_before = (value < 0);
4662 11574 : *is_zero = false;
4663 11574 : return cp + strlen(cp);
4664 : }
4665 :
4666 : /* Append a verbose-style interval field, but only if value isn't zero */
4667 : static char *
4668 42520 : AddVerboseIntPart(char *cp, int64 value, const char *units,
4669 : bool *is_zero, bool *is_before)
4670 : {
4671 42520 : if (value == 0)
4672 28226 : return cp;
4673 : /* first nonzero value sets is_before */
4674 14294 : if (*is_zero)
4675 : {
4676 7880 : *is_before = (value < 0);
4677 7880 : value = i64abs(value);
4678 : }
4679 6414 : else if (*is_before)
4680 1356 : value = -value;
4681 14294 : sprintf(cp, " %" PRId64 " %s%s", value, units, (value == 1) ? "" : "s");
4682 14294 : *is_zero = false;
4683 14294 : return cp + strlen(cp);
4684 : }
4685 :
4686 :
4687 : /* EncodeInterval()
4688 : * Interpret time structure as a delta time and convert to string.
4689 : *
4690 : * Support "traditional Postgres" and ISO-8601 styles.
4691 : * Actually, afaik ISO does not address time interval formatting,
4692 : * but this looks similar to the spec for absolute date/time.
4693 : * - thomas 1998-04-30
4694 : *
4695 : * Actually, afaik, ISO 8601 does specify formats for "time
4696 : * intervals...[of the]...format with time-unit designators", which
4697 : * are pretty ugly. The format looks something like
4698 : * P1Y1M1DT1H1M1.12345S
4699 : * but useful for exchanging data with computers instead of humans.
4700 : * - ron 2003-07-14
4701 : *
4702 : * And ISO's SQL 2008 standard specifies standards for
4703 : * "year-month literal"s (that look like '2-3') and
4704 : * "day-time literal"s (that look like ('4 5:6:7')
4705 : */
4706 : void
4707 16652 : EncodeInterval(struct pg_itm *itm, int style, char *str)
4708 : {
4709 16652 : char *cp = str;
4710 16652 : int year = itm->tm_year;
4711 16652 : int mon = itm->tm_mon;
4712 16652 : int64 mday = itm->tm_mday; /* tm_mday could be INT_MIN */
4713 16652 : int64 hour = itm->tm_hour;
4714 16652 : int min = itm->tm_min;
4715 16652 : int sec = itm->tm_sec;
4716 16652 : int fsec = itm->tm_usec;
4717 16652 : bool is_before = false;
4718 16652 : bool is_zero = true;
4719 :
4720 : /*
4721 : * The sign of year and month are guaranteed to match, since they are
4722 : * stored internally as "month". But we'll need to check for is_before and
4723 : * is_zero when determining the signs of day and hour/minute/seconds
4724 : * fields.
4725 : */
4726 16652 : switch (style)
4727 : {
4728 : /* SQL Standard interval format */
4729 126 : case INTSTYLE_SQL_STANDARD:
4730 : {
4731 96 : bool has_negative = year < 0 || mon < 0 ||
4732 66 : mday < 0 || hour < 0 ||
4733 222 : min < 0 || sec < 0 || fsec < 0;
4734 102 : bool has_positive = year > 0 || mon > 0 ||
4735 66 : mday > 0 || hour > 0 ||
4736 228 : min > 0 || sec > 0 || fsec > 0;
4737 126 : bool has_year_month = year != 0 || mon != 0;
4738 42 : bool has_day_time = mday != 0 || hour != 0 ||
4739 168 : min != 0 || sec != 0 || fsec != 0;
4740 126 : bool has_day = mday != 0;
4741 222 : bool sql_standard_value = !(has_negative && has_positive) &&
4742 96 : !(has_year_month && has_day_time);
4743 :
4744 : /*
4745 : * SQL Standard wants only 1 "<sign>" preceding the whole
4746 : * interval ... but can't do that if mixed signs.
4747 : */
4748 126 : if (has_negative && sql_standard_value)
4749 : {
4750 30 : *cp++ = '-';
4751 30 : year = -year;
4752 30 : mon = -mon;
4753 30 : mday = -mday;
4754 30 : hour = -hour;
4755 30 : min = -min;
4756 30 : sec = -sec;
4757 30 : fsec = -fsec;
4758 : }
4759 :
4760 126 : if (!has_negative && !has_positive)
4761 : {
4762 12 : sprintf(cp, "0");
4763 : }
4764 114 : else if (!sql_standard_value)
4765 : {
4766 : /*
4767 : * For non sql-standard interval values, force outputting
4768 : * the signs to avoid ambiguities with intervals with
4769 : * mixed sign components.
4770 : */
4771 54 : char year_sign = (year < 0 || mon < 0) ? '-' : '+';
4772 54 : char day_sign = (mday < 0) ? '-' : '+';
4773 78 : char sec_sign = (hour < 0 || min < 0 ||
4774 24 : sec < 0 || fsec < 0) ? '-' : '+';
4775 :
4776 54 : sprintf(cp, "%c%d-%d %c%" PRId64 " %c%" PRId64 ":%02d:",
4777 : year_sign, abs(year), abs(mon),
4778 : day_sign, i64abs(mday),
4779 : sec_sign, i64abs(hour), abs(min));
4780 54 : cp += strlen(cp);
4781 54 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4782 54 : *cp = '\0';
4783 : }
4784 60 : else if (has_year_month)
4785 : {
4786 18 : sprintf(cp, "%d-%d", year, mon);
4787 : }
4788 42 : else if (has_day)
4789 : {
4790 30 : sprintf(cp, "%" PRId64 " %" PRId64 ":%02d:",
4791 : mday, hour, min);
4792 30 : cp += strlen(cp);
4793 30 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4794 30 : *cp = '\0';
4795 : }
4796 : else
4797 : {
4798 12 : sprintf(cp, "%" PRId64 ":%02d:", hour, min);
4799 12 : cp += strlen(cp);
4800 12 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4801 12 : *cp = '\0';
4802 : }
4803 : }
4804 126 : break;
4805 :
4806 : /* ISO 8601 "time-intervals by duration only" */
4807 48 : case INTSTYLE_ISO_8601:
4808 : /* special-case zero to avoid printing nothing */
4809 48 : if (year == 0 && mon == 0 && mday == 0 &&
4810 6 : hour == 0 && min == 0 && sec == 0 && fsec == 0)
4811 : {
4812 6 : sprintf(cp, "PT0S");
4813 6 : break;
4814 : }
4815 42 : *cp++ = 'P';
4816 42 : cp = AddISO8601IntPart(cp, year, 'Y');
4817 42 : cp = AddISO8601IntPart(cp, mon, 'M');
4818 42 : cp = AddISO8601IntPart(cp, mday, 'D');
4819 42 : if (hour != 0 || min != 0 || sec != 0 || fsec != 0)
4820 36 : *cp++ = 'T';
4821 42 : cp = AddISO8601IntPart(cp, hour, 'H');
4822 42 : cp = AddISO8601IntPart(cp, min, 'M');
4823 42 : if (sec != 0 || fsec != 0)
4824 : {
4825 36 : if (sec < 0 || fsec < 0)
4826 12 : *cp++ = '-';
4827 36 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4828 36 : *cp++ = 'S';
4829 36 : *cp++ = '\0';
4830 : }
4831 42 : break;
4832 :
4833 : /* Compatible with postgresql < 8.4 when DateStyle = 'iso' */
4834 7974 : case INTSTYLE_POSTGRES:
4835 7974 : cp = AddPostgresIntPart(cp, year, "year", &is_zero, &is_before);
4836 :
4837 : /*
4838 : * Ideally we should spell out "month" like we do for "year" and
4839 : * "day". However, for backward compatibility, we can't easily
4840 : * fix this. bjm 2011-05-24
4841 : */
4842 7974 : cp = AddPostgresIntPart(cp, mon, "mon", &is_zero, &is_before);
4843 7974 : cp = AddPostgresIntPart(cp, mday, "day", &is_zero, &is_before);
4844 7974 : if (is_zero || hour != 0 || min != 0 || sec != 0 || fsec != 0)
4845 : {
4846 7000 : bool minus = (hour < 0 || min < 0 || sec < 0 || fsec < 0);
4847 :
4848 20780 : sprintf(cp, "%s%s%02" PRId64 ":%02d:",
4849 7000 : is_zero ? "" : " ",
4850 6780 : (minus ? "-" : (is_before ? "+" : "")),
4851 : i64abs(hour), abs(min));
4852 7000 : cp += strlen(cp);
4853 7000 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4854 7000 : *cp = '\0';
4855 : }
4856 7974 : break;
4857 :
4858 : /* Compatible with postgresql < 8.4 when DateStyle != 'iso' */
4859 8504 : case INTSTYLE_POSTGRES_VERBOSE:
4860 : default:
4861 8504 : strcpy(cp, "@");
4862 8504 : cp++;
4863 8504 : cp = AddVerboseIntPart(cp, year, "year", &is_zero, &is_before);
4864 8504 : cp = AddVerboseIntPart(cp, mon, "mon", &is_zero, &is_before);
4865 8504 : cp = AddVerboseIntPart(cp, mday, "day", &is_zero, &is_before);
4866 8504 : cp = AddVerboseIntPart(cp, hour, "hour", &is_zero, &is_before);
4867 8504 : cp = AddVerboseIntPart(cp, min, "min", &is_zero, &is_before);
4868 8504 : if (sec != 0 || fsec != 0)
4869 : {
4870 3282 : *cp++ = ' ';
4871 3282 : if (sec < 0 || (sec == 0 && fsec < 0))
4872 : {
4873 1020 : if (is_zero)
4874 342 : is_before = true;
4875 678 : else if (!is_before)
4876 6 : *cp++ = '-';
4877 : }
4878 2262 : else if (is_before)
4879 12 : *cp++ = '-';
4880 3282 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4881 : /* We output "ago", not negatives, so use abs(). */
4882 3282 : sprintf(cp, " sec%s",
4883 3282 : (abs(sec) != 1 || fsec != 0) ? "s" : "");
4884 3282 : is_zero = false;
4885 : }
4886 : /* identically zero? then put in a unitless zero... */
4887 8504 : if (is_zero)
4888 230 : strcat(cp, " 0");
4889 8504 : if (is_before)
4890 1414 : strcat(cp, " ago");
4891 8504 : break;
4892 : }
4893 16652 : }
4894 :
4895 :
4896 : /*
4897 : * We've been burnt by stupid errors in the ordering of the datetkn tables
4898 : * once too often. Arrange to check them during postmaster start.
4899 : */
4900 : static bool
4901 3384 : CheckDateTokenTable(const char *tablename, const datetkn *base, int nel)
4902 : {
4903 3384 : bool ok = true;
4904 : int i;
4905 :
4906 228420 : for (i = 0; i < nel; i++)
4907 : {
4908 : /* check for token strings that don't fit */
4909 225036 : if (strlen(base[i].token) > TOKMAXLEN)
4910 : {
4911 : /* %.*s is safe since all our tokens are ASCII */
4912 0 : elog(LOG, "token too long in %s table: \"%.*s\"",
4913 : tablename,
4914 : TOKMAXLEN + 1, base[i].token);
4915 0 : ok = false;
4916 0 : break; /* don't risk applying strcmp */
4917 : }
4918 : /* check for out of order */
4919 225036 : if (i > 0 &&
4920 221652 : strcmp(base[i - 1].token, base[i].token) >= 0)
4921 : {
4922 0 : elog(LOG, "ordering error in %s table: \"%s\" >= \"%s\"",
4923 : tablename,
4924 : base[i - 1].token,
4925 : base[i].token);
4926 0 : ok = false;
4927 : }
4928 : }
4929 3384 : return ok;
4930 : }
4931 :
4932 : bool
4933 1692 : CheckDateTokenTables(void)
4934 : {
4935 1692 : bool ok = true;
4936 :
4937 : Assert(UNIX_EPOCH_JDATE == date2j(1970, 1, 1));
4938 : Assert(POSTGRES_EPOCH_JDATE == date2j(2000, 1, 1));
4939 :
4940 1692 : ok &= CheckDateTokenTable("datetktbl", datetktbl, szdatetktbl);
4941 1692 : ok &= CheckDateTokenTable("deltatktbl", deltatktbl, szdeltatktbl);
4942 1692 : return ok;
4943 : }
4944 :
4945 : /*
4946 : * Common code for temporal prosupport functions: simplify, if possible,
4947 : * a call to a temporal type's length-coercion function.
4948 : *
4949 : * Types time, timetz, timestamp and timestamptz each have a range of allowed
4950 : * precisions. An unspecified precision is rigorously equivalent to the
4951 : * highest specifiable precision. We can replace the function call with a
4952 : * no-op RelabelType if it is coercing to the same or higher precision as the
4953 : * input is known to have.
4954 : *
4955 : * The input Node is always a FuncExpr, but to reduce the #include footprint
4956 : * of datetime.h, we declare it as Node *.
4957 : *
4958 : * Note: timestamp_scale throws an error when the typmod is out of range, but
4959 : * we can't get there from a cast: our typmodin will have caught it already.
4960 : */
4961 : Node *
4962 24 : TemporalSimplify(int32 max_precis, Node *node)
4963 : {
4964 24 : FuncExpr *expr = castNode(FuncExpr, node);
4965 24 : Node *ret = NULL;
4966 : Node *typmod;
4967 :
4968 : Assert(list_length(expr->args) >= 2);
4969 :
4970 24 : typmod = (Node *) lsecond(expr->args);
4971 :
4972 24 : if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
4973 : {
4974 24 : Node *source = (Node *) linitial(expr->args);
4975 24 : int32 old_precis = exprTypmod(source);
4976 24 : int32 new_precis = DatumGetInt32(((Const *) typmod)->constvalue);
4977 :
4978 24 : if (new_precis < 0 || new_precis == max_precis ||
4979 0 : (old_precis >= 0 && new_precis >= old_precis))
4980 0 : ret = relabel_to_typmod(source, new_precis);
4981 : }
4982 :
4983 24 : return ret;
4984 : }
4985 :
4986 : /*
4987 : * This function gets called during timezone config file load or reload
4988 : * to create the final array of timezone tokens. The argument array
4989 : * is already sorted in name order.
4990 : *
4991 : * The result is a TimeZoneAbbrevTable (which must be a single guc_malloc'd
4992 : * chunk) or NULL on alloc failure. No other error conditions are defined.
4993 : */
4994 : TimeZoneAbbrevTable *
4995 12762 : ConvertTimeZoneAbbrevs(struct tzEntry *abbrevs, int n)
4996 : {
4997 : TimeZoneAbbrevTable *tbl;
4998 : Size tbl_size;
4999 : int i;
5000 :
5001 : /* Space for fixed fields and datetkn array */
5002 12762 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
5003 12762 : n * sizeof(datetkn);
5004 12762 : tbl_size = MAXALIGN(tbl_size);
5005 : /* Count up space for dynamic abbreviations */
5006 2501364 : for (i = 0; i < n; i++)
5007 : {
5008 2488602 : struct tzEntry *abbr = abbrevs + i;
5009 :
5010 2488602 : if (abbr->zone != NULL)
5011 : {
5012 : Size dsize;
5013 :
5014 638094 : dsize = offsetof(DynamicZoneAbbrev, zone) +
5015 638094 : strlen(abbr->zone) + 1;
5016 638094 : tbl_size += MAXALIGN(dsize);
5017 : }
5018 : }
5019 :
5020 : /* Alloc the result ... */
5021 12762 : tbl = guc_malloc(LOG, tbl_size);
5022 12762 : if (!tbl)
5023 0 : return NULL;
5024 :
5025 : /* ... and fill it in */
5026 12762 : tbl->tblsize = tbl_size;
5027 12762 : tbl->numabbrevs = n;
5028 : /* in this loop, tbl_size reprises the space calculation above */
5029 12762 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
5030 12762 : n * sizeof(datetkn);
5031 12762 : tbl_size = MAXALIGN(tbl_size);
5032 2501364 : for (i = 0; i < n; i++)
5033 : {
5034 2488602 : struct tzEntry *abbr = abbrevs + i;
5035 2488602 : datetkn *dtoken = tbl->abbrevs + i;
5036 :
5037 : /* use strlcpy to truncate name if necessary */
5038 2488602 : strlcpy(dtoken->token, abbr->abbrev, TOKMAXLEN + 1);
5039 2488602 : if (abbr->zone != NULL)
5040 : {
5041 : /* Allocate a DynamicZoneAbbrev for this abbreviation */
5042 : DynamicZoneAbbrev *dtza;
5043 : Size dsize;
5044 :
5045 638094 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tbl_size);
5046 638094 : dtza->tz = NULL;
5047 638094 : strcpy(dtza->zone, abbr->zone);
5048 :
5049 638094 : dtoken->type = DYNTZ;
5050 : /* value is offset from table start to DynamicZoneAbbrev */
5051 638094 : dtoken->value = (int32) tbl_size;
5052 :
5053 638094 : dsize = offsetof(DynamicZoneAbbrev, zone) +
5054 638094 : strlen(abbr->zone) + 1;
5055 638094 : tbl_size += MAXALIGN(dsize);
5056 : }
5057 : else
5058 : {
5059 1850508 : dtoken->type = abbr->is_dst ? DTZ : TZ;
5060 1850508 : dtoken->value = abbr->offset;
5061 : }
5062 : }
5063 :
5064 : /* Assert the two loops above agreed on size calculations */
5065 : Assert(tbl->tblsize == tbl_size);
5066 :
5067 : /* Check the ordering, if testing */
5068 : Assert(CheckDateTokenTable("timezone abbreviations", tbl->abbrevs, n));
5069 :
5070 12762 : return tbl;
5071 : }
5072 :
5073 : /*
5074 : * Install a TimeZoneAbbrevTable as the active table.
5075 : *
5076 : * Caller is responsible that the passed table doesn't go away while in use.
5077 : */
5078 : void
5079 12564 : InstallTimeZoneAbbrevs(TimeZoneAbbrevTable *tbl)
5080 : {
5081 12564 : zoneabbrevtbl = tbl;
5082 : /* reset tzabbrevcache, which may contain results from old table */
5083 12564 : memset(tzabbrevcache, 0, sizeof(tzabbrevcache));
5084 12564 : }
5085 :
5086 : /*
5087 : * Helper subroutine to locate pg_tz timezone for a dynamic abbreviation.
5088 : *
5089 : * On failure, returns NULL and fills *extra for a DTERR_BAD_ZONE_ABBREV error.
5090 : */
5091 : static pg_tz *
5092 1230 : FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp,
5093 : DateTimeErrorExtra *extra)
5094 : {
5095 : DynamicZoneAbbrev *dtza;
5096 :
5097 : /* Just some sanity checks to prevent indexing off into nowhere */
5098 : Assert(tp->type == DYNTZ);
5099 : Assert(tp->value > 0 && tp->value < tbl->tblsize);
5100 :
5101 1230 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tp->value);
5102 :
5103 : /* Look up the underlying zone if we haven't already */
5104 1230 : if (dtza->tz == NULL)
5105 : {
5106 918 : dtza->tz = pg_tzset(dtza->zone);
5107 918 : if (dtza->tz == NULL)
5108 : {
5109 : /* Ooops, bogus zone name in config file entry */
5110 0 : extra->dtee_timezone = dtza->zone;
5111 0 : extra->dtee_abbrev = tp->token;
5112 : }
5113 : }
5114 1230 : return dtza->tz;
5115 : }
5116 :
5117 :
5118 : /*
5119 : * This set-returning function reads all the time zone abbreviations
5120 : * defined by the IANA data for the current timezone setting,
5121 : * and returns a set of (abbrev, utc_offset, is_dst).
5122 : */
5123 : Datum
5124 252 : pg_timezone_abbrevs_zone(PG_FUNCTION_ARGS)
5125 : {
5126 : FuncCallContext *funcctx;
5127 : int *pindex;
5128 : Datum result;
5129 : HeapTuple tuple;
5130 : Datum values[3];
5131 252 : bool nulls[3] = {0};
5132 252 : TimestampTz now = GetCurrentTransactionStartTimestamp();
5133 252 : pg_time_t t = timestamptz_to_time_t(now);
5134 : const char *abbrev;
5135 : long int gmtoff;
5136 : int isdst;
5137 : struct pg_itm_in itm_in;
5138 : Interval *resInterval;
5139 :
5140 : /* stuff done only on the first call of the function */
5141 252 : if (SRF_IS_FIRSTCALL())
5142 : {
5143 : TupleDesc tupdesc;
5144 : MemoryContext oldcontext;
5145 :
5146 : /* create a function context for cross-call persistence */
5147 42 : funcctx = SRF_FIRSTCALL_INIT();
5148 :
5149 : /*
5150 : * switch to memory context appropriate for multiple function calls
5151 : */
5152 42 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
5153 :
5154 : /* allocate memory for user context */
5155 42 : pindex = (int *) palloc(sizeof(int));
5156 42 : *pindex = 0;
5157 42 : funcctx->user_fctx = pindex;
5158 :
5159 42 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
5160 0 : elog(ERROR, "return type must be a row type");
5161 42 : funcctx->tuple_desc = tupdesc;
5162 :
5163 42 : MemoryContextSwitchTo(oldcontext);
5164 : }
5165 :
5166 : /* stuff done on every call of the function */
5167 252 : funcctx = SRF_PERCALL_SETUP();
5168 252 : pindex = (int *) funcctx->user_fctx;
5169 :
5170 252 : while ((abbrev = pg_get_next_timezone_abbrev(pindex,
5171 252 : session_timezone)) != NULL)
5172 : {
5173 : /* Ignore abbreviations that aren't all-alphabetic */
5174 210 : if (strspn(abbrev, "ABCDEFGHIJKLMNOPQRSTUVWXYZ") != strlen(abbrev))
5175 0 : continue;
5176 :
5177 : /* Determine the current meaning of the abbrev */
5178 210 : if (!pg_interpret_timezone_abbrev(abbrev,
5179 : &t,
5180 : &gmtoff,
5181 : &isdst,
5182 : session_timezone))
5183 0 : continue; /* hm, not actually used in this zone? */
5184 :
5185 210 : values[0] = CStringGetTextDatum(abbrev);
5186 :
5187 : /* Convert offset (in seconds) to an interval; can't overflow */
5188 840 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5189 210 : itm_in.tm_usec = (int64) gmtoff * USECS_PER_SEC;
5190 210 : resInterval = (Interval *) palloc(sizeof(Interval));
5191 210 : (void) itmin2interval(&itm_in, resInterval);
5192 210 : values[1] = IntervalPGetDatum(resInterval);
5193 :
5194 210 : values[2] = BoolGetDatum(isdst);
5195 :
5196 210 : tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
5197 210 : result = HeapTupleGetDatum(tuple);
5198 :
5199 210 : SRF_RETURN_NEXT(funcctx, result);
5200 : }
5201 :
5202 42 : SRF_RETURN_DONE(funcctx);
5203 : }
5204 :
5205 : /*
5206 : * This set-returning function reads all the time zone abbreviations
5207 : * defined by the timezone_abbreviations setting,
5208 : * and returns a set of (abbrev, utc_offset, is_dst).
5209 : */
5210 : Datum
5211 4716 : pg_timezone_abbrevs_abbrevs(PG_FUNCTION_ARGS)
5212 : {
5213 : FuncCallContext *funcctx;
5214 : int *pindex;
5215 : Datum result;
5216 : HeapTuple tuple;
5217 : Datum values[3];
5218 4716 : bool nulls[3] = {0};
5219 : const datetkn *tp;
5220 : char buffer[TOKMAXLEN + 1];
5221 : int gmtoffset;
5222 : bool is_dst;
5223 : unsigned char *p;
5224 : struct pg_itm_in itm_in;
5225 : Interval *resInterval;
5226 :
5227 : /* stuff done only on the first call of the function */
5228 4716 : if (SRF_IS_FIRSTCALL())
5229 : {
5230 : TupleDesc tupdesc;
5231 : MemoryContext oldcontext;
5232 :
5233 : /* create a function context for cross-call persistence */
5234 24 : funcctx = SRF_FIRSTCALL_INIT();
5235 :
5236 : /*
5237 : * switch to memory context appropriate for multiple function calls
5238 : */
5239 24 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
5240 :
5241 : /* allocate memory for user context */
5242 24 : pindex = (int *) palloc(sizeof(int));
5243 24 : *pindex = 0;
5244 24 : funcctx->user_fctx = pindex;
5245 :
5246 24 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
5247 0 : elog(ERROR, "return type must be a row type");
5248 24 : funcctx->tuple_desc = tupdesc;
5249 :
5250 24 : MemoryContextSwitchTo(oldcontext);
5251 : }
5252 :
5253 : /* stuff done on every call of the function */
5254 4716 : funcctx = SRF_PERCALL_SETUP();
5255 4716 : pindex = (int *) funcctx->user_fctx;
5256 :
5257 4716 : if (zoneabbrevtbl == NULL ||
5258 4716 : *pindex >= zoneabbrevtbl->numabbrevs)
5259 24 : SRF_RETURN_DONE(funcctx);
5260 :
5261 4692 : tp = zoneabbrevtbl->abbrevs + *pindex;
5262 :
5263 4692 : switch (tp->type)
5264 : {
5265 2346 : case TZ:
5266 2346 : gmtoffset = tp->value;
5267 2346 : is_dst = false;
5268 2346 : break;
5269 1152 : case DTZ:
5270 1152 : gmtoffset = tp->value;
5271 1152 : is_dst = true;
5272 1152 : break;
5273 1194 : case DYNTZ:
5274 : {
5275 : /* Determine the current meaning of the abbrev */
5276 : pg_tz *tzp;
5277 : DateTimeErrorExtra extra;
5278 : TimestampTz now;
5279 : int isdst;
5280 :
5281 1194 : tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp, &extra);
5282 1194 : if (tzp == NULL)
5283 0 : DateTimeParseError(DTERR_BAD_ZONE_ABBREV, &extra,
5284 : NULL, NULL, NULL);
5285 1194 : now = GetCurrentTransactionStartTimestamp();
5286 2388 : gmtoffset = -DetermineTimeZoneAbbrevOffsetTS(now,
5287 1194 : tp->token,
5288 : tzp,
5289 : &isdst);
5290 1194 : is_dst = (bool) isdst;
5291 1194 : break;
5292 : }
5293 0 : default:
5294 0 : elog(ERROR, "unrecognized timezone type %d", (int) tp->type);
5295 : gmtoffset = 0; /* keep compiler quiet */
5296 : is_dst = false;
5297 : break;
5298 : }
5299 :
5300 : /*
5301 : * Convert name to text, using upcasing conversion that is the inverse of
5302 : * what ParseDateTime() uses.
5303 : */
5304 4692 : strlcpy(buffer, tp->token, sizeof(buffer));
5305 21744 : for (p = (unsigned char *) buffer; *p; p++)
5306 17052 : *p = pg_toupper(*p);
5307 :
5308 4692 : values[0] = CStringGetTextDatum(buffer);
5309 :
5310 : /* Convert offset (in seconds) to an interval; can't overflow */
5311 18768 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5312 4692 : itm_in.tm_usec = (int64) gmtoffset * USECS_PER_SEC;
5313 4692 : resInterval = (Interval *) palloc(sizeof(Interval));
5314 4692 : (void) itmin2interval(&itm_in, resInterval);
5315 4692 : values[1] = IntervalPGetDatum(resInterval);
5316 :
5317 4692 : values[2] = BoolGetDatum(is_dst);
5318 :
5319 4692 : (*pindex)++;
5320 :
5321 4692 : tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
5322 4692 : result = HeapTupleGetDatum(tuple);
5323 :
5324 4692 : SRF_RETURN_NEXT(funcctx, result);
5325 : }
5326 :
5327 : /*
5328 : * This set-returning function reads all the available full time zones
5329 : * and returns a set of (name, abbrev, utc_offset, is_dst).
5330 : */
5331 : Datum
5332 16 : pg_timezone_names(PG_FUNCTION_ARGS)
5333 : {
5334 16 : ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
5335 : pg_tzenum *tzenum;
5336 : pg_tz *tz;
5337 : Datum values[4];
5338 16 : bool nulls[4] = {0};
5339 : int tzoff;
5340 : struct pg_tm tm;
5341 : fsec_t fsec;
5342 : const char *tzn;
5343 : Interval *resInterval;
5344 : struct pg_itm_in itm_in;
5345 :
5346 16 : InitMaterializedSRF(fcinfo, 0);
5347 :
5348 : /* initialize timezone scanning code */
5349 16 : tzenum = pg_tzenumerate_start();
5350 :
5351 : /* search for another zone to display */
5352 : for (;;)
5353 : {
5354 9584 : tz = pg_tzenumerate_next(tzenum);
5355 9584 : if (!tz)
5356 16 : break;
5357 :
5358 : /* Convert now() to local time in this zone */
5359 9568 : if (timestamp2tm(GetCurrentTransactionStartTimestamp(),
5360 : &tzoff, &tm, &fsec, &tzn, tz) != 0)
5361 0 : continue; /* ignore if conversion fails */
5362 :
5363 : /*
5364 : * IANA's rather silly "Factory" time zone used to emit ridiculously
5365 : * long "abbreviations" such as "Local time zone must be set--see zic
5366 : * manual page" or "Local time zone must be set--use tzsetup". While
5367 : * modern versions of tzdb emit the much saner "-00", it seems some
5368 : * benighted packagers are hacking the IANA data so that it continues
5369 : * to produce these strings. To prevent producing a weirdly wide
5370 : * abbrev column, reject ridiculously long abbreviations.
5371 : */
5372 9568 : if (tzn && strlen(tzn) > 31)
5373 0 : continue;
5374 :
5375 9568 : values[0] = CStringGetTextDatum(pg_get_timezone_name(tz));
5376 9568 : values[1] = CStringGetTextDatum(tzn ? tzn : "");
5377 :
5378 : /* Convert tzoff to an interval; can't overflow */
5379 38272 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5380 9568 : itm_in.tm_usec = (int64) -tzoff * USECS_PER_SEC;
5381 9568 : resInterval = (Interval *) palloc(sizeof(Interval));
5382 9568 : (void) itmin2interval(&itm_in, resInterval);
5383 9568 : values[2] = IntervalPGetDatum(resInterval);
5384 :
5385 9568 : values[3] = BoolGetDatum(tm.tm_isdst > 0);
5386 :
5387 9568 : tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls);
5388 : }
5389 :
5390 16 : pg_tzenumerate_end(tzenum);
5391 16 : return (Datum) 0;
5392 : }
|