Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * datetime.c
4 : * Support functions for date/time types.
5 : *
6 : * Portions Copyright (c) 1996-2026, 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/tuplestore.h"
34 : #include "utils/tzparser.h"
35 :
36 : static int DecodeNumber(int flen, char *str, bool haveTextMonth,
37 : int fmask, int *tmask,
38 : struct pg_tm *tm, fsec_t *fsec, bool *is2digits);
39 : static int DecodeNumberField(int len, char *str,
40 : int fmask, int *tmask,
41 : struct pg_tm *tm, fsec_t *fsec, bool *is2digits);
42 : static int DecodeTimeCommon(char *str, int fmask, int range,
43 : int *tmask, struct pg_itm *itm);
44 : static int DecodeTime(char *str, int fmask, int range,
45 : int *tmask, struct pg_tm *tm, fsec_t *fsec);
46 : static int DecodeTimeForInterval(char *str, int fmask, int range,
47 : int *tmask, struct pg_itm_in *itm_in);
48 : static const datetkn *datebsearch(const char *key, const datetkn *base, int nel);
49 : static int DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
50 : struct pg_tm *tm);
51 : static char *AppendSeconds(char *cp, int sec, fsec_t fsec,
52 : int precision, bool fillzeros);
53 : static bool int64_multiply_add(int64 val, int64 multiplier, int64 *sum);
54 : static bool AdjustFractMicroseconds(double frac, int64 scale,
55 : struct pg_itm_in *itm_in);
56 : static bool AdjustFractDays(double frac, int scale,
57 : struct pg_itm_in *itm_in);
58 : static bool AdjustFractYears(double frac, int scale,
59 : struct pg_itm_in *itm_in);
60 : static bool AdjustMicroseconds(int64 val, double fval, int64 scale,
61 : struct pg_itm_in *itm_in);
62 : static bool AdjustDays(int64 val, int scale,
63 : struct pg_itm_in *itm_in);
64 : static bool AdjustMonths(int64 val, struct pg_itm_in *itm_in);
65 : static bool AdjustYears(int64 val, int scale,
66 : struct pg_itm_in *itm_in);
67 : static int DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp,
68 : pg_time_t *tp);
69 : static bool DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t,
70 : const char *abbr, pg_tz *tzp,
71 : int *offset, int *isdst);
72 : static pg_tz *FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp,
73 : DateTimeErrorExtra *extra);
74 :
75 :
76 : const int day_tab[2][13] =
77 : {
78 : {31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0},
79 : {31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0}
80 : };
81 :
82 : const char *const months[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun",
83 : "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", NULL};
84 :
85 : const char *const days[] = {"Sunday", "Monday", "Tuesday", "Wednesday",
86 : "Thursday", "Friday", "Saturday", NULL};
87 :
88 :
89 : /*****************************************************************************
90 : * PRIVATE ROUTINES *
91 : *****************************************************************************/
92 :
93 : /*
94 : * datetktbl holds date/time keywords.
95 : *
96 : * Note that this table must be strictly alphabetically ordered to allow an
97 : * O(ln(N)) search algorithm to be used.
98 : *
99 : * The token field must be NUL-terminated; we truncate entries to TOKMAXLEN
100 : * characters to fit.
101 : *
102 : * The static table contains no TZ, DTZ, or DYNTZ entries; rather those
103 : * are loaded from configuration files and stored in zoneabbrevtbl, whose
104 : * abbrevs[] field has the same format as the static datetktbl.
105 : */
106 : static const datetkn datetktbl[] = {
107 : /* token, type, value */
108 : {"+infinity", RESERV, DTK_LATE}, /* same as "infinity" */
109 : {EARLY, RESERV, DTK_EARLY}, /* "-infinity" reserved for "early time" */
110 : {DA_D, ADBC, AD}, /* "ad" for years > 0 */
111 : {"allballs", RESERV, DTK_ZULU}, /* 00:00:00 */
112 : {"am", AMPM, AM},
113 : {"apr", MONTH, 4},
114 : {"april", MONTH, 4},
115 : {"at", IGNORE_DTF, 0}, /* "at" (throwaway) */
116 : {"aug", MONTH, 8},
117 : {"august", MONTH, 8},
118 : {DB_C, ADBC, BC}, /* "bc" for years <= 0 */
119 : {"d", UNITS, DTK_DAY}, /* "day of month" for ISO input */
120 : {"dec", MONTH, 12},
121 : {"december", MONTH, 12},
122 : {"dow", UNITS, DTK_DOW}, /* day of week */
123 : {"doy", UNITS, DTK_DOY}, /* day of year */
124 : {"dst", DTZMOD, SECS_PER_HOUR},
125 : {EPOCH, RESERV, DTK_EPOCH}, /* "epoch" reserved for system epoch time */
126 : {"feb", MONTH, 2},
127 : {"february", MONTH, 2},
128 : {"fri", DOW, 5},
129 : {"friday", DOW, 5},
130 : {"h", UNITS, DTK_HOUR}, /* "hour" */
131 : {LATE, RESERV, DTK_LATE}, /* "infinity" reserved for "late time" */
132 : {"isodow", UNITS, DTK_ISODOW}, /* ISO day of week, Sunday == 7 */
133 : {"isoyear", UNITS, DTK_ISOYEAR}, /* year in terms of the ISO week date */
134 : {"j", UNITS, DTK_JULIAN},
135 : {"jan", MONTH, 1},
136 : {"january", MONTH, 1},
137 : {"jd", UNITS, DTK_JULIAN},
138 : {"jul", MONTH, 7},
139 : {"julian", UNITS, DTK_JULIAN},
140 : {"july", MONTH, 7},
141 : {"jun", MONTH, 6},
142 : {"june", MONTH, 6},
143 : {"m", UNITS, DTK_MONTH}, /* "month" for ISO input */
144 : {"mar", MONTH, 3},
145 : {"march", MONTH, 3},
146 : {"may", MONTH, 5},
147 : {"mm", UNITS, DTK_MINUTE}, /* "minute" for ISO input */
148 : {"mon", DOW, 1},
149 : {"monday", DOW, 1},
150 : {"nov", MONTH, 11},
151 : {"november", MONTH, 11},
152 : {NOW, RESERV, DTK_NOW}, /* current transaction time */
153 : {"oct", MONTH, 10},
154 : {"october", MONTH, 10},
155 : {"on", IGNORE_DTF, 0}, /* "on" (throwaway) */
156 : {"pm", AMPM, PM},
157 : {"s", UNITS, DTK_SECOND}, /* "seconds" for ISO input */
158 : {"sat", DOW, 6},
159 : {"saturday", DOW, 6},
160 : {"sep", MONTH, 9},
161 : {"sept", MONTH, 9},
162 : {"september", MONTH, 9},
163 : {"sun", DOW, 0},
164 : {"sunday", DOW, 0},
165 : {"t", ISOTIME, DTK_TIME}, /* Filler for ISO time fields */
166 : {"thu", DOW, 4},
167 : {"thur", DOW, 4},
168 : {"thurs", DOW, 4},
169 : {"thursday", DOW, 4},
170 : {TODAY, RESERV, DTK_TODAY}, /* midnight */
171 : {TOMORROW, RESERV, DTK_TOMORROW}, /* tomorrow midnight */
172 : {"tue", DOW, 2},
173 : {"tues", DOW, 2},
174 : {"tuesday", DOW, 2},
175 : {"wed", DOW, 3},
176 : {"wednesday", DOW, 3},
177 : {"weds", DOW, 3},
178 : {"y", UNITS, DTK_YEAR}, /* "year" for ISO input */
179 : {YESTERDAY, RESERV, DTK_YESTERDAY} /* yesterday midnight */
180 : };
181 :
182 : static const int szdatetktbl = sizeof datetktbl / sizeof datetktbl[0];
183 :
184 : /*
185 : * deltatktbl: same format as datetktbl, but holds keywords used to represent
186 : * time units (eg, for intervals, and for EXTRACT).
187 : */
188 : static const datetkn deltatktbl[] = {
189 : /* token, type, value */
190 : {"@", IGNORE_DTF, 0}, /* postgres relative prefix */
191 : {DAGO, AGO, 0}, /* "ago" indicates negative time offset */
192 : {"c", UNITS, DTK_CENTURY}, /* "century" relative */
193 : {"cent", UNITS, DTK_CENTURY}, /* "century" relative */
194 : {"centuries", UNITS, DTK_CENTURY}, /* "centuries" relative */
195 : {DCENTURY, UNITS, DTK_CENTURY}, /* "century" relative */
196 : {"d", UNITS, DTK_DAY}, /* "day" relative */
197 : {DDAY, UNITS, DTK_DAY}, /* "day" relative */
198 : {"days", UNITS, DTK_DAY}, /* "days" relative */
199 : {"dec", UNITS, DTK_DECADE}, /* "decade" relative */
200 : {DDECADE, UNITS, DTK_DECADE}, /* "decade" relative */
201 : {"decades", UNITS, DTK_DECADE}, /* "decades" relative */
202 : {"decs", UNITS, DTK_DECADE}, /* "decades" relative */
203 : {"h", UNITS, DTK_HOUR}, /* "hour" relative */
204 : {DHOUR, UNITS, DTK_HOUR}, /* "hour" relative */
205 : {"hours", UNITS, DTK_HOUR}, /* "hours" relative */
206 : {"hr", UNITS, DTK_HOUR}, /* "hour" relative */
207 : {"hrs", UNITS, DTK_HOUR}, /* "hours" relative */
208 : {"m", UNITS, DTK_MINUTE}, /* "minute" relative */
209 : {"microsecon", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
210 : {"mil", UNITS, DTK_MILLENNIUM}, /* "millennium" relative */
211 : {"millennia", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */
212 : {DMILLENNIUM, UNITS, DTK_MILLENNIUM}, /* "millennium" relative */
213 : {"millisecon", UNITS, DTK_MILLISEC}, /* relative */
214 : {"mils", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */
215 : {"min", UNITS, DTK_MINUTE}, /* "minute" relative */
216 : {"mins", UNITS, DTK_MINUTE}, /* "minutes" relative */
217 : {DMINUTE, UNITS, DTK_MINUTE}, /* "minute" relative */
218 : {"minutes", UNITS, DTK_MINUTE}, /* "minutes" relative */
219 : {"mon", UNITS, DTK_MONTH}, /* "months" relative */
220 : {"mons", UNITS, DTK_MONTH}, /* "months" relative */
221 : {DMONTH, UNITS, DTK_MONTH}, /* "month" relative */
222 : {"months", UNITS, DTK_MONTH},
223 : {"ms", UNITS, DTK_MILLISEC},
224 : {"msec", UNITS, DTK_MILLISEC},
225 : {DMILLISEC, UNITS, DTK_MILLISEC},
226 : {"mseconds", UNITS, DTK_MILLISEC},
227 : {"msecs", UNITS, DTK_MILLISEC},
228 : {"qtr", UNITS, DTK_QUARTER}, /* "quarter" relative */
229 : {DQUARTER, UNITS, DTK_QUARTER}, /* "quarter" relative */
230 : {"s", UNITS, DTK_SECOND},
231 : {"sec", UNITS, DTK_SECOND},
232 : {DSECOND, UNITS, DTK_SECOND},
233 : {"seconds", UNITS, DTK_SECOND},
234 : {"secs", UNITS, DTK_SECOND},
235 : {DTIMEZONE, UNITS, DTK_TZ}, /* "timezone" time offset */
236 : {"timezone_h", UNITS, DTK_TZ_HOUR}, /* timezone hour units */
237 : {"timezone_m", UNITS, DTK_TZ_MINUTE}, /* timezone minutes units */
238 : {"us", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
239 : {"usec", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
240 : {DMICROSEC, UNITS, DTK_MICROSEC}, /* "microsecond" relative */
241 : {"useconds", UNITS, DTK_MICROSEC}, /* "microseconds" relative */
242 : {"usecs", UNITS, DTK_MICROSEC}, /* "microseconds" relative */
243 : {"w", UNITS, DTK_WEEK}, /* "week" relative */
244 : {DWEEK, UNITS, DTK_WEEK}, /* "week" relative */
245 : {"weeks", UNITS, DTK_WEEK}, /* "weeks" relative */
246 : {"y", UNITS, DTK_YEAR}, /* "year" relative */
247 : {DYEAR, UNITS, DTK_YEAR}, /* "year" relative */
248 : {"years", UNITS, DTK_YEAR}, /* "years" relative */
249 : {"yr", UNITS, DTK_YEAR}, /* "year" relative */
250 : {"yrs", UNITS, DTK_YEAR} /* "years" relative */
251 : };
252 :
253 : static const int szdeltatktbl = sizeof deltatktbl / sizeof deltatktbl[0];
254 :
255 : static TimeZoneAbbrevTable *zoneabbrevtbl = NULL;
256 :
257 : /* Caches of recent lookup results in the above tables */
258 :
259 : static const datetkn *datecache[MAXDATEFIELDS] = {NULL};
260 :
261 : static const datetkn *deltacache[MAXDATEFIELDS] = {NULL};
262 :
263 : /* Cache for results of timezone abbreviation lookups */
264 :
265 : typedef struct TzAbbrevCache
266 : {
267 : char abbrev[TOKMAXLEN + 1]; /* always NUL-terminated */
268 : char ftype; /* TZ, DTZ, or DYNTZ */
269 : int offset; /* GMT offset, if fixed-offset */
270 : pg_tz *tz; /* relevant zone, if variable-offset */
271 : } TzAbbrevCache;
272 :
273 : static TzAbbrevCache tzabbrevcache[MAXDATEFIELDS];
274 :
275 :
276 : /*
277 : * Calendar time to Julian date conversions.
278 : * Julian date is commonly used in astronomical applications,
279 : * since it is numerically accurate and computationally simple.
280 : * The algorithms here will accurately convert between Julian day
281 : * and calendar date for all non-negative Julian days
282 : * (i.e. from Nov 24, -4713 on).
283 : *
284 : * Rewritten to eliminate overflow problems. This now allows the
285 : * routines to work correctly for all Julian day counts from
286 : * 0 to 2147483647 (Nov 24, -4713 to Jun 3, 5874898) assuming
287 : * a 32-bit integer. Longer types should also work to the limits
288 : * of their precision.
289 : *
290 : * Actually, date2j() will work sanely, in the sense of producing
291 : * valid negative Julian dates, significantly before Nov 24, -4713.
292 : * We rely on it to do so back to Nov 1, -4713; see IS_VALID_JULIAN()
293 : * and associated commentary in timestamp.h.
294 : */
295 :
296 : int
297 283665 : date2j(int year, int month, int day)
298 : {
299 : int julian;
300 : int century;
301 :
302 283665 : if (month > 2)
303 : {
304 182353 : month += 1;
305 182353 : year += 4800;
306 : }
307 : else
308 : {
309 101312 : month += 13;
310 101312 : year += 4799;
311 : }
312 :
313 283665 : century = year / 100;
314 283665 : julian = year * 365 - 32167;
315 283665 : julian += year / 4 - century + century / 4;
316 283665 : julian += 7834 * month / 256 + day;
317 :
318 283665 : return julian;
319 : } /* date2j() */
320 :
321 : void
322 200801 : j2date(int jd, int *year, int *month, int *day)
323 : {
324 : unsigned int julian;
325 : unsigned int quad;
326 : unsigned int extra;
327 : int y;
328 :
329 200801 : julian = jd;
330 200801 : julian += 32044;
331 200801 : quad = julian / 146097;
332 200801 : extra = (julian - quad * 146097) * 4 + 3;
333 200801 : julian += 60 + quad * 3 + extra / 146097;
334 200801 : quad = julian / 1461;
335 200801 : julian -= quad * 1461;
336 200801 : y = julian * 4 / 1461;
337 401602 : julian = ((y != 0) ? ((julian + 305) % 365) : ((julian + 306) % 366))
338 200801 : + 123;
339 200801 : y += quad * 4;
340 200801 : *year = y - 4800;
341 200801 : quad = julian * 2141 / 65536;
342 200801 : *day = julian - 7834 * quad / 256;
343 200801 : *month = (quad + 10) % MONTHS_PER_YEAR + 1;
344 200801 : } /* j2date() */
345 :
346 :
347 : /*
348 : * j2day - convert Julian date to day-of-week (0..6 == Sun..Sat)
349 : *
350 : * Note: various places use the locution j2day(date - 1) to produce a
351 : * result according to the convention 0..6 = Mon..Sun. This is a bit of
352 : * a crock, but will work as long as the computation here is just a modulo.
353 : */
354 : int
355 34085 : j2day(int date)
356 : {
357 34085 : date += 1;
358 34085 : date %= 7;
359 : /* Cope if division truncates towards zero, as it probably does */
360 34085 : if (date < 0)
361 0 : date += 7;
362 :
363 34085 : return date;
364 : } /* j2day() */
365 :
366 :
367 : /*
368 : * GetCurrentDateTime()
369 : *
370 : * Get the transaction start time ("now()") broken down as a struct pg_tm,
371 : * converted according to the session timezone setting.
372 : *
373 : * This is just a convenience wrapper for GetCurrentTimeUsec, to cover the
374 : * case where caller doesn't need either fractional seconds or tz offset.
375 : */
376 : void
377 1941 : GetCurrentDateTime(struct pg_tm *tm)
378 : {
379 : fsec_t fsec;
380 :
381 1941 : GetCurrentTimeUsec(tm, &fsec, NULL);
382 1941 : }
383 :
384 : /*
385 : * GetCurrentTimeUsec()
386 : *
387 : * Get the transaction start time ("now()") broken down as a struct pg_tm,
388 : * including fractional seconds and timezone offset. The time is converted
389 : * according to the session timezone setting.
390 : *
391 : * Callers may pass tzp = NULL if they don't need the offset, but this does
392 : * not affect the conversion behavior (unlike timestamp2tm()).
393 : *
394 : * Internally, we cache the result, since this could be called many times
395 : * in a transaction, within which now() doesn't change.
396 : */
397 : void
398 2057 : GetCurrentTimeUsec(struct pg_tm *tm, fsec_t *fsec, int *tzp)
399 : {
400 2057 : TimestampTz cur_ts = GetCurrentTransactionStartTimestamp();
401 :
402 : /*
403 : * The cache key must include both current time and current timezone. By
404 : * representing the timezone by just a pointer, we're assuming that
405 : * distinct timezone settings could never have the same pointer value.
406 : * This is true by virtue of the hashtable used inside pg_tzset();
407 : * however, it might need another look if we ever allow entries in that
408 : * hash to be recycled.
409 : */
410 : static TimestampTz cache_ts = 0;
411 : static pg_tz *cache_timezone = NULL;
412 : static struct pg_tm cache_tm;
413 : static fsec_t cache_fsec;
414 : static int cache_tz;
415 :
416 2057 : if (cur_ts != cache_ts || session_timezone != cache_timezone)
417 : {
418 : /*
419 : * Make sure cache is marked invalid in case of error after partial
420 : * update within timestamp2tm.
421 : */
422 658 : cache_timezone = NULL;
423 :
424 : /*
425 : * Perform the computation, storing results into cache. We do not
426 : * really expect any error here, since current time surely ought to be
427 : * within range, but check just for sanity's sake.
428 : */
429 658 : if (timestamp2tm(cur_ts, &cache_tz, &cache_tm, &cache_fsec,
430 : NULL, session_timezone) != 0)
431 0 : ereport(ERROR,
432 : (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
433 : errmsg("timestamp out of range")));
434 :
435 : /* OK, so mark the cache valid. */
436 658 : cache_ts = cur_ts;
437 658 : cache_timezone = session_timezone;
438 : }
439 :
440 2057 : *tm = cache_tm;
441 2057 : *fsec = cache_fsec;
442 2057 : if (tzp != NULL)
443 108 : *tzp = cache_tz;
444 2057 : }
445 :
446 :
447 : /*
448 : * Append seconds and fractional seconds (if any) at *cp.
449 : *
450 : * precision is the max number of fraction digits, fillzeros says to
451 : * pad to two integral-seconds digits.
452 : *
453 : * Returns a pointer to the new end of string. No NUL terminator is put
454 : * there; callers are responsible for NUL terminating str themselves.
455 : *
456 : * Note that any sign is stripped from the input sec and fsec values.
457 : */
458 : static char *
459 77643 : AppendSeconds(char *cp, int sec, fsec_t fsec, int precision, bool fillzeros)
460 : {
461 : Assert(precision >= 0);
462 :
463 77643 : if (fillzeros)
464 75445 : cp = pg_ultostr_zeropad(cp, abs(sec), 2);
465 : else
466 2198 : cp = pg_ultostr(cp, abs(sec));
467 :
468 : /* fsec_t is just an int32 */
469 77643 : if (fsec != 0)
470 : {
471 12313 : int32 value = abs(fsec);
472 12313 : char *end = &cp[precision + 1];
473 12313 : bool gotnonzero = false;
474 :
475 12313 : *cp++ = '.';
476 :
477 : /*
478 : * Append the fractional seconds part. Note that we don't want any
479 : * trailing zeros here, so since we're building the number in reverse
480 : * we'll skip appending zeros until we've output a non-zero digit.
481 : */
482 86191 : while (precision--)
483 : {
484 73878 : int32 oldval = value;
485 : int32 remainder;
486 :
487 73878 : value /= 10;
488 73878 : remainder = oldval - value * 10;
489 :
490 : /* check if we got a non-zero */
491 73878 : if (remainder)
492 53368 : gotnonzero = true;
493 :
494 73878 : if (gotnonzero)
495 61986 : cp[precision] = '0' + remainder;
496 : else
497 11892 : end = &cp[precision];
498 : }
499 :
500 : /*
501 : * If we still have a non-zero value then precision must have not been
502 : * enough to print the number. We punt the problem to pg_ultostr(),
503 : * which will generate a correct answer in the minimum valid width.
504 : */
505 12313 : if (value)
506 0 : return pg_ultostr(cp, abs(fsec));
507 :
508 12313 : return end;
509 : }
510 : else
511 65330 : return cp;
512 : }
513 :
514 :
515 : /*
516 : * Variant of above that's specialized to timestamp case.
517 : *
518 : * Returns a pointer to the new end of string. No NUL terminator is put
519 : * there; callers are responsible for NUL terminating str themselves.
520 : */
521 : static char *
522 64153 : AppendTimestampSeconds(char *cp, struct pg_tm *tm, fsec_t fsec)
523 : {
524 64153 : return AppendSeconds(cp, tm->tm_sec, fsec, MAX_TIMESTAMP_PRECISION, true);
525 : }
526 :
527 :
528 : /*
529 : * Add val * multiplier to *sum.
530 : * Returns true if successful, false on overflow.
531 : */
532 : static bool
533 5793 : int64_multiply_add(int64 val, int64 multiplier, int64 *sum)
534 : {
535 : int64 product;
536 :
537 11546 : if (pg_mul_s64_overflow(val, multiplier, &product) ||
538 5753 : pg_add_s64_overflow(*sum, product, sum))
539 104 : return false;
540 5689 : return true;
541 : }
542 :
543 : /*
544 : * Multiply frac by scale (to produce microseconds) and add to itm_in->tm_usec.
545 : * Returns true if successful, false if itm_in overflows.
546 : */
547 : static bool
548 5900 : AdjustFractMicroseconds(double frac, int64 scale,
549 : struct pg_itm_in *itm_in)
550 : {
551 : int64 usec;
552 :
553 : /* Fast path for common case */
554 5900 : if (frac == 0)
555 5592 : return true;
556 :
557 : /*
558 : * We assume the input frac has abs value less than 1, so overflow of frac
559 : * or usec is not an issue for interesting values of scale.
560 : */
561 308 : frac *= scale;
562 308 : usec = (int64) frac;
563 :
564 : /* Round off any fractional microsecond */
565 308 : frac -= usec;
566 308 : if (frac > 0.5)
567 16 : usec++;
568 292 : else if (frac < -0.5)
569 20 : usec--;
570 :
571 308 : return !pg_add_s64_overflow(itm_in->tm_usec, usec, &itm_in->tm_usec);
572 : }
573 :
574 : /*
575 : * Multiply frac by scale (to produce days). Add the integral part of the
576 : * result to itm_in->tm_mday, the fractional part to itm_in->tm_usec.
577 : * Returns true if successful, false if itm_in overflows.
578 : */
579 : static bool
580 868 : AdjustFractDays(double frac, int scale,
581 : struct pg_itm_in *itm_in)
582 : {
583 : int extra_days;
584 :
585 : /* Fast path for common case */
586 868 : if (frac == 0)
587 744 : return true;
588 :
589 : /*
590 : * We assume the input frac has abs value less than 1, so overflow of frac
591 : * or extra_days is not an issue.
592 : */
593 124 : frac *= scale;
594 124 : extra_days = (int) frac;
595 :
596 : /* ... but this could overflow, if tm_mday is already nonzero */
597 124 : if (pg_add_s32_overflow(itm_in->tm_mday, extra_days, &itm_in->tm_mday))
598 32 : return false;
599 :
600 : /* Handle any fractional day */
601 92 : frac -= extra_days;
602 92 : return AdjustFractMicroseconds(frac, USECS_PER_DAY, itm_in);
603 : }
604 :
605 : /*
606 : * Multiply frac by scale (to produce years), then further scale up to months.
607 : * Add the integral part of the result to itm_in->tm_mon, discarding any
608 : * fractional part.
609 : * Returns true if successful, false if itm_in overflows.
610 : */
611 : static bool
612 36484 : AdjustFractYears(double frac, int scale,
613 : struct pg_itm_in *itm_in)
614 : {
615 : /*
616 : * As above, we assume abs(frac) < 1, so this can't overflow for any
617 : * interesting value of scale.
618 : */
619 36484 : int extra_months = (int) rint(frac * scale * MONTHS_PER_YEAR);
620 :
621 36484 : return !pg_add_s32_overflow(itm_in->tm_mon, extra_months, &itm_in->tm_mon);
622 : }
623 :
624 : /*
625 : * Add (val + fval) * scale to itm_in->tm_usec.
626 : * Returns true if successful, false if itm_in overflows.
627 : */
628 : static bool
629 1842 : AdjustMicroseconds(int64 val, double fval, int64 scale,
630 : struct pg_itm_in *itm_in)
631 : {
632 : /* Handle the integer part */
633 1842 : if (!int64_multiply_add(val, scale, &itm_in->tm_usec))
634 100 : return false;
635 : /* Handle the float part */
636 1742 : return AdjustFractMicroseconds(fval, scale, itm_in);
637 : }
638 :
639 : /*
640 : * Multiply val by scale (to produce days) and add to itm_in->tm_mday.
641 : * Returns true if successful, false if itm_in overflows.
642 : */
643 : static bool
644 4238 : AdjustDays(int64 val, int scale, struct pg_itm_in *itm_in)
645 : {
646 : int days;
647 :
648 4238 : if (val < INT_MIN || val > INT_MAX)
649 24 : return false;
650 8420 : return !pg_mul_s32_overflow((int32) val, scale, &days) &&
651 4206 : !pg_add_s32_overflow(itm_in->tm_mday, days, &itm_in->tm_mday);
652 : }
653 :
654 : /*
655 : * Add val to itm_in->tm_mon (no need for scale here, as val is always
656 : * in months already).
657 : * Returns true if successful, false if itm_in overflows.
658 : */
659 : static bool
660 848 : AdjustMonths(int64 val, struct pg_itm_in *itm_in)
661 : {
662 848 : if (val < INT_MIN || val > INT_MAX)
663 8 : return false;
664 840 : return !pg_add_s32_overflow(itm_in->tm_mon, (int32) val, &itm_in->tm_mon);
665 : }
666 :
667 : /*
668 : * Multiply val by scale (to produce years) and add to itm_in->tm_year.
669 : * Returns true if successful, false if itm_in overflows.
670 : */
671 : static bool
672 36576 : AdjustYears(int64 val, int scale,
673 : struct pg_itm_in *itm_in)
674 : {
675 : int years;
676 :
677 36576 : if (val < INT_MIN || val > INT_MAX)
678 16 : return false;
679 73096 : return !pg_mul_s32_overflow((int32) val, scale, &years) &&
680 36536 : !pg_add_s32_overflow(itm_in->tm_year, years, &itm_in->tm_year);
681 : }
682 :
683 :
684 : /*
685 : * Parse the fractional part of a number (decimal point and optional digits,
686 : * followed by end of string). Returns the fractional value into *frac.
687 : *
688 : * Returns 0 if successful, DTERR code if bogus input detected.
689 : */
690 : static int
691 12120 : ParseFraction(char *cp, double *frac)
692 : {
693 : /* Caller should always pass the start of the fraction part */
694 : Assert(*cp == '.');
695 :
696 : /*
697 : * We want to allow just "." with no digits, but some versions of strtod
698 : * will report EINVAL for that, so special-case it.
699 : */
700 12120 : if (cp[1] == '\0')
701 : {
702 0 : *frac = 0;
703 : }
704 : else
705 : {
706 : /*
707 : * On the other hand, let's reject anything that's not digits after
708 : * the ".". strtod is happy with input like ".123e9", but that'd
709 : * break callers' expectation that the result is in 0..1. (It's quite
710 : * difficult to get here with such input, but not impossible.)
711 : */
712 12120 : if (strspn(cp + 1, "0123456789") != strlen(cp + 1))
713 8 : return DTERR_BAD_FORMAT;
714 :
715 12112 : errno = 0;
716 12112 : *frac = strtod(cp, &cp);
717 : /* check for parse failure (probably redundant given prior check) */
718 12112 : if (*cp != '\0' || errno != 0)
719 0 : return DTERR_BAD_FORMAT;
720 : }
721 12112 : return 0;
722 : }
723 :
724 : /*
725 : * Fetch a fractional-second value with suitable error checking.
726 : * Same as ParseFraction except we convert the result to integer microseconds.
727 : */
728 : static int
729 11788 : ParseFractionalSecond(char *cp, fsec_t *fsec)
730 : {
731 : double frac;
732 : int dterr;
733 :
734 11788 : dterr = ParseFraction(cp, &frac);
735 11788 : if (dterr)
736 8 : return dterr;
737 11780 : *fsec = rint(frac * 1000000);
738 11780 : return 0;
739 : }
740 :
741 :
742 : /*
743 : * ParseDateTime()
744 : * Break string into tokens based on a date/time context.
745 : * Returns 0 if successful, DTERR code if bogus input detected.
746 : *
747 : * timestr - the input string
748 : * workbuf - workspace for field string storage. This must be
749 : * larger than the largest legal input for this datetime type --
750 : * some additional space will be needed to NUL terminate fields.
751 : * buflen - the size of workbuf
752 : * field[] - pointers to field strings are returned in this array
753 : * ftype[] - field type indicators are returned in this array
754 : * maxfields - dimensions of the above two arrays
755 : * *numfields - set to the actual number of fields detected
756 : *
757 : * The fields extracted from the input are stored as separate,
758 : * null-terminated strings in the workspace at workbuf. Any text is
759 : * converted to lower case.
760 : *
761 : * Several field types are assigned:
762 : * DTK_NUMBER - digits and (possibly) a decimal point
763 : * DTK_DATE - digits and two delimiters, or digits and text
764 : * DTK_TIME - digits, colon delimiters, and possibly a decimal point
765 : * DTK_STRING - text (no digits or punctuation)
766 : * DTK_SPECIAL - leading "+" or "-" followed by text
767 : * DTK_TZ - leading "+" or "-" followed by digits (also eats ':', '.', '-')
768 : *
769 : * Note that some field types can hold unexpected items:
770 : * DTK_NUMBER can hold date fields (yy.ddd)
771 : * DTK_STRING can hold months (January) and time zones (PST)
772 : * DTK_DATE can hold time zone names (America/New_York, GMT-8)
773 : */
774 : int
775 88942 : ParseDateTime(const char *timestr, char *workbuf, size_t buflen,
776 : char **field, int *ftype, int maxfields, int *numfields)
777 : {
778 88942 : int nf = 0;
779 88942 : const char *cp = timestr;
780 88942 : char *bufp = workbuf;
781 88942 : const char *bufend = workbuf + buflen;
782 :
783 : /*
784 : * Set the character pointed-to by "bufptr" to "newchar", and increment
785 : * "bufptr". "end" gives the end of the buffer -- we return an error if
786 : * there is no space left to append a character to the buffer. Note that
787 : * "bufptr" is evaluated twice.
788 : */
789 : #define APPEND_CHAR(bufptr, end, newchar) \
790 : do \
791 : { \
792 : if (((bufptr) + 1) >= (end)) \
793 : return DTERR_BAD_FORMAT; \
794 : *(bufptr)++ = newchar; \
795 : } while (0)
796 :
797 : /* outer loop through fields */
798 361994 : while (*cp != '\0')
799 : {
800 : /* Ignore spaces between fields */
801 273052 : if (isspace((unsigned char) *cp))
802 : {
803 82102 : cp++;
804 82102 : continue;
805 : }
806 :
807 : /* Record start of current field */
808 190950 : if (nf >= maxfields)
809 0 : return DTERR_BAD_FORMAT;
810 190950 : field[nf] = bufp;
811 :
812 : /* leading digit? then date or time */
813 190950 : if (isdigit((unsigned char) *cp))
814 : {
815 118288 : APPEND_CHAR(bufp, bufend, *cp++);
816 389505 : while (isdigit((unsigned char) *cp))
817 271217 : APPEND_CHAR(bufp, bufend, *cp++);
818 :
819 : /* time field? */
820 118288 : if (*cp == ':')
821 : {
822 32555 : ftype[nf] = DTK_TIME;
823 32555 : APPEND_CHAR(bufp, bufend, *cp++);
824 32555 : while (isdigit((unsigned char) *cp) ||
825 272249 : (*cp == ':') || (*cp == '.'))
826 239694 : APPEND_CHAR(bufp, bufend, *cp++);
827 : }
828 : /* date field? allow embedded text month */
829 85733 : else if (*cp == '-' || *cp == '/' || *cp == '.')
830 40743 : {
831 : /* save delimiting character to use later */
832 40743 : char delim = *cp;
833 :
834 40743 : APPEND_CHAR(bufp, bufend, *cp++);
835 : /* second field is all digits? then no embedded text month */
836 40743 : if (isdigit((unsigned char) *cp))
837 : {
838 40679 : ftype[nf] = ((delim == '.') ? DTK_NUMBER : DTK_DATE);
839 122061 : while (isdigit((unsigned char) *cp))
840 81382 : APPEND_CHAR(bufp, bufend, *cp++);
841 :
842 : /*
843 : * insist that the delimiters match to get a three-field
844 : * date.
845 : */
846 40679 : if (*cp == delim)
847 : {
848 40275 : ftype[nf] = DTK_DATE;
849 40275 : APPEND_CHAR(bufp, bufend, *cp++);
850 122781 : while (isdigit((unsigned char) *cp) || *cp == delim)
851 82506 : APPEND_CHAR(bufp, bufend, *cp++);
852 : }
853 : }
854 : else
855 : {
856 64 : ftype[nf] = DTK_DATE;
857 504 : while (isalnum((unsigned char) *cp) || *cp == delim)
858 440 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
859 : }
860 : }
861 :
862 : /*
863 : * otherwise, number only and will determine year, month, day, or
864 : * concatenated fields later...
865 : */
866 : else
867 44990 : ftype[nf] = DTK_NUMBER;
868 : }
869 : /* Leading decimal point? Then fractional seconds... */
870 72662 : else if (*cp == '.')
871 : {
872 0 : APPEND_CHAR(bufp, bufend, *cp++);
873 0 : while (isdigit((unsigned char) *cp))
874 0 : APPEND_CHAR(bufp, bufend, *cp++);
875 :
876 0 : ftype[nf] = DTK_NUMBER;
877 : }
878 :
879 : /*
880 : * text? then date string, month, day of week, special, or timezone
881 : */
882 72662 : else if (isalpha((unsigned char) *cp))
883 : {
884 : bool is_date;
885 :
886 50350 : ftype[nf] = DTK_STRING;
887 50350 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
888 238176 : while (isalpha((unsigned char) *cp))
889 187826 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
890 :
891 : /*
892 : * Dates can have embedded '-', '/', or '.' separators. It could
893 : * also be a timezone name containing embedded '/', '+', '-', '_',
894 : * or ':' (but '_' or ':' can't be the first punctuation). If the
895 : * next character is a digit or '+', we need to check whether what
896 : * we have so far is a recognized non-timezone keyword --- if so,
897 : * don't believe that this is the start of a timezone.
898 : */
899 50350 : is_date = false;
900 50350 : if (*cp == '-' || *cp == '/' || *cp == '.')
901 836 : is_date = true;
902 49514 : else if (*cp == '+' || isdigit((unsigned char) *cp))
903 : {
904 1089 : *bufp = '\0'; /* null-terminate current field value */
905 : /* we need search only the core token table, not TZ names */
906 1089 : if (datebsearch(field[nf], datetktbl, szdatetktbl) == NULL)
907 821 : is_date = true;
908 : }
909 50350 : if (is_date)
910 : {
911 1657 : ftype[nf] = DTK_DATE;
912 : do
913 : {
914 8744 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
915 8744 : } while (*cp == '+' || *cp == '-' ||
916 8548 : *cp == '/' || *cp == '_' ||
917 17188 : *cp == '.' || *cp == ':' ||
918 8164 : isalnum((unsigned char) *cp));
919 : }
920 : }
921 : /* sign? then special or numeric timezone */
922 22312 : else if (*cp == '+' || *cp == '-')
923 : {
924 21936 : APPEND_CHAR(bufp, bufend, *cp++);
925 : /* soak up leading whitespace */
926 21952 : while (isspace((unsigned char) *cp))
927 16 : cp++;
928 : /* numeric timezone? */
929 : /* note that "DTK_TZ" could also be a signed float or yyyy-mm */
930 43872 : if (isdigit((unsigned char) *cp))
931 : {
932 21329 : ftype[nf] = DTK_TZ;
933 21329 : APPEND_CHAR(bufp, bufend, *cp++);
934 21329 : while (isdigit((unsigned char) *cp) ||
935 50941 : *cp == ':' || *cp == '.' || *cp == '-')
936 29612 : APPEND_CHAR(bufp, bufend, *cp++);
937 : }
938 : /* special? */
939 607 : else if (isalpha((unsigned char) *cp))
940 : {
941 607 : ftype[nf] = DTK_SPECIAL;
942 607 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
943 4856 : while (isalpha((unsigned char) *cp))
944 4249 : APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
945 : }
946 : /* otherwise something wrong... */
947 : else
948 0 : return DTERR_BAD_FORMAT;
949 : }
950 : /* ignore other punctuation but use as delimiter */
951 376 : else if (ispunct((unsigned char) *cp))
952 : {
953 376 : cp++;
954 376 : continue;
955 : }
956 : /* otherwise, something is not right... */
957 : else
958 0 : return DTERR_BAD_FORMAT;
959 :
960 : /* force in a delimiter after each field */
961 190574 : *bufp++ = '\0';
962 190574 : nf++;
963 : }
964 :
965 88942 : *numfields = nf;
966 :
967 88942 : return 0;
968 : }
969 :
970 :
971 : /*
972 : * DecodeDateTime()
973 : * Interpret previously parsed fields for general date and time.
974 : * Return 0 if full date, 1 if only time, and negative DTERR code if problems.
975 : * (Currently, all callers treat 1 as an error return too.)
976 : *
977 : * Inputs are field[] and ftype[] arrays, of length nf.
978 : * Other arguments are outputs.
979 : *
980 : * External format(s):
981 : * "<weekday> <month>-<day>-<year> <hour>:<minute>:<second>"
982 : * "Fri Feb-7-1997 15:23:27"
983 : * "Feb-7-1997 15:23:27"
984 : * "2-7-1997 15:23:27"
985 : * "1997-2-7 15:23:27"
986 : * "1997.038 15:23:27" (day of year 1-366)
987 : * Also supports input in compact time:
988 : * "970207 152327"
989 : * "97038 152327"
990 : * "20011225T040506.789-07"
991 : *
992 : * Use the system-provided functions to get the current time zone
993 : * if not specified in the input string.
994 : *
995 : * If the date is outside the range of pg_time_t (in practice that could only
996 : * happen if pg_time_t is just 32 bits), then assume UTC time zone - thomas
997 : * 1997-05-27
998 : */
999 : int
1000 43304 : DecodeDateTime(char **field, int *ftype, int nf,
1001 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
1002 : DateTimeErrorExtra *extra)
1003 : {
1004 43304 : int fmask = 0,
1005 : tmask,
1006 : type;
1007 43304 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
1008 : int i;
1009 : int val;
1010 : int dterr;
1011 43304 : int mer = HR24;
1012 43304 : bool haveTextMonth = false;
1013 43304 : bool isjulian = false;
1014 43304 : bool is2digits = false;
1015 43304 : bool bc = false;
1016 43304 : pg_tz *namedTz = NULL;
1017 43304 : pg_tz *abbrevTz = NULL;
1018 : pg_tz *valtz;
1019 43304 : char *abbrev = NULL;
1020 : struct pg_tm cur_tm;
1021 :
1022 : /*
1023 : * We'll insist on at least all of the date fields, but initialize the
1024 : * remaining fields in case they are not set later...
1025 : */
1026 43304 : *dtype = DTK_DATE;
1027 43304 : tm->tm_hour = 0;
1028 43304 : tm->tm_min = 0;
1029 43304 : tm->tm_sec = 0;
1030 43304 : *fsec = 0;
1031 : /* don't know daylight savings time status apriori */
1032 43304 : tm->tm_isdst = -1;
1033 43304 : if (tzp != NULL)
1034 43304 : *tzp = 0;
1035 :
1036 141332 : for (i = 0; i < nf; i++)
1037 : {
1038 98248 : switch (ftype[i])
1039 : {
1040 41007 : case DTK_DATE:
1041 :
1042 : /*
1043 : * Integral julian day with attached time zone? All other
1044 : * forms with JD will be separated into distinct fields, so we
1045 : * handle just this case here.
1046 : */
1047 41007 : if (ptype == DTK_JULIAN)
1048 : {
1049 : char *cp;
1050 : int jday;
1051 :
1052 4 : if (tzp == NULL)
1053 0 : return DTERR_BAD_FORMAT;
1054 :
1055 4 : errno = 0;
1056 4 : jday = strtoint(field[i], &cp, 10);
1057 4 : if (errno == ERANGE || jday < 0)
1058 0 : return DTERR_FIELD_OVERFLOW;
1059 :
1060 4 : j2date(jday, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1061 4 : isjulian = true;
1062 :
1063 : /* Get the time zone from the end of the string */
1064 4 : dterr = DecodeTimezone(cp, tzp);
1065 4 : if (dterr)
1066 0 : return dterr;
1067 :
1068 4 : tmask = DTK_DATE_M | DTK_TIME_M | DTK_M(TZ);
1069 4 : ptype = 0;
1070 4 : break;
1071 : }
1072 :
1073 : /*
1074 : * Already have a date? Then this might be a time zone name
1075 : * with embedded punctuation (e.g. "America/New_York") or a
1076 : * run-together time with trailing time zone (e.g. hhmmss-zz).
1077 : * - thomas 2001-12-25
1078 : *
1079 : * We consider it a time zone if we already have month & day.
1080 : * This is to allow the form "mmm dd hhmmss tz year", which
1081 : * we've historically accepted.
1082 : */
1083 41003 : else if (ptype != 0 ||
1084 40987 : ((fmask & (DTK_M(MONTH) | DTK_M(DAY))) ==
1085 : (DTK_M(MONTH) | DTK_M(DAY))))
1086 : {
1087 : /* No time zone accepted? Then quit... */
1088 772 : if (tzp == NULL)
1089 0 : return DTERR_BAD_FORMAT;
1090 :
1091 1512 : if (isdigit((unsigned char) *field[i]) || ptype != 0)
1092 12 : {
1093 : char *cp;
1094 :
1095 : /*
1096 : * Allow a preceding "t" field, but no other units.
1097 : */
1098 20 : if (ptype != 0)
1099 : {
1100 : /* Sanity check; should not fail this test */
1101 16 : if (ptype != DTK_TIME)
1102 0 : return DTERR_BAD_FORMAT;
1103 16 : ptype = 0;
1104 : }
1105 :
1106 : /*
1107 : * Starts with a digit but we already have a time
1108 : * field? Then we are in trouble with a date and time
1109 : * already...
1110 : */
1111 20 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1112 0 : return DTERR_BAD_FORMAT;
1113 :
1114 20 : if ((cp = strchr(field[i], '-')) == NULL)
1115 0 : return DTERR_BAD_FORMAT;
1116 :
1117 : /* Get the time zone from the end of the string */
1118 20 : dterr = DecodeTimezone(cp, tzp);
1119 20 : if (dterr)
1120 0 : return dterr;
1121 20 : *cp = '\0';
1122 :
1123 : /*
1124 : * Then read the rest of the field as a concatenated
1125 : * time
1126 : */
1127 20 : dterr = DecodeNumberField(strlen(field[i]), field[i],
1128 : fmask,
1129 : &tmask, tm,
1130 : fsec, &is2digits);
1131 20 : if (dterr < 0)
1132 8 : return dterr;
1133 :
1134 : /*
1135 : * modify tmask after returning from
1136 : * DecodeNumberField()
1137 : */
1138 12 : tmask |= DTK_M(TZ);
1139 : }
1140 : else
1141 : {
1142 752 : namedTz = pg_tzset(field[i]);
1143 752 : if (!namedTz)
1144 : {
1145 24 : extra->dtee_timezone = field[i];
1146 24 : return DTERR_BAD_TIMEZONE;
1147 : }
1148 : /* we'll apply the zone setting below */
1149 728 : tmask = DTK_M(TZ);
1150 : }
1151 : }
1152 : else
1153 : {
1154 40231 : dterr = DecodeDate(field[i], fmask,
1155 : &tmask, &is2digits, tm);
1156 40231 : if (dterr)
1157 24 : return dterr;
1158 : }
1159 40947 : break;
1160 :
1161 28963 : case DTK_TIME:
1162 :
1163 : /*
1164 : * This might be an ISO time following a "t" field.
1165 : */
1166 28963 : if (ptype != 0)
1167 : {
1168 : /* Sanity check; should not fail this test */
1169 8 : if (ptype != DTK_TIME)
1170 0 : return DTERR_BAD_FORMAT;
1171 8 : ptype = 0;
1172 : }
1173 28963 : dterr = DecodeTime(field[i], fmask, INTERVAL_FULL_RANGE,
1174 : &tmask, tm, fsec);
1175 28963 : if (dterr)
1176 0 : return dterr;
1177 :
1178 : /* check for time overflow */
1179 28963 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec,
1180 : *fsec))
1181 0 : return DTERR_FIELD_OVERFLOW;
1182 28963 : break;
1183 :
1184 18939 : case DTK_TZ:
1185 : {
1186 : int tz;
1187 :
1188 18939 : if (tzp == NULL)
1189 8 : return DTERR_BAD_FORMAT;
1190 :
1191 18939 : dterr = DecodeTimezone(field[i], &tz);
1192 18939 : if (dterr)
1193 8 : return dterr;
1194 18931 : *tzp = tz;
1195 18931 : tmask = DTK_M(TZ);
1196 : }
1197 18931 : break;
1198 :
1199 3733 : case DTK_NUMBER:
1200 :
1201 : /*
1202 : * Deal with cases where previous field labeled this one
1203 : */
1204 3733 : if (ptype != 0)
1205 : {
1206 : char *cp;
1207 : int value;
1208 :
1209 96 : errno = 0;
1210 96 : value = strtoint(field[i], &cp, 10);
1211 96 : if (errno == ERANGE)
1212 8 : return DTERR_FIELD_OVERFLOW;
1213 96 : if (*cp != '.' && *cp != '\0')
1214 0 : return DTERR_BAD_FORMAT;
1215 :
1216 96 : switch (ptype)
1217 : {
1218 64 : case DTK_JULIAN:
1219 : /* previous field was a label for "julian date" */
1220 64 : if (value < 0)
1221 0 : return DTERR_FIELD_OVERFLOW;
1222 64 : tmask = DTK_DATE_M;
1223 64 : j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1224 64 : isjulian = true;
1225 :
1226 : /* fractional Julian Day? */
1227 64 : if (*cp == '.')
1228 : {
1229 : double time;
1230 :
1231 8 : dterr = ParseFraction(cp, &time);
1232 8 : if (dterr)
1233 0 : return dterr;
1234 8 : time *= USECS_PER_DAY;
1235 8 : dt2time(time,
1236 : &tm->tm_hour, &tm->tm_min,
1237 : &tm->tm_sec, fsec);
1238 8 : tmask |= DTK_TIME_M;
1239 : }
1240 64 : break;
1241 :
1242 24 : case DTK_TIME:
1243 : /* previous field was "t" for ISO time */
1244 24 : dterr = DecodeNumberField(strlen(field[i]), field[i],
1245 : (fmask | DTK_DATE_M),
1246 : &tmask, tm,
1247 : fsec, &is2digits);
1248 24 : if (dterr < 0)
1249 0 : return dterr;
1250 24 : if (tmask != DTK_TIME_M)
1251 0 : return DTERR_BAD_FORMAT;
1252 24 : break;
1253 :
1254 8 : default:
1255 8 : return DTERR_BAD_FORMAT;
1256 : break;
1257 : }
1258 :
1259 88 : ptype = 0;
1260 88 : *dtype = DTK_DATE;
1261 : }
1262 : else
1263 : {
1264 : char *cp;
1265 : int flen;
1266 :
1267 3637 : flen = strlen(field[i]);
1268 3637 : cp = strchr(field[i], '.');
1269 :
1270 : /* Embedded decimal and no date yet? */
1271 3637 : if (cp != NULL && !(fmask & DTK_DATE_M))
1272 : {
1273 20 : dterr = DecodeDate(field[i], fmask,
1274 : &tmask, &is2digits, tm);
1275 20 : if (dterr)
1276 0 : return dterr;
1277 : }
1278 : /* embedded decimal and several digits before? */
1279 3617 : else if (cp != NULL && flen - strlen(cp) > 2)
1280 : {
1281 : /*
1282 : * Interpret as a concatenated date or time Set the
1283 : * type field to allow decoding other fields later.
1284 : * Example: 20011223 or 040506
1285 : */
1286 8 : dterr = DecodeNumberField(flen, field[i], fmask,
1287 : &tmask, tm,
1288 : fsec, &is2digits);
1289 8 : if (dterr < 0)
1290 0 : return dterr;
1291 : }
1292 :
1293 : /*
1294 : * Is this a YMD or HMS specification, or a year number?
1295 : * YMD and HMS are required to be six digits or more, so
1296 : * if it is 5 digits, it is a year. If it is six or more
1297 : * digits, we assume it is YMD or HMS unless no date and
1298 : * no time values have been specified. This forces 6+
1299 : * digit years to be at the end of the string, or to use
1300 : * the ISO date specification.
1301 : */
1302 3609 : else if (flen >= 6 && (!(fmask & DTK_DATE_M) ||
1303 64 : !(fmask & DTK_TIME_M)))
1304 : {
1305 224 : dterr = DecodeNumberField(flen, field[i], fmask,
1306 : &tmask, tm,
1307 : fsec, &is2digits);
1308 224 : if (dterr < 0)
1309 0 : return dterr;
1310 : }
1311 : /* otherwise it is a single date/time field... */
1312 : else
1313 : {
1314 3385 : dterr = DecodeNumber(flen, field[i],
1315 : haveTextMonth, fmask,
1316 : &tmask, tm,
1317 : fsec, &is2digits);
1318 3385 : if (dterr)
1319 8 : return dterr;
1320 : }
1321 : }
1322 3717 : break;
1323 :
1324 5606 : case DTK_STRING:
1325 : case DTK_SPECIAL:
1326 : /* timezone abbrevs take precedence over built-in tokens */
1327 5606 : dterr = DecodeTimezoneAbbrev(i, field[i],
1328 : &type, &val, &valtz, extra);
1329 5606 : if (dterr)
1330 0 : return dterr;
1331 5606 : if (type == UNKNOWN_FIELD)
1332 4218 : type = DecodeSpecial(i, field[i], &val);
1333 5606 : if (type == IGNORE_DTF)
1334 0 : continue;
1335 :
1336 5606 : tmask = DTK_M(type);
1337 5606 : switch (type)
1338 : {
1339 1199 : case RESERV:
1340 1199 : switch (val)
1341 : {
1342 76 : case DTK_NOW:
1343 76 : tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1344 76 : *dtype = DTK_DATE;
1345 76 : GetCurrentTimeUsec(tm, fsec, tzp);
1346 76 : break;
1347 :
1348 68 : case DTK_YESTERDAY:
1349 68 : tmask = DTK_DATE_M;
1350 68 : *dtype = DTK_DATE;
1351 68 : GetCurrentDateTime(&cur_tm);
1352 68 : j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) - 1,
1353 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1354 68 : break;
1355 :
1356 96 : case DTK_TODAY:
1357 96 : tmask = DTK_DATE_M;
1358 96 : *dtype = DTK_DATE;
1359 96 : GetCurrentDateTime(&cur_tm);
1360 96 : tm->tm_year = cur_tm.tm_year;
1361 96 : tm->tm_mon = cur_tm.tm_mon;
1362 96 : tm->tm_mday = cur_tm.tm_mday;
1363 96 : break;
1364 :
1365 100 : case DTK_TOMORROW:
1366 100 : tmask = DTK_DATE_M;
1367 100 : *dtype = DTK_DATE;
1368 100 : GetCurrentDateTime(&cur_tm);
1369 100 : j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) + 1,
1370 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1371 100 : break;
1372 :
1373 4 : case DTK_ZULU:
1374 4 : tmask = (DTK_TIME_M | DTK_M(TZ));
1375 4 : *dtype = DTK_DATE;
1376 4 : tm->tm_hour = 0;
1377 4 : tm->tm_min = 0;
1378 4 : tm->tm_sec = 0;
1379 4 : if (tzp != NULL)
1380 4 : *tzp = 0;
1381 4 : break;
1382 :
1383 855 : case DTK_EPOCH:
1384 : case DTK_LATE:
1385 : case DTK_EARLY:
1386 855 : tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1387 855 : *dtype = val;
1388 : /* caller ignores tm for these dtype codes */
1389 855 : break;
1390 :
1391 0 : default:
1392 0 : elog(ERROR, "unrecognized RESERV datetime token: %d",
1393 : val);
1394 : }
1395 :
1396 1199 : break;
1397 :
1398 1533 : case MONTH:
1399 :
1400 : /*
1401 : * already have a (numeric) month? then see if we can
1402 : * substitute...
1403 : */
1404 1533 : if ((fmask & DTK_M(MONTH)) && !haveTextMonth &&
1405 55 : !(fmask & DTK_M(DAY)) && tm->tm_mon >= 1 &&
1406 47 : tm->tm_mon <= 31)
1407 : {
1408 43 : tm->tm_mday = tm->tm_mon;
1409 43 : tmask = DTK_M(DAY);
1410 : }
1411 1533 : haveTextMonth = true;
1412 1533 : tm->tm_mon = val;
1413 1533 : break;
1414 :
1415 4 : case DTZMOD:
1416 :
1417 : /*
1418 : * daylight savings time modifier (solves "MET DST"
1419 : * syntax)
1420 : */
1421 4 : tmask |= DTK_M(DTZ);
1422 4 : tm->tm_isdst = 1;
1423 4 : if (tzp == NULL)
1424 0 : return DTERR_BAD_FORMAT;
1425 4 : *tzp -= val;
1426 4 : break;
1427 :
1428 940 : case DTZ:
1429 :
1430 : /*
1431 : * set mask for TZ here _or_ check for DTZ later when
1432 : * getting default timezone
1433 : */
1434 940 : tmask |= DTK_M(TZ);
1435 940 : tm->tm_isdst = 1;
1436 940 : if (tzp == NULL)
1437 0 : return DTERR_BAD_FORMAT;
1438 940 : *tzp = -val;
1439 940 : break;
1440 :
1441 392 : case TZ:
1442 392 : tm->tm_isdst = 0;
1443 392 : if (tzp == NULL)
1444 0 : return DTERR_BAD_FORMAT;
1445 392 : *tzp = -val;
1446 392 : break;
1447 :
1448 56 : case DYNTZ:
1449 56 : tmask |= DTK_M(TZ);
1450 56 : if (tzp == NULL)
1451 0 : return DTERR_BAD_FORMAT;
1452 : /* we'll determine the actual offset later */
1453 56 : abbrevTz = valtz;
1454 56 : abbrev = field[i];
1455 56 : break;
1456 :
1457 24 : case AMPM:
1458 24 : mer = val;
1459 24 : break;
1460 :
1461 224 : case ADBC:
1462 224 : bc = (val == BC);
1463 224 : break;
1464 :
1465 1058 : case DOW:
1466 1058 : tm->tm_wday = val;
1467 1058 : break;
1468 :
1469 92 : case UNITS:
1470 92 : tmask = 0;
1471 : /* reject consecutive unhandled units */
1472 92 : if (ptype != 0)
1473 8 : return DTERR_BAD_FORMAT;
1474 84 : ptype = val;
1475 84 : break;
1476 :
1477 48 : case ISOTIME:
1478 :
1479 : /*
1480 : * This is a filler field "t" indicating that the next
1481 : * field is time. Try to verify that this is sensible.
1482 : */
1483 48 : tmask = 0;
1484 :
1485 : /* No preceding date? Then quit... */
1486 48 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1487 0 : return DTERR_BAD_FORMAT;
1488 :
1489 : /* reject consecutive unhandled units */
1490 48 : if (ptype != 0)
1491 0 : return DTERR_BAD_FORMAT;
1492 48 : ptype = val;
1493 48 : break;
1494 :
1495 36 : case UNKNOWN_FIELD:
1496 :
1497 : /*
1498 : * Before giving up and declaring error, check to see
1499 : * if it is an all-alpha timezone name.
1500 : */
1501 36 : namedTz = pg_tzset(field[i]);
1502 36 : if (!namedTz)
1503 36 : return DTERR_BAD_FORMAT;
1504 : /* we'll apply the zone setting below */
1505 0 : tmask = DTK_M(TZ);
1506 0 : break;
1507 :
1508 0 : default:
1509 0 : return DTERR_BAD_FORMAT;
1510 : }
1511 5562 : break;
1512 :
1513 0 : default:
1514 0 : return DTERR_BAD_FORMAT;
1515 : }
1516 :
1517 98124 : if (tmask & fmask)
1518 96 : return DTERR_BAD_FORMAT;
1519 98028 : fmask |= tmask;
1520 : } /* end loop over fields */
1521 :
1522 : /* reject if prefix type appeared and was never handled */
1523 43084 : if (ptype != 0)
1524 0 : return DTERR_BAD_FORMAT;
1525 :
1526 : /* do additional checking for normal date specs (but not "infinity" etc) */
1527 43084 : if (*dtype == DTK_DATE)
1528 : {
1529 : /* do final checking/adjustment of Y/M/D fields */
1530 42325 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
1531 42325 : if (dterr)
1532 132 : return dterr;
1533 :
1534 : /* handle AM/PM */
1535 42193 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
1536 0 : return DTERR_FIELD_OVERFLOW;
1537 42193 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
1538 0 : tm->tm_hour = 0;
1539 42193 : else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
1540 24 : tm->tm_hour += HOURS_PER_DAY / 2;
1541 :
1542 : /* check for incomplete input */
1543 42193 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1544 : {
1545 4 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1546 0 : return 1;
1547 4 : return DTERR_BAD_FORMAT;
1548 : }
1549 :
1550 : /*
1551 : * If we had a full timezone spec, compute the offset (we could not do
1552 : * it before, because we need the date to resolve DST status).
1553 : */
1554 42189 : if (namedTz != NULL)
1555 : {
1556 : /* daylight savings time modifier disallowed with full TZ */
1557 728 : if (fmask & DTK_M(DTZMOD))
1558 0 : return DTERR_BAD_FORMAT;
1559 :
1560 728 : *tzp = DetermineTimeZoneOffset(tm, namedTz);
1561 : }
1562 :
1563 : /*
1564 : * Likewise, if we had a dynamic timezone abbreviation, resolve it
1565 : * now.
1566 : */
1567 42189 : if (abbrevTz != NULL)
1568 : {
1569 : /* daylight savings time modifier disallowed with dynamic TZ */
1570 56 : if (fmask & DTK_M(DTZMOD))
1571 0 : return DTERR_BAD_FORMAT;
1572 :
1573 56 : *tzp = DetermineTimeZoneAbbrevOffset(tm, abbrev, abbrevTz);
1574 : }
1575 :
1576 : /* timezone not specified? then use session timezone */
1577 42189 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
1578 : {
1579 : /*
1580 : * daylight savings time modifier but no standard timezone? then
1581 : * error
1582 : */
1583 21062 : if (fmask & DTK_M(DTZMOD))
1584 0 : return DTERR_BAD_FORMAT;
1585 :
1586 21062 : *tzp = DetermineTimeZoneOffset(tm, session_timezone);
1587 : }
1588 : }
1589 :
1590 42948 : return 0;
1591 : }
1592 :
1593 :
1594 : /*
1595 : * DetermineTimeZoneOffset()
1596 : *
1597 : * Given a struct pg_tm in which tm_year, tm_mon, tm_mday, tm_hour, tm_min,
1598 : * and tm_sec fields are set, and a zic-style time zone definition, determine
1599 : * the applicable GMT offset and daylight-savings status at that time.
1600 : * Set the struct pg_tm's tm_isdst field accordingly, and return the GMT
1601 : * offset as the function result.
1602 : *
1603 : * Note: if the date is out of the range we can deal with, we return zero
1604 : * as the GMT offset and set tm_isdst = 0. We don't throw an error here,
1605 : * though probably some higher-level code will.
1606 : */
1607 : int
1608 76198 : DetermineTimeZoneOffset(struct pg_tm *tm, pg_tz *tzp)
1609 : {
1610 : pg_time_t t;
1611 :
1612 76198 : return DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1613 : }
1614 :
1615 :
1616 : /*
1617 : * DetermineTimeZoneOffsetInternal()
1618 : *
1619 : * As above, but also return the actual UTC time imputed to the date/time
1620 : * into *tp.
1621 : *
1622 : * In event of an out-of-range date, we punt by returning zero into *tp.
1623 : * This is okay for the immediate callers but is a good reason for not
1624 : * exposing this worker function globally.
1625 : *
1626 : * Note: it might seem that we should use mktime() for this, but bitter
1627 : * experience teaches otherwise. This code is much faster than most versions
1628 : * of mktime(), anyway.
1629 : */
1630 : static int
1631 76336 : DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp, pg_time_t *tp)
1632 : {
1633 : int date,
1634 : sec;
1635 : pg_time_t day,
1636 : mytime,
1637 : prevtime,
1638 : boundary,
1639 : beforetime,
1640 : aftertime;
1641 : long int before_gmtoff,
1642 : after_gmtoff;
1643 : int before_isdst,
1644 : after_isdst;
1645 : int res;
1646 :
1647 : /*
1648 : * First, generate the pg_time_t value corresponding to the given
1649 : * y/m/d/h/m/s taken as GMT time. If this overflows, punt and decide the
1650 : * timezone is GMT. (For a valid Julian date, integer overflow should be
1651 : * impossible with 64-bit pg_time_t, but let's check for safety.)
1652 : */
1653 76336 : if (!IS_VALID_JULIAN(tm->tm_year, tm->tm_mon, tm->tm_mday))
1654 16 : goto overflow;
1655 76320 : date = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday) - UNIX_EPOCH_JDATE;
1656 :
1657 76320 : day = ((pg_time_t) date) * SECS_PER_DAY;
1658 76320 : if (day / SECS_PER_DAY != date)
1659 0 : goto overflow;
1660 76320 : sec = tm->tm_sec + (tm->tm_min + tm->tm_hour * MINS_PER_HOUR) * SECS_PER_MINUTE;
1661 76320 : mytime = day + sec;
1662 : /* since sec >= 0, overflow could only be from +day to -mytime */
1663 76320 : if (mytime < 0 && day > 0)
1664 0 : goto overflow;
1665 :
1666 : /*
1667 : * Find the DST time boundary just before or following the target time. We
1668 : * assume that all zones have GMT offsets less than 24 hours, and that DST
1669 : * boundaries can't be closer together than 48 hours, so backing up 24
1670 : * hours and finding the "next" boundary will work.
1671 : */
1672 76320 : prevtime = mytime - SECS_PER_DAY;
1673 76320 : if (mytime < 0 && prevtime > 0)
1674 0 : goto overflow;
1675 :
1676 76320 : res = pg_next_dst_boundary(&prevtime,
1677 : &before_gmtoff, &before_isdst,
1678 : &boundary,
1679 : &after_gmtoff, &after_isdst,
1680 : tzp);
1681 76320 : if (res < 0)
1682 0 : goto overflow; /* failure? */
1683 :
1684 76320 : if (res == 0)
1685 : {
1686 : /* Non-DST zone, life is simple */
1687 3351 : tm->tm_isdst = before_isdst;
1688 3351 : *tp = mytime - before_gmtoff;
1689 3351 : return -(int) before_gmtoff;
1690 : }
1691 :
1692 : /*
1693 : * Form the candidate pg_time_t values with local-time adjustment
1694 : */
1695 72969 : beforetime = mytime - before_gmtoff;
1696 72969 : if ((before_gmtoff > 0 &&
1697 8 : mytime < 0 && beforetime > 0) ||
1698 72969 : (before_gmtoff <= 0 &&
1699 64396 : mytime > 0 && beforetime < 0))
1700 0 : goto overflow;
1701 72969 : aftertime = mytime - after_gmtoff;
1702 72969 : if ((after_gmtoff > 0 &&
1703 8 : mytime < 0 && aftertime > 0) ||
1704 72969 : (after_gmtoff <= 0 &&
1705 64396 : mytime > 0 && aftertime < 0))
1706 0 : goto overflow;
1707 :
1708 : /*
1709 : * If both before or both after the boundary time, we know what to do. The
1710 : * boundary time itself is considered to be after the transition, which
1711 : * means we can accept aftertime == boundary in the second case.
1712 : */
1713 72969 : if (beforetime < boundary && aftertime < boundary)
1714 : {
1715 72368 : tm->tm_isdst = before_isdst;
1716 72368 : *tp = beforetime;
1717 72368 : return -(int) before_gmtoff;
1718 : }
1719 601 : if (beforetime > boundary && aftertime >= boundary)
1720 : {
1721 489 : tm->tm_isdst = after_isdst;
1722 489 : *tp = aftertime;
1723 489 : return -(int) after_gmtoff;
1724 : }
1725 :
1726 : /*
1727 : * It's an invalid or ambiguous time due to timezone transition. In a
1728 : * spring-forward transition, prefer the "before" interpretation; in a
1729 : * fall-back transition, prefer "after". (We used to define and implement
1730 : * this test as "prefer the standard-time interpretation", but that rule
1731 : * does not help to resolve the behavior when both times are reported as
1732 : * standard time; which does happen, eg Europe/Moscow in Oct 2014. Also,
1733 : * in some zones such as Europe/Dublin, there is widespread confusion
1734 : * about which time offset is "standard" time, so it's fortunate that our
1735 : * behavior doesn't depend on that.)
1736 : */
1737 112 : if (beforetime > aftertime)
1738 : {
1739 54 : tm->tm_isdst = before_isdst;
1740 54 : *tp = beforetime;
1741 54 : return -(int) before_gmtoff;
1742 : }
1743 58 : tm->tm_isdst = after_isdst;
1744 58 : *tp = aftertime;
1745 58 : return -(int) after_gmtoff;
1746 :
1747 16 : overflow:
1748 : /* Given date is out of range, so assume UTC */
1749 16 : tm->tm_isdst = 0;
1750 16 : *tp = 0;
1751 16 : return 0;
1752 : }
1753 :
1754 :
1755 : /*
1756 : * DetermineTimeZoneAbbrevOffset()
1757 : *
1758 : * Determine the GMT offset and DST flag to be attributed to a dynamic
1759 : * time zone abbreviation, that is one whose meaning has changed over time.
1760 : * *tm contains the local time at which the meaning should be determined,
1761 : * and tm->tm_isdst receives the DST flag.
1762 : *
1763 : * This differs from the behavior of DetermineTimeZoneOffset() in that a
1764 : * standard-time or daylight-time abbreviation forces use of the corresponding
1765 : * GMT offset even when the zone was then in DS or standard time respectively.
1766 : * (However, that happens only if we can match the given abbreviation to some
1767 : * abbreviation that appears in the IANA timezone data. Otherwise, we fall
1768 : * back to doing DetermineTimeZoneOffset().)
1769 : */
1770 : int
1771 138 : DetermineTimeZoneAbbrevOffset(struct pg_tm *tm, const char *abbr, pg_tz *tzp)
1772 : {
1773 : pg_time_t t;
1774 : int zone_offset;
1775 : int abbr_offset;
1776 : int abbr_isdst;
1777 :
1778 : /*
1779 : * Compute the UTC time we want to probe at. (In event of overflow, we'll
1780 : * probe at the epoch, which is a bit random but probably doesn't matter.)
1781 : */
1782 138 : zone_offset = DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1783 :
1784 : /*
1785 : * Try to match the abbreviation to something in the zone definition.
1786 : */
1787 138 : if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1788 : &abbr_offset, &abbr_isdst))
1789 : {
1790 : /* Success, so use the abbrev-specific answers. */
1791 138 : tm->tm_isdst = abbr_isdst;
1792 138 : return abbr_offset;
1793 : }
1794 :
1795 : /*
1796 : * No match, so use the answers we already got from
1797 : * DetermineTimeZoneOffsetInternal.
1798 : */
1799 0 : return zone_offset;
1800 : }
1801 :
1802 :
1803 : /*
1804 : * DetermineTimeZoneAbbrevOffsetTS()
1805 : *
1806 : * As above but the probe time is specified as a TimestampTz (hence, UTC time),
1807 : * and DST status is returned into *isdst rather than into tm->tm_isdst.
1808 : */
1809 : int
1810 856 : DetermineTimeZoneAbbrevOffsetTS(TimestampTz ts, const char *abbr,
1811 : pg_tz *tzp, int *isdst)
1812 : {
1813 856 : pg_time_t t = timestamptz_to_time_t(ts);
1814 : int zone_offset;
1815 : int abbr_offset;
1816 : int tz;
1817 : struct pg_tm tm;
1818 : fsec_t fsec;
1819 :
1820 : /*
1821 : * If the abbrev matches anything in the zone data, this is pretty easy.
1822 : */
1823 856 : if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1824 : &abbr_offset, isdst))
1825 76 : return abbr_offset;
1826 :
1827 : /*
1828 : * Else, break down the timestamp so we can use DetermineTimeZoneOffset.
1829 : */
1830 780 : if (timestamp2tm(ts, &tz, &tm, &fsec, NULL, tzp) != 0)
1831 0 : ereport(ERROR,
1832 : (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
1833 : errmsg("timestamp out of range")));
1834 :
1835 780 : zone_offset = DetermineTimeZoneOffset(&tm, tzp);
1836 780 : *isdst = tm.tm_isdst;
1837 780 : return zone_offset;
1838 : }
1839 :
1840 :
1841 : /*
1842 : * DetermineTimeZoneAbbrevOffsetInternal()
1843 : *
1844 : * Workhorse for above two functions: work from a pg_time_t probe instant.
1845 : * On success, return GMT offset and DST status into *offset and *isdst.
1846 : */
1847 : static bool
1848 994 : DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t, const char *abbr, pg_tz *tzp,
1849 : int *offset, int *isdst)
1850 : {
1851 : char upabbr[TZ_STRLEN_MAX + 1];
1852 : unsigned char *p;
1853 : long int gmtoff;
1854 :
1855 : /* We need to force the abbrev to upper case */
1856 994 : strlcpy(upabbr, abbr, sizeof(upabbr));
1857 4660 : for (p = (unsigned char *) upabbr; *p; p++)
1858 3666 : *p = pg_toupper(*p);
1859 :
1860 : /* Look up the abbrev's meaning at this time in this zone */
1861 994 : if (pg_interpret_timezone_abbrev(upabbr,
1862 : &t,
1863 : &gmtoff,
1864 : isdst,
1865 : tzp))
1866 : {
1867 : /* Change sign to agree with DetermineTimeZoneOffset() */
1868 214 : *offset = (int) -gmtoff;
1869 214 : return true;
1870 : }
1871 780 : return false;
1872 : }
1873 :
1874 :
1875 : /*
1876 : * TimeZoneAbbrevIsKnown()
1877 : *
1878 : * Detect whether the given string is a time zone abbreviation that's known
1879 : * in the specified TZDB timezone, and if so whether it's fixed or varying
1880 : * meaning. The match is not case-sensitive.
1881 : */
1882 : static bool
1883 4961 : TimeZoneAbbrevIsKnown(const char *abbr, pg_tz *tzp,
1884 : bool *isfixed, int *offset, int *isdst)
1885 : {
1886 : char upabbr[TZ_STRLEN_MAX + 1];
1887 : unsigned char *p;
1888 : long int gmtoff;
1889 :
1890 : /* We need to force the abbrev to upper case */
1891 4961 : strlcpy(upabbr, abbr, sizeof(upabbr));
1892 27539 : for (p = (unsigned char *) upabbr; *p; p++)
1893 22578 : *p = pg_toupper(*p);
1894 :
1895 : /* Look up the abbrev's meaning in this zone */
1896 4961 : if (pg_timezone_abbrev_is_known(upabbr,
1897 : isfixed,
1898 : &gmtoff,
1899 : isdst,
1900 : tzp))
1901 : {
1902 : /* Change sign to agree with DetermineTimeZoneOffset() */
1903 153 : *offset = (int) -gmtoff;
1904 153 : return true;
1905 : }
1906 4808 : return false;
1907 : }
1908 :
1909 :
1910 : /*
1911 : * DecodeTimeOnly()
1912 : * Interpret parsed string as time fields only.
1913 : * Returns 0 if successful, DTERR code if bogus input detected.
1914 : *
1915 : * Inputs are field[] and ftype[] arrays, of length nf.
1916 : * Other arguments are outputs.
1917 : *
1918 : * Note that support for time zone is here for
1919 : * SQL TIME WITH TIME ZONE, but it reveals
1920 : * bogosity with SQL date/time standards, since
1921 : * we must infer a time zone from current time.
1922 : * - thomas 2000-03-10
1923 : * Allow specifying date to get a better time zone,
1924 : * if time zones are allowed. - thomas 2001-12-26
1925 : */
1926 : int
1927 2788 : DecodeTimeOnly(char **field, int *ftype, int nf,
1928 : int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp,
1929 : DateTimeErrorExtra *extra)
1930 : {
1931 2788 : int fmask = 0,
1932 : tmask,
1933 : type;
1934 2788 : int ptype = 0; /* "prefix type" for ISO and Julian formats */
1935 : int i;
1936 : int val;
1937 : int dterr;
1938 2788 : bool isjulian = false;
1939 2788 : bool is2digits = false;
1940 2788 : bool bc = false;
1941 2788 : int mer = HR24;
1942 2788 : pg_tz *namedTz = NULL;
1943 2788 : pg_tz *abbrevTz = NULL;
1944 2788 : char *abbrev = NULL;
1945 : pg_tz *valtz;
1946 :
1947 2788 : *dtype = DTK_TIME;
1948 2788 : tm->tm_hour = 0;
1949 2788 : tm->tm_min = 0;
1950 2788 : tm->tm_sec = 0;
1951 2788 : *fsec = 0;
1952 : /* don't know daylight savings time status apriori */
1953 2788 : tm->tm_isdst = -1;
1954 :
1955 2788 : if (tzp != NULL)
1956 2788 : *tzp = 0;
1957 :
1958 7154 : for (i = 0; i < nf; i++)
1959 : {
1960 4382 : switch (ftype[i])
1961 : {
1962 749 : case DTK_DATE:
1963 :
1964 : /*
1965 : * Time zone not allowed? Then should not accept dates or time
1966 : * zones no matter what else!
1967 : */
1968 749 : if (tzp == NULL)
1969 0 : return DTERR_BAD_FORMAT;
1970 :
1971 : /* Under limited circumstances, we will accept a date... */
1972 749 : if (i == 0 && nf >= 2 &&
1973 140 : (ftype[nf - 1] == DTK_DATE || ftype[1] == DTK_TIME))
1974 : {
1975 140 : dterr = DecodeDate(field[i], fmask,
1976 : &tmask, &is2digits, tm);
1977 140 : if (dterr)
1978 0 : return dterr;
1979 : }
1980 : /* otherwise, this is a time and/or time zone */
1981 : else
1982 : {
1983 609 : if (isdigit((unsigned char) *field[i]))
1984 : {
1985 : char *cp;
1986 :
1987 : /*
1988 : * Starts with a digit but we already have a time
1989 : * field? Then we are in trouble with time already...
1990 : */
1991 0 : if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1992 0 : return DTERR_BAD_FORMAT;
1993 :
1994 : /*
1995 : * Should not get here and fail. Sanity check only...
1996 : */
1997 0 : if ((cp = strchr(field[i], '-')) == NULL)
1998 0 : return DTERR_BAD_FORMAT;
1999 :
2000 : /* Get the time zone from the end of the string */
2001 0 : dterr = DecodeTimezone(cp, tzp);
2002 0 : if (dterr)
2003 0 : return dterr;
2004 0 : *cp = '\0';
2005 :
2006 : /*
2007 : * Then read the rest of the field as a concatenated
2008 : * time
2009 : */
2010 0 : dterr = DecodeNumberField(strlen(field[i]), field[i],
2011 : (fmask | DTK_DATE_M),
2012 : &tmask, tm,
2013 : fsec, &is2digits);
2014 0 : if (dterr < 0)
2015 0 : return dterr;
2016 0 : ftype[i] = dterr;
2017 :
2018 0 : tmask |= DTK_M(TZ);
2019 : }
2020 : else
2021 : {
2022 609 : namedTz = pg_tzset(field[i]);
2023 609 : if (!namedTz)
2024 : {
2025 0 : extra->dtee_timezone = field[i];
2026 0 : return DTERR_BAD_TIMEZONE;
2027 : }
2028 : /* we'll apply the zone setting below */
2029 609 : ftype[i] = DTK_TZ;
2030 609 : tmask = DTK_M(TZ);
2031 : }
2032 : }
2033 749 : break;
2034 :
2035 2672 : case DTK_TIME:
2036 :
2037 : /*
2038 : * This might be an ISO time following a "t" field.
2039 : */
2040 2672 : if (ptype != 0)
2041 : {
2042 24 : if (ptype != DTK_TIME)
2043 0 : return DTERR_BAD_FORMAT;
2044 24 : ptype = 0;
2045 : }
2046 :
2047 2672 : dterr = DecodeTime(field[i], (fmask | DTK_DATE_M),
2048 : INTERVAL_FULL_RANGE,
2049 : &tmask, tm, fsec);
2050 2672 : if (dterr)
2051 0 : return dterr;
2052 2672 : break;
2053 :
2054 550 : case DTK_TZ:
2055 : {
2056 : int tz;
2057 :
2058 550 : if (tzp == NULL)
2059 0 : return DTERR_BAD_FORMAT;
2060 :
2061 550 : dterr = DecodeTimezone(field[i], &tz);
2062 550 : if (dterr)
2063 0 : return dterr;
2064 550 : *tzp = tz;
2065 550 : tmask = DTK_M(TZ);
2066 : }
2067 550 : break;
2068 :
2069 120 : case DTK_NUMBER:
2070 :
2071 : /*
2072 : * Deal with cases where previous field labeled this one
2073 : */
2074 120 : if (ptype != 0)
2075 : {
2076 : char *cp;
2077 : int value;
2078 :
2079 80 : errno = 0;
2080 80 : value = strtoint(field[i], &cp, 10);
2081 80 : if (errno == ERANGE)
2082 16 : return DTERR_FIELD_OVERFLOW;
2083 80 : if (*cp != '.' && *cp != '\0')
2084 0 : return DTERR_BAD_FORMAT;
2085 :
2086 80 : switch (ptype)
2087 : {
2088 4 : case DTK_JULIAN:
2089 : /* previous field was a label for "julian date" */
2090 4 : if (tzp == NULL)
2091 0 : return DTERR_BAD_FORMAT;
2092 4 : if (value < 0)
2093 0 : return DTERR_FIELD_OVERFLOW;
2094 4 : tmask = DTK_DATE_M;
2095 4 : j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2096 4 : isjulian = true;
2097 :
2098 4 : if (*cp == '.')
2099 : {
2100 : double time;
2101 :
2102 0 : dterr = ParseFraction(cp, &time);
2103 0 : if (dterr)
2104 0 : return dterr;
2105 0 : time *= USECS_PER_DAY;
2106 0 : dt2time(time,
2107 : &tm->tm_hour, &tm->tm_min,
2108 : &tm->tm_sec, fsec);
2109 0 : tmask |= DTK_TIME_M;
2110 : }
2111 4 : break;
2112 :
2113 68 : case DTK_TIME:
2114 : /* previous field was "t" for ISO time */
2115 68 : dterr = DecodeNumberField(strlen(field[i]), field[i],
2116 : (fmask | DTK_DATE_M),
2117 : &tmask, tm,
2118 : fsec, &is2digits);
2119 68 : if (dterr < 0)
2120 8 : return dterr;
2121 60 : ftype[i] = dterr;
2122 :
2123 60 : if (tmask != DTK_TIME_M)
2124 0 : return DTERR_BAD_FORMAT;
2125 60 : break;
2126 :
2127 8 : default:
2128 8 : return DTERR_BAD_FORMAT;
2129 : break;
2130 : }
2131 :
2132 64 : ptype = 0;
2133 64 : *dtype = DTK_DATE;
2134 : }
2135 : else
2136 : {
2137 : char *cp;
2138 : int flen;
2139 :
2140 40 : flen = strlen(field[i]);
2141 40 : cp = strchr(field[i], '.');
2142 :
2143 : /* Embedded decimal? */
2144 40 : if (cp != NULL)
2145 : {
2146 : /*
2147 : * Under limited circumstances, we will accept a
2148 : * date...
2149 : */
2150 24 : if (i == 0 && nf >= 2 && ftype[nf - 1] == DTK_DATE)
2151 : {
2152 0 : dterr = DecodeDate(field[i], fmask,
2153 : &tmask, &is2digits, tm);
2154 0 : if (dterr)
2155 0 : return dterr;
2156 : }
2157 : /* embedded decimal and several digits before? */
2158 24 : else if (flen - strlen(cp) > 2)
2159 : {
2160 : /*
2161 : * Interpret as a concatenated date or time Set
2162 : * the type field to allow decoding other fields
2163 : * later. Example: 20011223 or 040506
2164 : */
2165 24 : dterr = DecodeNumberField(flen, field[i],
2166 : (fmask | DTK_DATE_M),
2167 : &tmask, tm,
2168 : fsec, &is2digits);
2169 24 : if (dterr < 0)
2170 0 : return dterr;
2171 24 : ftype[i] = dterr;
2172 : }
2173 : else
2174 0 : return DTERR_BAD_FORMAT;
2175 : }
2176 16 : else if (flen > 4)
2177 : {
2178 8 : dterr = DecodeNumberField(flen, field[i],
2179 : (fmask | DTK_DATE_M),
2180 : &tmask, tm,
2181 : fsec, &is2digits);
2182 8 : if (dterr < 0)
2183 0 : return dterr;
2184 8 : ftype[i] = dterr;
2185 : }
2186 : /* otherwise it is a single date/time field... */
2187 : else
2188 : {
2189 8 : dterr = DecodeNumber(flen, field[i],
2190 : false,
2191 : (fmask | DTK_DATE_M),
2192 : &tmask, tm,
2193 : fsec, &is2digits);
2194 8 : if (dterr)
2195 0 : return dterr;
2196 : }
2197 : }
2198 104 : break;
2199 :
2200 291 : case DTK_STRING:
2201 : case DTK_SPECIAL:
2202 : /* timezone abbrevs take precedence over built-in tokens */
2203 291 : dterr = DecodeTimezoneAbbrev(i, field[i],
2204 : &type, &val, &valtz, extra);
2205 291 : if (dterr)
2206 0 : return dterr;
2207 291 : if (type == UNKNOWN_FIELD)
2208 120 : type = DecodeSpecial(i, field[i], &val);
2209 291 : if (type == IGNORE_DTF)
2210 0 : continue;
2211 :
2212 291 : tmask = DTK_M(type);
2213 291 : switch (type)
2214 : {
2215 8 : case RESERV:
2216 8 : switch (val)
2217 : {
2218 8 : case DTK_NOW:
2219 8 : tmask = DTK_TIME_M;
2220 8 : *dtype = DTK_TIME;
2221 8 : GetCurrentTimeUsec(tm, fsec, NULL);
2222 8 : break;
2223 :
2224 0 : case DTK_ZULU:
2225 0 : tmask = (DTK_TIME_M | DTK_M(TZ));
2226 0 : *dtype = DTK_TIME;
2227 0 : tm->tm_hour = 0;
2228 0 : tm->tm_min = 0;
2229 0 : tm->tm_sec = 0;
2230 0 : tm->tm_isdst = 0;
2231 0 : break;
2232 :
2233 0 : default:
2234 0 : return DTERR_BAD_FORMAT;
2235 : }
2236 :
2237 8 : break;
2238 :
2239 0 : case DTZMOD:
2240 :
2241 : /*
2242 : * daylight savings time modifier (solves "MET DST"
2243 : * syntax)
2244 : */
2245 0 : tmask |= DTK_M(DTZ);
2246 0 : tm->tm_isdst = 1;
2247 0 : if (tzp == NULL)
2248 0 : return DTERR_BAD_FORMAT;
2249 0 : *tzp -= val;
2250 0 : break;
2251 :
2252 132 : case DTZ:
2253 :
2254 : /*
2255 : * set mask for TZ here _or_ check for DTZ later when
2256 : * getting default timezone
2257 : */
2258 132 : tmask |= DTK_M(TZ);
2259 132 : tm->tm_isdst = 1;
2260 132 : if (tzp == NULL)
2261 0 : return DTERR_BAD_FORMAT;
2262 132 : *tzp = -val;
2263 132 : ftype[i] = DTK_TZ;
2264 132 : break;
2265 :
2266 35 : case TZ:
2267 35 : tm->tm_isdst = 0;
2268 35 : if (tzp == NULL)
2269 0 : return DTERR_BAD_FORMAT;
2270 35 : *tzp = -val;
2271 35 : ftype[i] = DTK_TZ;
2272 35 : break;
2273 :
2274 4 : case DYNTZ:
2275 4 : tmask |= DTK_M(TZ);
2276 4 : if (tzp == NULL)
2277 0 : return DTERR_BAD_FORMAT;
2278 : /* we'll determine the actual offset later */
2279 4 : abbrevTz = valtz;
2280 4 : abbrev = field[i];
2281 4 : ftype[i] = DTK_TZ;
2282 4 : break;
2283 :
2284 8 : case AMPM:
2285 8 : mer = val;
2286 8 : break;
2287 :
2288 0 : case ADBC:
2289 0 : bc = (val == BC);
2290 0 : break;
2291 :
2292 12 : case UNITS:
2293 12 : tmask = 0;
2294 : /* reject consecutive unhandled units */
2295 12 : if (ptype != 0)
2296 0 : return DTERR_BAD_FORMAT;
2297 12 : ptype = val;
2298 12 : break;
2299 :
2300 92 : case ISOTIME:
2301 92 : tmask = 0;
2302 : /* reject consecutive unhandled units */
2303 92 : if (ptype != 0)
2304 0 : return DTERR_BAD_FORMAT;
2305 92 : ptype = val;
2306 92 : break;
2307 :
2308 0 : case UNKNOWN_FIELD:
2309 :
2310 : /*
2311 : * Before giving up and declaring error, check to see
2312 : * if it is an all-alpha timezone name.
2313 : */
2314 0 : namedTz = pg_tzset(field[i]);
2315 0 : if (!namedTz)
2316 0 : return DTERR_BAD_FORMAT;
2317 : /* we'll apply the zone setting below */
2318 0 : tmask = DTK_M(TZ);
2319 0 : break;
2320 :
2321 0 : default:
2322 0 : return DTERR_BAD_FORMAT;
2323 : }
2324 291 : break;
2325 :
2326 0 : default:
2327 0 : return DTERR_BAD_FORMAT;
2328 : }
2329 :
2330 4366 : if (tmask & fmask)
2331 0 : return DTERR_BAD_FORMAT;
2332 4366 : fmask |= tmask;
2333 : } /* end loop over fields */
2334 :
2335 : /* reject if prefix type appeared and was never handled */
2336 2772 : if (ptype != 0)
2337 0 : return DTERR_BAD_FORMAT;
2338 :
2339 : /* do final checking/adjustment of Y/M/D fields */
2340 2772 : dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
2341 2772 : if (dterr)
2342 0 : return dterr;
2343 :
2344 : /* handle AM/PM */
2345 2772 : if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
2346 0 : return DTERR_FIELD_OVERFLOW;
2347 2772 : if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
2348 0 : tm->tm_hour = 0;
2349 2772 : else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
2350 8 : tm->tm_hour += HOURS_PER_DAY / 2;
2351 :
2352 : /* check for time overflow */
2353 2772 : if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec, *fsec))
2354 48 : return DTERR_FIELD_OVERFLOW;
2355 :
2356 2724 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2357 0 : return DTERR_BAD_FORMAT;
2358 :
2359 : /*
2360 : * If we had a full timezone spec, compute the offset (we could not do it
2361 : * before, because we may need the date to resolve DST status).
2362 : */
2363 2724 : if (namedTz != NULL)
2364 : {
2365 : long int gmtoff;
2366 :
2367 : /* daylight savings time modifier disallowed with full TZ */
2368 609 : if (fmask & DTK_M(DTZMOD))
2369 28 : return DTERR_BAD_FORMAT;
2370 :
2371 : /* if non-DST zone, we do not need to know the date */
2372 609 : if (pg_get_timezone_offset(namedTz, &gmtoff))
2373 : {
2374 557 : *tzp = -(int) gmtoff;
2375 : }
2376 : else
2377 : {
2378 : /* a date has to be specified */
2379 52 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2380 28 : return DTERR_BAD_FORMAT;
2381 24 : *tzp = DetermineTimeZoneOffset(tm, namedTz);
2382 : }
2383 : }
2384 :
2385 : /*
2386 : * Likewise, if we had a dynamic timezone abbreviation, resolve it now.
2387 : */
2388 2696 : if (abbrevTz != NULL)
2389 : {
2390 : struct pg_tm tt,
2391 0 : *tmp = &tt;
2392 :
2393 : /*
2394 : * daylight savings time modifier but no standard timezone? then error
2395 : */
2396 0 : if (fmask & DTK_M(DTZMOD))
2397 0 : return DTERR_BAD_FORMAT;
2398 :
2399 0 : if ((fmask & DTK_DATE_M) == 0)
2400 0 : GetCurrentDateTime(tmp);
2401 : else
2402 : {
2403 : /* a date has to be specified */
2404 0 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2405 0 : return DTERR_BAD_FORMAT;
2406 0 : tmp->tm_year = tm->tm_year;
2407 0 : tmp->tm_mon = tm->tm_mon;
2408 0 : tmp->tm_mday = tm->tm_mday;
2409 : }
2410 0 : tmp->tm_hour = tm->tm_hour;
2411 0 : tmp->tm_min = tm->tm_min;
2412 0 : tmp->tm_sec = tm->tm_sec;
2413 0 : *tzp = DetermineTimeZoneAbbrevOffset(tmp, abbrev, abbrevTz);
2414 0 : tm->tm_isdst = tmp->tm_isdst;
2415 : }
2416 :
2417 : /* timezone not specified? then use session timezone */
2418 2696 : if (tzp != NULL && !(fmask & DTK_M(TZ)))
2419 : {
2420 : struct pg_tm tt,
2421 1422 : *tmp = &tt;
2422 :
2423 : /*
2424 : * daylight savings time modifier but no standard timezone? then error
2425 : */
2426 1422 : if (fmask & DTK_M(DTZMOD))
2427 0 : return DTERR_BAD_FORMAT;
2428 :
2429 1422 : if ((fmask & DTK_DATE_M) == 0)
2430 1370 : GetCurrentDateTime(tmp);
2431 : else
2432 : {
2433 : /* a date has to be specified */
2434 52 : if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2435 0 : return DTERR_BAD_FORMAT;
2436 52 : tmp->tm_year = tm->tm_year;
2437 52 : tmp->tm_mon = tm->tm_mon;
2438 52 : tmp->tm_mday = tm->tm_mday;
2439 : }
2440 1422 : tmp->tm_hour = tm->tm_hour;
2441 1422 : tmp->tm_min = tm->tm_min;
2442 1422 : tmp->tm_sec = tm->tm_sec;
2443 1422 : *tzp = DetermineTimeZoneOffset(tmp, session_timezone);
2444 1422 : tm->tm_isdst = tmp->tm_isdst;
2445 : }
2446 :
2447 2696 : return 0;
2448 : }
2449 :
2450 : /*
2451 : * DecodeDate()
2452 : * Decode date string which includes delimiters.
2453 : * Return 0 if okay, a DTERR code if not.
2454 : *
2455 : * str: field to be parsed
2456 : * fmask: bitmask for field types already seen
2457 : * *tmask: receives bitmask for fields found here
2458 : * *is2digits: set to true if we find 2-digit year
2459 : * *tm: field values are stored into appropriate members of this struct
2460 : */
2461 : static int
2462 40391 : DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
2463 : struct pg_tm *tm)
2464 : {
2465 : fsec_t fsec;
2466 40391 : int nf = 0;
2467 : int i,
2468 : len;
2469 : int dterr;
2470 40391 : bool haveTextMonth = false;
2471 : int type,
2472 : val,
2473 40391 : dmask = 0;
2474 : char *field[MAXDATEFIELDS];
2475 :
2476 40391 : *tmask = 0;
2477 :
2478 : /* parse this string... */
2479 161520 : while (*str != '\0' && nf < MAXDATEFIELDS)
2480 : {
2481 : /* skip field separators */
2482 121129 : while (*str != '\0' && !isalnum((unsigned char) *str))
2483 0 : str++;
2484 :
2485 121129 : if (*str == '\0')
2486 0 : return DTERR_BAD_FORMAT; /* end of string after separator */
2487 :
2488 121129 : field[nf] = str;
2489 121129 : if (isdigit((unsigned char) *str))
2490 : {
2491 443851 : while (isdigit((unsigned char) *str))
2492 322818 : str++;
2493 : }
2494 96 : else if (isalpha((unsigned char) *str))
2495 : {
2496 384 : while (isalpha((unsigned char) *str))
2497 288 : str++;
2498 : }
2499 :
2500 : /* Just get rid of any non-digit, non-alpha characters... */
2501 121129 : if (*str != '\0')
2502 80762 : *str++ = '\0';
2503 121129 : nf++;
2504 : }
2505 :
2506 : /* look first for text fields, since that will be unambiguous month */
2507 161520 : for (i = 0; i < nf; i++)
2508 : {
2509 121129 : if (isalpha((unsigned char) *field[i]))
2510 : {
2511 96 : type = DecodeSpecial(i, field[i], &val);
2512 96 : if (type == IGNORE_DTF)
2513 0 : continue;
2514 :
2515 96 : dmask = DTK_M(type);
2516 96 : switch (type)
2517 : {
2518 96 : case MONTH:
2519 96 : tm->tm_mon = val;
2520 96 : haveTextMonth = true;
2521 96 : break;
2522 :
2523 0 : default:
2524 0 : return DTERR_BAD_FORMAT;
2525 : }
2526 96 : if (fmask & dmask)
2527 0 : return DTERR_BAD_FORMAT;
2528 :
2529 96 : fmask |= dmask;
2530 96 : *tmask |= dmask;
2531 :
2532 : /* mark this field as being completed */
2533 96 : field[i] = NULL;
2534 : }
2535 : }
2536 :
2537 : /* now pick up remaining numeric fields */
2538 161520 : for (i = 0; i < nf; i++)
2539 : {
2540 121129 : if (field[i] == NULL)
2541 96 : continue;
2542 :
2543 121033 : if ((len = strlen(field[i])) <= 0)
2544 0 : return DTERR_BAD_FORMAT;
2545 :
2546 121033 : dterr = DecodeNumber(len, field[i], haveTextMonth, fmask,
2547 : &dmask, tm,
2548 : &fsec, is2digits);
2549 121033 : if (dterr)
2550 0 : return dterr;
2551 :
2552 121033 : if (fmask & dmask)
2553 0 : return DTERR_BAD_FORMAT;
2554 :
2555 121033 : fmask |= dmask;
2556 121033 : *tmask |= dmask;
2557 : }
2558 :
2559 40391 : if ((fmask & ~(DTK_M(DOY) | DTK_M(TZ))) != DTK_DATE_M)
2560 24 : return DTERR_BAD_FORMAT;
2561 :
2562 : /* validation of the field values must wait until ValidateDate() */
2563 :
2564 40367 : return 0;
2565 : }
2566 :
2567 : /*
2568 : * ValidateDate()
2569 : * Check valid year/month/day values, handle BC and DOY cases
2570 : * Return 0 if okay, a DTERR code if not.
2571 : */
2572 : int
2573 49298 : ValidateDate(int fmask, bool isjulian, bool is2digits, bool bc,
2574 : struct pg_tm *tm)
2575 : {
2576 49298 : if (fmask & DTK_M(YEAR))
2577 : {
2578 46662 : if (isjulian)
2579 : {
2580 : /* tm_year is correct and should not be touched */
2581 : }
2582 42586 : else if (bc)
2583 : {
2584 : /* there is no year zero in AD/BC notation */
2585 222 : if (tm->tm_year <= 0)
2586 0 : return DTERR_FIELD_OVERFLOW;
2587 : /* internally, we represent 1 BC as year zero, 2 BC as -1, etc */
2588 222 : tm->tm_year = -(tm->tm_year - 1);
2589 : }
2590 42364 : else if (is2digits)
2591 : {
2592 : /* process 1 or 2-digit input as 1970-2069 AD, allow '0' and '00' */
2593 236 : if (tm->tm_year < 0) /* just paranoia */
2594 0 : return DTERR_FIELD_OVERFLOW;
2595 236 : if (tm->tm_year < 70)
2596 116 : tm->tm_year += 2000;
2597 120 : else if (tm->tm_year < 100)
2598 120 : tm->tm_year += 1900;
2599 : }
2600 : else
2601 : {
2602 : /* there is no year zero in AD/BC notation */
2603 42128 : if (tm->tm_year <= 0)
2604 8 : return DTERR_FIELD_OVERFLOW;
2605 : }
2606 : }
2607 :
2608 : /* now that we have correct year, decode DOY */
2609 49290 : if (fmask & DTK_M(DOY))
2610 : {
2611 20 : j2date(date2j(tm->tm_year, 1, 1) + tm->tm_yday - 1,
2612 : &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2613 : }
2614 :
2615 : /* check for valid month */
2616 49290 : if (fmask & DTK_M(MONTH))
2617 : {
2618 46638 : if (tm->tm_mon < 1 || tm->tm_mon > MONTHS_PER_YEAR)
2619 52 : return DTERR_MD_FIELD_OVERFLOW;
2620 : }
2621 :
2622 : /* minimal check for valid day */
2623 49238 : if (fmask & DTK_M(DAY))
2624 : {
2625 46562 : if (tm->tm_mday < 1 || tm->tm_mday > 31)
2626 92 : return DTERR_MD_FIELD_OVERFLOW;
2627 : }
2628 :
2629 49146 : if ((fmask & DTK_DATE_M) == DTK_DATE_M)
2630 : {
2631 : /*
2632 : * Check for valid day of month, now that we know for sure the month
2633 : * and year. Note we don't use MD_FIELD_OVERFLOW here, since it seems
2634 : * unlikely that "Feb 29" is a YMD-order error.
2635 : */
2636 46458 : if (tm->tm_mday > day_tab[isleap(tm->tm_year)][tm->tm_mon - 1])
2637 32 : return DTERR_FIELD_OVERFLOW;
2638 : }
2639 :
2640 49114 : return 0;
2641 : }
2642 :
2643 :
2644 : /*
2645 : * DecodeTimeCommon()
2646 : * Decode time string which includes delimiters.
2647 : * Return 0 if okay, a DTERR code if not.
2648 : * tmask and itm are output parameters.
2649 : *
2650 : * This code is shared between the timestamp and interval cases.
2651 : * We return a struct pg_itm (of which only the tm_usec, tm_sec, tm_min,
2652 : * and tm_hour fields are used) and let the wrapper functions below
2653 : * convert and range-check as necessary.
2654 : */
2655 : static int
2656 32960 : DecodeTimeCommon(char *str, int fmask, int range,
2657 : int *tmask, struct pg_itm *itm)
2658 : {
2659 : char *cp;
2660 : int dterr;
2661 32960 : fsec_t fsec = 0;
2662 :
2663 32960 : *tmask = DTK_TIME_M;
2664 :
2665 32960 : errno = 0;
2666 32960 : itm->tm_hour = strtoi64(str, &cp, 10);
2667 32960 : if (errno == ERANGE)
2668 0 : return DTERR_FIELD_OVERFLOW;
2669 32960 : if (*cp != ':')
2670 0 : return DTERR_BAD_FORMAT;
2671 32960 : errno = 0;
2672 32960 : itm->tm_min = strtoint(cp + 1, &cp, 10);
2673 32960 : if (errno == ERANGE)
2674 0 : return DTERR_FIELD_OVERFLOW;
2675 32960 : if (*cp == '\0')
2676 : {
2677 1076 : itm->tm_sec = 0;
2678 : /* If it's a MINUTE TO SECOND interval, take 2 fields as being mm:ss */
2679 1076 : if (range == (INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND)))
2680 : {
2681 12 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2682 0 : return DTERR_FIELD_OVERFLOW;
2683 12 : itm->tm_sec = itm->tm_min;
2684 12 : itm->tm_min = (int) itm->tm_hour;
2685 12 : itm->tm_hour = 0;
2686 : }
2687 : }
2688 31884 : else if (*cp == '.')
2689 : {
2690 : /* always assume mm:ss.sss is MINUTE TO SECOND */
2691 32 : dterr = ParseFractionalSecond(cp, &fsec);
2692 32 : if (dterr)
2693 8 : return dterr;
2694 24 : if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN)
2695 0 : return DTERR_FIELD_OVERFLOW;
2696 24 : itm->tm_sec = itm->tm_min;
2697 24 : itm->tm_min = (int) itm->tm_hour;
2698 24 : itm->tm_hour = 0;
2699 : }
2700 31852 : else if (*cp == ':')
2701 : {
2702 31852 : errno = 0;
2703 31852 : itm->tm_sec = strtoint(cp + 1, &cp, 10);
2704 31852 : if (errno == ERANGE)
2705 0 : return DTERR_FIELD_OVERFLOW;
2706 31852 : if (*cp == '.')
2707 : {
2708 11664 : dterr = ParseFractionalSecond(cp, &fsec);
2709 11664 : if (dterr)
2710 0 : return dterr;
2711 : }
2712 20188 : else if (*cp != '\0')
2713 0 : return DTERR_BAD_FORMAT;
2714 : }
2715 : else
2716 0 : return DTERR_BAD_FORMAT;
2717 :
2718 : /* do a sanity check; but caller must check the range of tm_hour */
2719 32952 : if (itm->tm_hour < 0 ||
2720 32952 : itm->tm_min < 0 || itm->tm_min > MINS_PER_HOUR - 1 ||
2721 32952 : itm->tm_sec < 0 || itm->tm_sec > SECS_PER_MINUTE ||
2722 32952 : fsec < 0 || fsec > USECS_PER_SEC)
2723 0 : return DTERR_FIELD_OVERFLOW;
2724 :
2725 32952 : itm->tm_usec = (int) fsec;
2726 :
2727 32952 : return 0;
2728 : }
2729 :
2730 : /*
2731 : * DecodeTime()
2732 : * Decode time string which includes delimiters.
2733 : * Return 0 if okay, a DTERR code if not.
2734 : *
2735 : * This version is used for timestamps. The results are returned into
2736 : * the tm_hour/tm_min/tm_sec fields of *tm, and microseconds into *fsec.
2737 : */
2738 : static int
2739 31635 : DecodeTime(char *str, int fmask, int range,
2740 : int *tmask, struct pg_tm *tm, fsec_t *fsec)
2741 : {
2742 : struct pg_itm itm;
2743 : int dterr;
2744 :
2745 31635 : dterr = DecodeTimeCommon(str, fmask, range,
2746 : tmask, &itm);
2747 31635 : if (dterr)
2748 0 : return dterr;
2749 :
2750 31635 : if (itm.tm_hour > INT_MAX)
2751 0 : return DTERR_FIELD_OVERFLOW;
2752 31635 : tm->tm_hour = (int) itm.tm_hour;
2753 31635 : tm->tm_min = itm.tm_min;
2754 31635 : tm->tm_sec = itm.tm_sec;
2755 31635 : *fsec = itm.tm_usec;
2756 :
2757 31635 : return 0;
2758 : }
2759 :
2760 : /*
2761 : * DecodeTimeForInterval()
2762 : * Decode time string which includes delimiters.
2763 : * Return 0 if okay, a DTERR code if not.
2764 : *
2765 : * This version is used for intervals. The results are returned into
2766 : * itm_in->tm_usec.
2767 : */
2768 : static int
2769 1325 : DecodeTimeForInterval(char *str, int fmask, int range,
2770 : int *tmask, struct pg_itm_in *itm_in)
2771 : {
2772 : struct pg_itm itm;
2773 : int dterr;
2774 :
2775 1325 : dterr = DecodeTimeCommon(str, fmask, range,
2776 : tmask, &itm);
2777 1325 : if (dterr)
2778 8 : return dterr;
2779 :
2780 1317 : itm_in->tm_usec = itm.tm_usec;
2781 1317 : if (!int64_multiply_add(itm.tm_hour, USECS_PER_HOUR, &itm_in->tm_usec) ||
2782 1317 : !int64_multiply_add(itm.tm_min, USECS_PER_MINUTE, &itm_in->tm_usec) ||
2783 1317 : !int64_multiply_add(itm.tm_sec, USECS_PER_SEC, &itm_in->tm_usec))
2784 4 : return DTERR_FIELD_OVERFLOW;
2785 :
2786 1313 : return 0;
2787 : }
2788 :
2789 :
2790 : /*
2791 : * DecodeNumber()
2792 : * Interpret plain numeric field as a date value in context.
2793 : * Return 0 if okay, a DTERR code if not.
2794 : */
2795 : static int
2796 124426 : DecodeNumber(int flen, char *str, bool haveTextMonth, int fmask,
2797 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2798 : {
2799 : int val;
2800 : char *cp;
2801 : int dterr;
2802 :
2803 124426 : *tmask = 0;
2804 :
2805 124426 : errno = 0;
2806 124426 : val = strtoint(str, &cp, 10);
2807 124426 : if (errno == ERANGE)
2808 0 : return DTERR_FIELD_OVERFLOW;
2809 124426 : if (cp == str)
2810 0 : return DTERR_BAD_FORMAT;
2811 :
2812 124426 : if (*cp == '.')
2813 : {
2814 : /*
2815 : * More than two digits before decimal point? Then could be a date or
2816 : * a run-together time: 2001.360 20011225 040506.789
2817 : */
2818 0 : if (cp - str > 2)
2819 : {
2820 0 : dterr = DecodeNumberField(flen, str,
2821 : (fmask | DTK_DATE_M),
2822 : tmask, tm,
2823 : fsec, is2digits);
2824 0 : if (dterr < 0)
2825 0 : return dterr;
2826 0 : return 0;
2827 : }
2828 :
2829 0 : dterr = ParseFractionalSecond(cp, fsec);
2830 0 : if (dterr)
2831 0 : return dterr;
2832 : }
2833 124426 : else if (*cp != '\0')
2834 0 : return DTERR_BAD_FORMAT;
2835 :
2836 : /* Special case for day of year */
2837 124426 : if (flen == 3 && (fmask & DTK_DATE_M) == DTK_M(YEAR) && val >= 1 &&
2838 : val <= 366)
2839 : {
2840 36 : *tmask = (DTK_M(DOY) | DTK_M(MONTH) | DTK_M(DAY));
2841 36 : tm->tm_yday = val;
2842 : /* tm_mon and tm_mday can't actually be set yet ... */
2843 36 : return 0;
2844 : }
2845 :
2846 : /* Switch based on what we have so far */
2847 124390 : switch (fmask & DTK_DATE_M)
2848 : {
2849 40511 : case 0:
2850 :
2851 : /*
2852 : * Nothing so far; make a decision about what we think the input
2853 : * is. There used to be lots of heuristics here, but the
2854 : * consensus now is to be paranoid. It *must* be either
2855 : * YYYY-MM-DD (with a more-than-two-digit year field), or the
2856 : * field order defined by DateOrder.
2857 : */
2858 40511 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2859 : {
2860 39331 : *tmask = DTK_M(YEAR);
2861 39331 : tm->tm_year = val;
2862 : }
2863 1180 : else if (DateOrder == DATEORDER_DMY)
2864 : {
2865 107 : *tmask = DTK_M(DAY);
2866 107 : tm->tm_mday = val;
2867 : }
2868 : else
2869 : {
2870 1073 : *tmask = DTK_M(MONTH);
2871 1073 : tm->tm_mon = val;
2872 : }
2873 40511 : break;
2874 :
2875 39259 : case (DTK_M(YEAR)):
2876 : /* Must be at second field of YY-MM-DD */
2877 39259 : *tmask = DTK_M(MONTH);
2878 39259 : tm->tm_mon = val;
2879 39259 : break;
2880 :
2881 2568 : case (DTK_M(MONTH)):
2882 2568 : if (haveTextMonth)
2883 : {
2884 : /*
2885 : * We are at the first numeric field of a date that included a
2886 : * textual month name. We want to support the variants
2887 : * MON-DD-YYYY, DD-MON-YYYY, and YYYY-MON-DD as unambiguous
2888 : * inputs. We will also accept MON-DD-YY or DD-MON-YY in
2889 : * either DMY or MDY modes, as well as YY-MON-DD in YMD mode.
2890 : */
2891 1526 : if (flen >= 3 || DateOrder == DATEORDER_YMD)
2892 : {
2893 48 : *tmask = DTK_M(YEAR);
2894 48 : tm->tm_year = val;
2895 : }
2896 : else
2897 : {
2898 1478 : *tmask = DTK_M(DAY);
2899 1478 : tm->tm_mday = val;
2900 : }
2901 : }
2902 : else
2903 : {
2904 : /* Must be at second field of MM-DD-YY */
2905 1042 : *tmask = DTK_M(DAY);
2906 1042 : tm->tm_mday = val;
2907 : }
2908 2568 : break;
2909 :
2910 39311 : case (DTK_M(YEAR) | DTK_M(MONTH)):
2911 39311 : if (haveTextMonth)
2912 : {
2913 : /* Need to accept DD-MON-YYYY even in YMD mode */
2914 84 : if (flen >= 3 && *is2digits)
2915 : {
2916 : /* Guess that first numeric field is day was wrong */
2917 20 : *tmask = DTK_M(DAY); /* YEAR is already set */
2918 20 : tm->tm_mday = tm->tm_year;
2919 20 : tm->tm_year = val;
2920 20 : *is2digits = false;
2921 : }
2922 : else
2923 : {
2924 64 : *tmask = DTK_M(DAY);
2925 64 : tm->tm_mday = val;
2926 : }
2927 : }
2928 : else
2929 : {
2930 : /* Must be at third field of YY-MM-DD */
2931 39227 : *tmask = DTK_M(DAY);
2932 39227 : tm->tm_mday = val;
2933 : }
2934 39311 : break;
2935 :
2936 95 : case (DTK_M(DAY)):
2937 : /* Must be at second field of DD-MM-YY */
2938 95 : *tmask = DTK_M(MONTH);
2939 95 : tm->tm_mon = val;
2940 95 : break;
2941 :
2942 2630 : case (DTK_M(MONTH) | DTK_M(DAY)):
2943 : /* Must be at third field of DD-MM-YY or MM-DD-YY */
2944 2630 : *tmask = DTK_M(YEAR);
2945 2630 : tm->tm_year = val;
2946 2630 : break;
2947 :
2948 16 : case (DTK_M(YEAR) | DTK_M(MONTH) | DTK_M(DAY)):
2949 : /* we have all the date, so it must be a time field */
2950 16 : dterr = DecodeNumberField(flen, str, fmask,
2951 : tmask, tm,
2952 : fsec, is2digits);
2953 16 : if (dterr < 0)
2954 8 : return dterr;
2955 8 : return 0;
2956 :
2957 0 : default:
2958 : /* Anything else is bogus input */
2959 0 : return DTERR_BAD_FORMAT;
2960 : }
2961 :
2962 : /*
2963 : * When processing a year field, mark it for adjustment if it's only one
2964 : * or two digits.
2965 : */
2966 124374 : if (*tmask == DTK_M(YEAR))
2967 42009 : *is2digits = (flen <= 2);
2968 :
2969 124374 : return 0;
2970 : }
2971 :
2972 :
2973 : /*
2974 : * DecodeNumberField()
2975 : * Interpret numeric string as a concatenated date or time field.
2976 : * Return a DTK token (>= 0) if successful, a DTERR code (< 0) if not.
2977 : *
2978 : * Use the context of previously decoded fields to help with
2979 : * the interpretation.
2980 : */
2981 : static int
2982 392 : DecodeNumberField(int len, char *str, int fmask,
2983 : int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2984 : {
2985 : char *cp;
2986 :
2987 : /*
2988 : * This function was originally meant to cope only with DTK_NUMBER fields,
2989 : * but we now sometimes abuse it to parse (parts of) DTK_DATE fields,
2990 : * which can contain letters and other punctuation. Reject if it's not a
2991 : * valid DTK_NUMBER, that is digits and decimal point(s). (ParseFraction
2992 : * will reject if there's more than one decimal point.)
2993 : */
2994 392 : if (strspn(str, "0123456789.") != len)
2995 8 : return DTERR_BAD_FORMAT;
2996 :
2997 : /*
2998 : * Have a decimal point? Then this is a date or something with a seconds
2999 : * field...
3000 : */
3001 384 : if ((cp = strchr(str, '.')) != NULL)
3002 : {
3003 : int dterr;
3004 :
3005 : /* Convert the fraction and store at *fsec */
3006 92 : dterr = ParseFractionalSecond(cp, fsec);
3007 92 : if (dterr)
3008 0 : return dterr;
3009 : /* Now truncate off the fraction for further processing */
3010 92 : *cp = '\0';
3011 92 : len = strlen(str);
3012 : }
3013 : /* No decimal point and no complete date yet? */
3014 292 : else if ((fmask & DTK_DATE_M) != DTK_DATE_M)
3015 : {
3016 164 : if (len >= 6)
3017 : {
3018 164 : *tmask = DTK_DATE_M;
3019 :
3020 : /*
3021 : * Start from end and consider first 2 as Day, next 2 as Month,
3022 : * and the rest as Year.
3023 : */
3024 164 : tm->tm_mday = atoi(str + (len - 2));
3025 164 : *(str + (len - 2)) = '\0';
3026 164 : tm->tm_mon = atoi(str + (len - 4));
3027 164 : *(str + (len - 4)) = '\0';
3028 164 : tm->tm_year = atoi(str);
3029 164 : if ((len - 4) == 2)
3030 12 : *is2digits = true;
3031 :
3032 164 : return DTK_DATE;
3033 : }
3034 : }
3035 :
3036 : /* not all time fields are specified? */
3037 220 : if ((fmask & DTK_TIME_M) != DTK_TIME_M)
3038 : {
3039 : /* hhmmss */
3040 220 : if (len == 6)
3041 : {
3042 188 : *tmask = DTK_TIME_M;
3043 188 : tm->tm_sec = atoi(str + 4);
3044 188 : *(str + 4) = '\0';
3045 188 : tm->tm_min = atoi(str + 2);
3046 188 : *(str + 2) = '\0';
3047 188 : tm->tm_hour = atoi(str);
3048 :
3049 188 : return DTK_TIME;
3050 : }
3051 : /* hhmm? */
3052 32 : else if (len == 4)
3053 : {
3054 16 : *tmask = DTK_TIME_M;
3055 16 : tm->tm_sec = 0;
3056 16 : tm->tm_min = atoi(str + 2);
3057 16 : *(str + 2) = '\0';
3058 16 : tm->tm_hour = atoi(str);
3059 :
3060 16 : return DTK_TIME;
3061 : }
3062 : }
3063 :
3064 16 : return DTERR_BAD_FORMAT;
3065 : }
3066 :
3067 :
3068 : /*
3069 : * DecodeTimezone()
3070 : * Interpret string as a numeric timezone.
3071 : *
3072 : * Return 0 if okay (and set *tzp), a DTERR code if not okay.
3073 : */
3074 : int
3075 19641 : DecodeTimezone(const char *str, int *tzp)
3076 : {
3077 : int tz;
3078 : int hr,
3079 : min,
3080 19641 : sec = 0;
3081 : char *cp;
3082 :
3083 : /* leading character must be "+" or "-" */
3084 19641 : if (*str != '+' && *str != '-')
3085 48 : return DTERR_BAD_FORMAT;
3086 :
3087 19593 : errno = 0;
3088 19593 : hr = strtoint(str + 1, &cp, 10);
3089 19593 : if (errno == ERANGE)
3090 0 : return DTERR_TZDISP_OVERFLOW;
3091 :
3092 : /* explicit delimiter? */
3093 19593 : if (*cp == ':')
3094 : {
3095 72 : errno = 0;
3096 72 : min = strtoint(cp + 1, &cp, 10);
3097 72 : if (errno == ERANGE)
3098 0 : return DTERR_TZDISP_OVERFLOW;
3099 72 : if (*cp == ':')
3100 : {
3101 16 : errno = 0;
3102 16 : sec = strtoint(cp + 1, &cp, 10);
3103 16 : if (errno == ERANGE)
3104 0 : return DTERR_TZDISP_OVERFLOW;
3105 : }
3106 : }
3107 : /* otherwise, might have run things together... */
3108 19521 : else if (*cp == '\0' && strlen(str) > 3)
3109 : {
3110 48 : min = hr % 100;
3111 48 : hr = hr / 100;
3112 : /* we could, but don't, support a run-together hhmmss format */
3113 : }
3114 : else
3115 19473 : min = 0;
3116 :
3117 : /* Range-check the values; see notes in datatype/timestamp.h */
3118 19593 : if (hr < 0 || hr > MAX_TZDISP_HOUR)
3119 8 : return DTERR_TZDISP_OVERFLOW;
3120 19585 : if (min < 0 || min >= MINS_PER_HOUR)
3121 8 : return DTERR_TZDISP_OVERFLOW;
3122 19577 : if (sec < 0 || sec >= SECS_PER_MINUTE)
3123 0 : return DTERR_TZDISP_OVERFLOW;
3124 :
3125 19577 : tz = (hr * MINS_PER_HOUR + min) * SECS_PER_MINUTE + sec;
3126 19577 : if (*str == '-')
3127 565 : tz = -tz;
3128 :
3129 19577 : *tzp = -tz;
3130 :
3131 19577 : if (*cp != '\0')
3132 0 : return DTERR_BAD_FORMAT;
3133 :
3134 19577 : return 0;
3135 : }
3136 :
3137 :
3138 : /*
3139 : * DecodeTimezoneAbbrev()
3140 : * Interpret string as a timezone abbreviation, if possible.
3141 : *
3142 : * Sets *ftype to an abbreviation type (TZ, DTZ, or DYNTZ), or UNKNOWN_FIELD if
3143 : * string is not any known abbreviation. On success, set *offset and *tz to
3144 : * represent the UTC offset (for TZ or DTZ) or underlying zone (for DYNTZ).
3145 : * Note that full timezone names (such as America/New_York) are not handled
3146 : * here, mostly for historical reasons.
3147 : *
3148 : * The function result is 0 or a DTERR code; in the latter case, *extra
3149 : * is filled as needed. Note that unknown-abbreviation is not considered
3150 : * an error case. Also note that many callers assume that the DTERR code
3151 : * is one that DateTimeParseError does not require "str" or "datatype"
3152 : * strings for.
3153 : *
3154 : * Given string must be lowercased already.
3155 : *
3156 : * Implement a cache lookup since it is likely that dates
3157 : * will be related in format.
3158 : */
3159 : int
3160 6514 : DecodeTimezoneAbbrev(int field, const char *lowtoken,
3161 : int *ftype, int *offset, pg_tz **tz,
3162 : DateTimeErrorExtra *extra)
3163 : {
3164 6514 : TzAbbrevCache *tzc = &tzabbrevcache[field];
3165 : bool isfixed;
3166 : int isdst;
3167 : const datetkn *tp;
3168 :
3169 : /*
3170 : * Do we have a cached result? Use strncmp so that we match truncated
3171 : * names, although we shouldn't really see that happen with normal
3172 : * abbreviations.
3173 : */
3174 6514 : if (strncmp(lowtoken, tzc->abbrev, TOKMAXLEN) == 0)
3175 : {
3176 1605 : *ftype = tzc->ftype;
3177 1605 : *offset = tzc->offset;
3178 1605 : *tz = tzc->tz;
3179 1605 : return 0;
3180 : }
3181 :
3182 : /*
3183 : * See if the current session_timezone recognizes it. Checking this
3184 : * before zoneabbrevtbl allows us to correctly handle abbreviations whose
3185 : * meaning varies across zones, such as "LMT".
3186 : */
3187 9818 : if (session_timezone &&
3188 4909 : TimeZoneAbbrevIsKnown(lowtoken, session_timezone,
3189 : &isfixed, offset, &isdst))
3190 : {
3191 149 : *ftype = (isfixed ? (isdst ? DTZ : TZ) : DYNTZ);
3192 149 : *tz = (isfixed ? NULL : session_timezone);
3193 : /* flip sign to agree with the convention used in zoneabbrevtbl */
3194 149 : *offset = -(*offset);
3195 : /* cache result; use strlcpy to truncate name if necessary */
3196 149 : strlcpy(tzc->abbrev, lowtoken, TOKMAXLEN + 1);
3197 149 : tzc->ftype = *ftype;
3198 149 : tzc->offset = *offset;
3199 149 : tzc->tz = *tz;
3200 149 : return 0;
3201 : }
3202 :
3203 : /* Nope, so look in zoneabbrevtbl */
3204 4760 : if (zoneabbrevtbl)
3205 4760 : tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3206 4760 : zoneabbrevtbl->numabbrevs);
3207 : else
3208 0 : tp = NULL;
3209 4760 : if (tp == NULL)
3210 : {
3211 4605 : *ftype = UNKNOWN_FIELD;
3212 4605 : *offset = 0;
3213 4605 : *tz = NULL;
3214 : /* failure results are not cached */
3215 : }
3216 : else
3217 : {
3218 155 : *ftype = tp->type;
3219 155 : if (tp->type == DYNTZ)
3220 : {
3221 20 : *offset = 0;
3222 20 : *tz = FetchDynamicTimeZone(zoneabbrevtbl, tp, extra);
3223 20 : if (*tz == NULL)
3224 0 : return DTERR_BAD_ZONE_ABBREV;
3225 : }
3226 : else
3227 : {
3228 135 : *offset = tp->value;
3229 135 : *tz = NULL;
3230 : }
3231 :
3232 : /* cache result; use strlcpy to truncate name if necessary */
3233 155 : strlcpy(tzc->abbrev, lowtoken, TOKMAXLEN + 1);
3234 155 : tzc->ftype = *ftype;
3235 155 : tzc->offset = *offset;
3236 155 : tzc->tz = *tz;
3237 : }
3238 :
3239 4760 : return 0;
3240 : }
3241 :
3242 : /*
3243 : * Reset tzabbrevcache after a change in session_timezone.
3244 : */
3245 : void
3246 10349 : ClearTimeZoneAbbrevCache(void)
3247 : {
3248 10349 : memset(tzabbrevcache, 0, sizeof(tzabbrevcache));
3249 10349 : }
3250 :
3251 :
3252 : /*
3253 : * DecodeSpecial()
3254 : * Decode text string using lookup table.
3255 : *
3256 : * Recognizes the keywords listed in datetktbl.
3257 : * Note: at one time this would also recognize timezone abbreviations,
3258 : * but no more; use DecodeTimezoneAbbrev for that.
3259 : *
3260 : * Given string must be lowercased already.
3261 : *
3262 : * Implement a cache lookup since it is likely that dates
3263 : * will be related in format.
3264 : */
3265 : int
3266 23409 : DecodeSpecial(int field, const char *lowtoken, int *val)
3267 : {
3268 : int type;
3269 : const datetkn *tp;
3270 :
3271 23409 : tp = datecache[field];
3272 : /* use strncmp so that we match truncated tokens */
3273 23409 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3274 : {
3275 6564 : tp = datebsearch(lowtoken, datetktbl, szdatetktbl);
3276 : }
3277 23409 : if (tp == NULL)
3278 : {
3279 76 : type = UNKNOWN_FIELD;
3280 76 : *val = 0;
3281 : }
3282 : else
3283 : {
3284 23333 : datecache[field] = tp;
3285 23333 : type = tp->type;
3286 23333 : *val = tp->value;
3287 : }
3288 :
3289 23409 : return type;
3290 : }
3291 :
3292 :
3293 : /*
3294 : * DecodeTimezoneName()
3295 : * Interpret string as a timezone abbreviation or name.
3296 : * Throw error if the name is not recognized.
3297 : *
3298 : * The return value indicates what kind of zone identifier it is:
3299 : * TZNAME_FIXED_OFFSET: fixed offset from UTC
3300 : * TZNAME_DYNTZ: dynamic timezone abbreviation
3301 : * TZNAME_ZONE: full tzdb zone name
3302 : *
3303 : * For TZNAME_FIXED_OFFSET, *offset receives the UTC offset (in seconds,
3304 : * with ISO sign convention: positive is east of Greenwich).
3305 : * For the other two cases, *tz receives the timezone struct representing
3306 : * the zone name or the abbreviation's underlying zone.
3307 : */
3308 : int
3309 617 : DecodeTimezoneName(const char *tzname, int *offset, pg_tz **tz)
3310 : {
3311 : char *lowzone;
3312 : int dterr,
3313 : type;
3314 : DateTimeErrorExtra extra;
3315 :
3316 : /*
3317 : * First we look in the timezone abbreviation table (to handle cases like
3318 : * "EST"), and if that fails, we look in the timezone database (to handle
3319 : * cases like "America/New_York"). This matches the order in which
3320 : * timestamp input checks the cases; it's important because the timezone
3321 : * database unwisely uses a few zone names that are identical to offset
3322 : * abbreviations.
3323 : */
3324 :
3325 : /* DecodeTimezoneAbbrev requires lowercase input */
3326 617 : lowzone = downcase_truncate_identifier(tzname,
3327 617 : strlen(tzname),
3328 : false);
3329 :
3330 617 : dterr = DecodeTimezoneAbbrev(0, lowzone, &type, offset, tz, &extra);
3331 617 : if (dterr)
3332 0 : DateTimeParseError(dterr, &extra, NULL, NULL, NULL);
3333 :
3334 617 : if (type == TZ || type == DTZ)
3335 : {
3336 : /* fixed-offset abbreviation, return the offset */
3337 207 : return TZNAME_FIXED_OFFSET;
3338 : }
3339 410 : else if (type == DYNTZ)
3340 : {
3341 : /* dynamic-offset abbreviation, return its referenced timezone */
3342 143 : return TZNAME_DYNTZ;
3343 : }
3344 : else
3345 : {
3346 : /* try it as a full zone name */
3347 267 : *tz = pg_tzset(tzname);
3348 267 : if (*tz == NULL)
3349 8 : ereport(ERROR,
3350 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
3351 : errmsg("time zone \"%s\" not recognized", tzname)));
3352 259 : return TZNAME_ZONE;
3353 : }
3354 : }
3355 :
3356 : /*
3357 : * DecodeTimezoneNameToTz()
3358 : * Interpret string as a timezone abbreviation or name.
3359 : * Throw error if the name is not recognized.
3360 : *
3361 : * This is a simple wrapper for DecodeTimezoneName that produces a pg_tz *
3362 : * result in all cases.
3363 : */
3364 : pg_tz *
3365 70 : DecodeTimezoneNameToTz(const char *tzname)
3366 : {
3367 : pg_tz *result;
3368 : int offset;
3369 :
3370 70 : if (DecodeTimezoneName(tzname, &offset, &result) == TZNAME_FIXED_OFFSET)
3371 : {
3372 : /* fixed-offset abbreviation, get a pg_tz descriptor for that */
3373 22 : result = pg_tzset_offset(-offset); /* flip to POSIX sign convention */
3374 : }
3375 70 : return result;
3376 : }
3377 :
3378 : /*
3379 : * DecodeTimezoneAbbrevPrefix()
3380 : * Interpret prefix of string as a timezone abbreviation, if possible.
3381 : *
3382 : * This has roughly the same functionality as DecodeTimezoneAbbrev(),
3383 : * but the API is adapted to the needs of formatting.c. Notably,
3384 : * we will match the longest possible prefix of the given string
3385 : * rather than insisting on a complete match, and downcasing is applied
3386 : * here rather than in the caller.
3387 : *
3388 : * Returns the length of the timezone abbreviation, or -1 if not recognized.
3389 : * On success, sets *offset to the GMT offset for the abbreviation if it
3390 : * is a fixed-offset abbreviation, or sets *tz to the pg_tz struct for
3391 : * a dynamic abbreviation.
3392 : */
3393 : int
3394 2374 : DecodeTimezoneAbbrevPrefix(const char *str, int *offset, pg_tz **tz)
3395 : {
3396 : char lowtoken[TOKMAXLEN + 1];
3397 : int len;
3398 :
3399 2374 : *offset = 0; /* avoid uninitialized vars on failure */
3400 2374 : *tz = NULL;
3401 :
3402 : /* Downcase as much of the string as we could need */
3403 2466 : for (len = 0; len < TOKMAXLEN; len++)
3404 : {
3405 2466 : if (*str == '\0' || !isalpha((unsigned char) *str))
3406 : break;
3407 92 : lowtoken[len] = pg_tolower((unsigned char) *str++);
3408 : }
3409 2374 : lowtoken[len] = '\0';
3410 :
3411 : /*
3412 : * We could avoid doing repeated binary searches if we cared to duplicate
3413 : * datebsearch here, but it's not clear that such an optimization would be
3414 : * worth the trouble. In common cases there's probably not anything after
3415 : * the zone abbrev anyway. So just search with successively truncated
3416 : * strings.
3417 : */
3418 2402 : while (len > 0)
3419 : {
3420 : bool isfixed;
3421 : int isdst;
3422 : const datetkn *tp;
3423 :
3424 : /* See if the current session_timezone recognizes it. */
3425 104 : if (session_timezone &&
3426 52 : TimeZoneAbbrevIsKnown(lowtoken, session_timezone,
3427 : &isfixed, offset, &isdst))
3428 : {
3429 4 : if (isfixed)
3430 : {
3431 : /* flip sign to agree with the convention in zoneabbrevtbl */
3432 4 : *offset = -(*offset);
3433 : }
3434 : else
3435 : {
3436 : /* Caller must resolve the abbrev's current meaning */
3437 0 : *tz = session_timezone;
3438 : }
3439 24 : return len;
3440 : }
3441 :
3442 : /* Known in zoneabbrevtbl? */
3443 48 : if (zoneabbrevtbl)
3444 48 : tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
3445 48 : zoneabbrevtbl->numabbrevs);
3446 : else
3447 0 : tp = NULL;
3448 48 : if (tp != NULL)
3449 : {
3450 20 : if (tp->type == DYNTZ)
3451 : {
3452 : DateTimeErrorExtra extra;
3453 4 : pg_tz *tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp,
3454 : &extra);
3455 :
3456 4 : if (tzp != NULL)
3457 : {
3458 : /* Caller must resolve the abbrev's current meaning */
3459 4 : *tz = tzp;
3460 4 : return len;
3461 : }
3462 : }
3463 : else
3464 : {
3465 : /* Fixed-offset zone abbrev, so it's easy */
3466 16 : *offset = tp->value;
3467 16 : return len;
3468 : }
3469 : }
3470 :
3471 : /* Nope, try the next shorter string. */
3472 28 : lowtoken[--len] = '\0';
3473 : }
3474 :
3475 : /* Did not find a match */
3476 2350 : return -1;
3477 : }
3478 :
3479 :
3480 : /*
3481 : * ClearPgItmIn
3482 : *
3483 : * Zero out a pg_itm_in
3484 : */
3485 : static inline void
3486 43258 : ClearPgItmIn(struct pg_itm_in *itm_in)
3487 : {
3488 43258 : itm_in->tm_usec = 0;
3489 43258 : itm_in->tm_mday = 0;
3490 43258 : itm_in->tm_mon = 0;
3491 43258 : itm_in->tm_year = 0;
3492 43258 : }
3493 :
3494 :
3495 : /*
3496 : * DecodeInterval()
3497 : * Interpret previously parsed fields for general time interval.
3498 : * Returns 0 if successful, DTERR code if bogus input detected.
3499 : * dtype and itm_in are output parameters.
3500 : *
3501 : * Allow "date" field DTK_DATE since this could be just
3502 : * an unsigned floating point number. - thomas 1997-11-16
3503 : *
3504 : * Allow ISO-style time span, with implicit units on number of days
3505 : * preceding an hh:mm:ss field. - thomas 1998-04-30
3506 : *
3507 : * itm_in remains undefined for infinite interval values for which dtype alone
3508 : * suffices.
3509 : */
3510 : int
3511 42850 : DecodeInterval(char **field, int *ftype, int nf, int range,
3512 : int *dtype, struct pg_itm_in *itm_in)
3513 : {
3514 42850 : bool force_negative = false;
3515 42850 : bool is_before = false;
3516 42850 : bool parsing_unit_val = false;
3517 : char *cp;
3518 42850 : int fmask = 0,
3519 : tmask,
3520 : type,
3521 : uval;
3522 : int i;
3523 : int dterr;
3524 : int64 val;
3525 : double fval;
3526 :
3527 42850 : *dtype = DTK_DELTA;
3528 42850 : type = IGNORE_DTF;
3529 42850 : ClearPgItmIn(itm_in);
3530 :
3531 : /*----------
3532 : * The SQL standard defines the interval literal
3533 : * '-1 1:00:00'
3534 : * to mean "negative 1 days and negative 1 hours", while Postgres
3535 : * traditionally treats this as meaning "negative 1 days and positive
3536 : * 1 hours". In SQL_STANDARD intervalstyle, we apply the leading sign
3537 : * to all fields if there are no other explicit signs.
3538 : *
3539 : * We leave the signs alone if there are additional explicit signs.
3540 : * This protects us against misinterpreting postgres-style dump output,
3541 : * since the postgres-style output code has always put an explicit sign on
3542 : * all fields following a negative field. But note that SQL-spec output
3543 : * is ambiguous and can be misinterpreted on load! (So it's best practice
3544 : * to dump in postgres style, not SQL style.)
3545 : *----------
3546 : */
3547 42850 : if (IntervalStyle == INTSTYLE_SQL_STANDARD && nf > 0 && *field[0] == '-')
3548 : {
3549 25 : force_negative = true;
3550 : /* Check for additional explicit signs */
3551 162 : for (i = 1; i < nf; i++)
3552 : {
3553 149 : if (*field[i] == '-' || *field[i] == '+')
3554 : {
3555 12 : force_negative = false;
3556 12 : break;
3557 : }
3558 : }
3559 : }
3560 :
3561 : /* read through list backwards to pick up units before values */
3562 129878 : for (i = nf - 1; i >= 0; i--)
3563 : {
3564 87756 : switch (ftype[i])
3565 : {
3566 920 : case DTK_TIME:
3567 920 : dterr = DecodeTimeForInterval(field[i], fmask, range,
3568 : &tmask, itm_in);
3569 920 : if (dterr)
3570 12 : return dterr;
3571 908 : if (force_negative &&
3572 1 : itm_in->tm_usec > 0)
3573 1 : itm_in->tm_usec = -itm_in->tm_usec;
3574 908 : type = DTK_DAY;
3575 908 : parsing_unit_val = false;
3576 908 : break;
3577 :
3578 1828 : case DTK_TZ:
3579 :
3580 : /*
3581 : * Timezone means a token with a leading sign character and at
3582 : * least one digit; there could be ':', '.', '-' embedded in
3583 : * it as well.
3584 : */
3585 : Assert(*field[i] == '-' || *field[i] == '+');
3586 :
3587 : /*
3588 : * Check for signed hh:mm or hh:mm:ss. If so, process exactly
3589 : * like DTK_TIME case above, plus handling the sign.
3590 : */
3591 2233 : if (strchr(field[i] + 1, ':') != NULL &&
3592 405 : DecodeTimeForInterval(field[i] + 1, fmask, range,
3593 : &tmask, itm_in) == 0)
3594 : {
3595 405 : if (*field[i] == '-')
3596 : {
3597 : /* flip the sign on time field */
3598 365 : if (itm_in->tm_usec == PG_INT64_MIN)
3599 0 : return DTERR_FIELD_OVERFLOW;
3600 365 : itm_in->tm_usec = -itm_in->tm_usec;
3601 : }
3602 :
3603 405 : if (force_negative &&
3604 0 : itm_in->tm_usec > 0)
3605 0 : itm_in->tm_usec = -itm_in->tm_usec;
3606 :
3607 : /*
3608 : * Set the next type to be a day, if units are not
3609 : * specified. This handles the case of '1 +02:03' since we
3610 : * are reading right to left.
3611 : */
3612 405 : type = DTK_DAY;
3613 405 : parsing_unit_val = false;
3614 405 : break;
3615 : }
3616 :
3617 : /*
3618 : * Otherwise, fall through to DTK_NUMBER case, which can
3619 : * handle signed float numbers and signed year-month values.
3620 : */
3621 :
3622 : pg_fallthrough;
3623 :
3624 : case DTK_DATE:
3625 : case DTK_NUMBER:
3626 43120 : if (type == IGNORE_DTF)
3627 : {
3628 : /* use typmod to decide what rightmost field is */
3629 400 : switch (range)
3630 : {
3631 4 : case INTERVAL_MASK(YEAR):
3632 4 : type = DTK_YEAR;
3633 4 : break;
3634 20 : case INTERVAL_MASK(MONTH):
3635 : case INTERVAL_MASK(YEAR) | INTERVAL_MASK(MONTH):
3636 20 : type = DTK_MONTH;
3637 20 : break;
3638 12 : case INTERVAL_MASK(DAY):
3639 12 : type = DTK_DAY;
3640 12 : break;
3641 16 : case INTERVAL_MASK(HOUR):
3642 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR):
3643 16 : type = DTK_HOUR;
3644 16 : break;
3645 16 : case INTERVAL_MASK(MINUTE):
3646 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3647 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3648 16 : type = DTK_MINUTE;
3649 16 : break;
3650 40 : case INTERVAL_MASK(SECOND):
3651 : case INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3652 : case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3653 : case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3654 40 : type = DTK_SECOND;
3655 40 : break;
3656 292 : default:
3657 292 : type = DTK_SECOND;
3658 292 : break;
3659 : }
3660 : }
3661 :
3662 43120 : errno = 0;
3663 43120 : val = strtoi64(field[i], &cp, 10);
3664 43120 : if (errno == ERANGE)
3665 8 : return DTERR_FIELD_OVERFLOW;
3666 :
3667 43112 : if (*cp == '-')
3668 : {
3669 : /* SQL "years-months" syntax */
3670 : int val2;
3671 :
3672 40 : val2 = strtoint(cp + 1, &cp, 10);
3673 40 : if (errno == ERANGE || val2 < 0 || val2 >= MONTHS_PER_YEAR)
3674 0 : return DTERR_FIELD_OVERFLOW;
3675 40 : if (*cp != '\0')
3676 0 : return DTERR_BAD_FORMAT;
3677 40 : type = DTK_MONTH;
3678 40 : if (*field[i] == '-')
3679 4 : val2 = -val2;
3680 40 : if (pg_mul_s64_overflow(val, MONTHS_PER_YEAR, &val))
3681 0 : return DTERR_FIELD_OVERFLOW;
3682 40 : if (pg_add_s64_overflow(val, val2, &val))
3683 0 : return DTERR_FIELD_OVERFLOW;
3684 40 : fval = 0;
3685 : }
3686 43072 : else if (*cp == '.')
3687 : {
3688 324 : dterr = ParseFraction(cp, &fval);
3689 324 : if (dterr)
3690 0 : return dterr;
3691 324 : if (*field[i] == '-')
3692 92 : fval = -fval;
3693 : }
3694 42748 : else if (*cp == '\0')
3695 42492 : fval = 0;
3696 : else
3697 256 : return DTERR_BAD_FORMAT;
3698 :
3699 42856 : tmask = 0; /* DTK_M(type); */
3700 :
3701 42856 : if (force_negative)
3702 : {
3703 : /* val and fval should be of same sign, but test anyway */
3704 53 : if (val > 0)
3705 40 : val = -val;
3706 53 : if (fval > 0)
3707 12 : fval = -fval;
3708 : }
3709 :
3710 42856 : switch (type)
3711 : {
3712 220 : case DTK_MICROSEC:
3713 220 : if (!AdjustMicroseconds(val, fval, 1, itm_in))
3714 24 : return DTERR_FIELD_OVERFLOW;
3715 196 : tmask = DTK_M(MICROSECOND);
3716 196 : break;
3717 :
3718 70 : case DTK_MILLISEC:
3719 70 : if (!AdjustMicroseconds(val, fval, 1000, itm_in))
3720 8 : return DTERR_FIELD_OVERFLOW;
3721 62 : tmask = DTK_M(MILLISECOND);
3722 62 : break;
3723 :
3724 608 : case DTK_SECOND:
3725 608 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
3726 8 : return DTERR_FIELD_OVERFLOW;
3727 :
3728 : /*
3729 : * If any subseconds were specified, consider this
3730 : * microsecond and millisecond input as well.
3731 : */
3732 600 : if (fval == 0)
3733 492 : tmask = DTK_M(SECOND);
3734 : else
3735 108 : tmask = DTK_ALL_SECS_M;
3736 600 : break;
3737 :
3738 245 : case DTK_MINUTE:
3739 245 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
3740 8 : return DTERR_FIELD_OVERFLOW;
3741 237 : tmask = DTK_M(MINUTE);
3742 237 : break;
3743 :
3744 439 : case DTK_HOUR:
3745 439 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
3746 8 : return DTERR_FIELD_OVERFLOW;
3747 431 : tmask = DTK_M(HOUR);
3748 431 : type = DTK_DAY; /* set for next field */
3749 431 : break;
3750 :
3751 4018 : case DTK_DAY:
3752 4018 : if (!AdjustDays(val, 1, itm_in) ||
3753 3970 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
3754 60 : return DTERR_FIELD_OVERFLOW;
3755 3958 : tmask = DTK_M(DAY);
3756 3958 : break;
3757 :
3758 64 : case DTK_WEEK:
3759 64 : if (!AdjustDays(val, 7, itm_in) ||
3760 48 : !AdjustFractDays(fval, 7, itm_in))
3761 32 : return DTERR_FIELD_OVERFLOW;
3762 32 : tmask = DTK_M(WEEK);
3763 32 : break;
3764 :
3765 732 : case DTK_MONTH:
3766 732 : if (!AdjustMonths(val, itm_in) ||
3767 692 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
3768 56 : return DTERR_FIELD_OVERFLOW;
3769 676 : tmask = DTK_M(MONTH);
3770 676 : break;
3771 :
3772 36328 : case DTK_YEAR:
3773 36328 : if (!AdjustYears(val, 1, itm_in) ||
3774 36296 : !AdjustFractYears(fval, 1, itm_in))
3775 40 : return DTERR_FIELD_OVERFLOW;
3776 36288 : tmask = DTK_M(YEAR);
3777 36288 : break;
3778 :
3779 44 : case DTK_DECADE:
3780 44 : if (!AdjustYears(val, 10, itm_in) ||
3781 28 : !AdjustFractYears(fval, 10, itm_in))
3782 24 : return DTERR_FIELD_OVERFLOW;
3783 20 : tmask = DTK_M(DECADE);
3784 20 : break;
3785 :
3786 44 : case DTK_CENTURY:
3787 44 : if (!AdjustYears(val, 100, itm_in) ||
3788 28 : !AdjustFractYears(fval, 100, itm_in))
3789 24 : return DTERR_FIELD_OVERFLOW;
3790 20 : tmask = DTK_M(CENTURY);
3791 20 : break;
3792 :
3793 44 : case DTK_MILLENNIUM:
3794 44 : if (!AdjustYears(val, 1000, itm_in) ||
3795 28 : !AdjustFractYears(fval, 1000, itm_in))
3796 24 : return DTERR_FIELD_OVERFLOW;
3797 20 : tmask = DTK_M(MILLENNIUM);
3798 20 : break;
3799 :
3800 0 : default:
3801 0 : return DTERR_BAD_FORMAT;
3802 : }
3803 42540 : parsing_unit_val = false;
3804 42540 : break;
3805 :
3806 43311 : case DTK_STRING:
3807 : case DTK_SPECIAL:
3808 : /* reject consecutive unhandled units */
3809 43311 : if (parsing_unit_val)
3810 8 : return DTERR_BAD_FORMAT;
3811 43303 : type = DecodeUnits(i, field[i], &uval);
3812 43303 : if (type == UNKNOWN_FIELD)
3813 708 : type = DecodeSpecial(i, field[i], &uval);
3814 43303 : if (type == IGNORE_DTF)
3815 0 : continue;
3816 :
3817 43303 : tmask = 0; /* DTK_M(type); */
3818 43303 : switch (type)
3819 : {
3820 42527 : case UNITS:
3821 42527 : type = uval;
3822 42527 : parsing_unit_val = true;
3823 42527 : break;
3824 :
3825 68 : case AGO:
3826 :
3827 : /*
3828 : * "ago" is only allowed to appear at the end of the
3829 : * interval.
3830 : */
3831 68 : if (i != nf - 1)
3832 8 : return DTERR_BAD_FORMAT;
3833 60 : is_before = true;
3834 60 : type = uval;
3835 60 : break;
3836 :
3837 684 : case RESERV:
3838 684 : tmask = (DTK_DATE_M | DTK_TIME_M);
3839 :
3840 : /*
3841 : * Only reserved words corresponding to infinite
3842 : * intervals are accepted.
3843 : */
3844 684 : if (uval != DTK_LATE && uval != DTK_EARLY)
3845 24 : return DTERR_BAD_FORMAT;
3846 :
3847 : /*
3848 : * Infinity cannot be followed by anything else. We
3849 : * could allow "ago" to reverse the sign of infinity
3850 : * but using signed infinity is more intuitive.
3851 : */
3852 660 : if (i != nf - 1)
3853 8 : return DTERR_BAD_FORMAT;
3854 :
3855 652 : *dtype = uval;
3856 652 : break;
3857 :
3858 24 : default:
3859 24 : return DTERR_BAD_FORMAT;
3860 : }
3861 43239 : break;
3862 :
3863 0 : default:
3864 0 : return DTERR_BAD_FORMAT;
3865 : }
3866 :
3867 87092 : if (tmask & fmask)
3868 64 : return DTERR_BAD_FORMAT;
3869 87028 : fmask |= tmask;
3870 : }
3871 :
3872 : /* ensure that at least one time field has been found */
3873 42122 : if (fmask == 0)
3874 4 : return DTERR_BAD_FORMAT;
3875 :
3876 : /* reject if unit appeared and was never handled */
3877 42118 : if (parsing_unit_val)
3878 4 : return DTERR_BAD_FORMAT;
3879 :
3880 : /* finally, AGO negates everything */
3881 42114 : if (is_before)
3882 : {
3883 28 : if (itm_in->tm_usec == PG_INT64_MIN ||
3884 20 : itm_in->tm_mday == INT_MIN ||
3885 16 : itm_in->tm_mon == INT_MIN ||
3886 12 : itm_in->tm_year == INT_MIN)
3887 16 : return DTERR_FIELD_OVERFLOW;
3888 :
3889 12 : itm_in->tm_usec = -itm_in->tm_usec;
3890 12 : itm_in->tm_mday = -itm_in->tm_mday;
3891 12 : itm_in->tm_mon = -itm_in->tm_mon;
3892 12 : itm_in->tm_year = -itm_in->tm_year;
3893 : }
3894 :
3895 42098 : return 0;
3896 : }
3897 :
3898 :
3899 : /*
3900 : * Helper functions to avoid duplicated code in DecodeISO8601Interval.
3901 : *
3902 : * Parse a decimal value and break it into integer and fractional parts.
3903 : * Set *endptr to end+1 of the parsed substring.
3904 : * Returns 0 or DTERR code.
3905 : */
3906 : static int
3907 636 : ParseISO8601Number(char *str, char **endptr, int64 *ipart, double *fpart)
3908 : {
3909 : double val;
3910 :
3911 : /*
3912 : * Historically this has accepted anything that strtod() would take,
3913 : * notably including "e" notation, so continue doing that. This is
3914 : * slightly annoying because the precision of double is less than that of
3915 : * int64, so we would lose accuracy for inputs larger than 2^53 or so.
3916 : * However, historically we rejected inputs outside the int32 range,
3917 : * making that concern moot. What we do now is reject abs(val) above
3918 : * 1.0e15 (a round number a bit less than 2^50), so that any accepted
3919 : * value will have an exact integer part, and thereby a fraction part with
3920 : * abs(*fpart) less than 1. In the absence of field complaints it doesn't
3921 : * seem worth working harder.
3922 : */
3923 636 : if (!(isdigit((unsigned char) *str) || *str == '-' || *str == '.'))
3924 0 : return DTERR_BAD_FORMAT;
3925 636 : errno = 0;
3926 636 : val = strtod(str, endptr);
3927 : /* did we not see anything that looks like a double? */
3928 636 : if (*endptr == str || errno != 0)
3929 4 : return DTERR_BAD_FORMAT;
3930 : /* watch out for overflow, including infinities; reject NaN too */
3931 632 : if (isnan(val) || val < -1.0e15 || val > 1.0e15)
3932 0 : return DTERR_FIELD_OVERFLOW;
3933 : /* be very sure we truncate towards zero (cf dtrunc()) */
3934 632 : if (val >= 0)
3935 488 : *ipart = (int64) floor(val);
3936 : else
3937 144 : *ipart = (int64) -floor(-val);
3938 632 : *fpart = val - *ipart;
3939 : /* Callers expect this to hold */
3940 : Assert(*fpart > -1.0 && *fpart < 1.0);
3941 632 : return 0;
3942 : }
3943 :
3944 : /*
3945 : * Determine number of integral digits in a valid ISO 8601 number field
3946 : * (we should ignore sign and any fraction part)
3947 : */
3948 : static int
3949 44 : ISO8601IntegerWidth(char *fieldstart)
3950 : {
3951 : /* We might have had a leading '-' */
3952 44 : if (*fieldstart == '-')
3953 12 : fieldstart++;
3954 44 : return strspn(fieldstart, "0123456789");
3955 : }
3956 :
3957 :
3958 : /*
3959 : * DecodeISO8601Interval()
3960 : * Decode an ISO 8601 time interval of the "format with designators"
3961 : * (section 4.4.3.2) or "alternative format" (section 4.4.3.3)
3962 : * Examples: P1D for 1 day
3963 : * PT1H for 1 hour
3964 : * P2Y6M7DT1H30M for 2 years, 6 months, 7 days 1 hour 30 min
3965 : * P0002-06-07T01:30:00 the same value in alternative format
3966 : *
3967 : * Returns 0 if successful, DTERR code if bogus input detected.
3968 : * Note: error code should be DTERR_BAD_FORMAT if input doesn't look like
3969 : * ISO8601, otherwise this could cause unexpected error messages.
3970 : * dtype and itm_in are output parameters.
3971 : *
3972 : * A couple exceptions from the spec:
3973 : * - a week field ('W') may coexist with other units
3974 : * - allows decimals in fields other than the least significant unit.
3975 : */
3976 : int
3977 408 : DecodeISO8601Interval(char *str,
3978 : int *dtype, struct pg_itm_in *itm_in)
3979 : {
3980 408 : bool datepart = true;
3981 408 : bool havefield = false;
3982 :
3983 408 : *dtype = DTK_DELTA;
3984 408 : ClearPgItmIn(itm_in);
3985 :
3986 408 : if (strlen(str) < 2 || str[0] != 'P')
3987 152 : return DTERR_BAD_FORMAT;
3988 :
3989 256 : str++;
3990 748 : while (*str)
3991 : {
3992 : char *fieldstart;
3993 : int64 val;
3994 : double fval;
3995 : char unit;
3996 : int dterr;
3997 :
3998 676 : if (*str == 'T') /* T indicates the beginning of the time part */
3999 : {
4000 132 : datepart = false;
4001 132 : havefield = false;
4002 132 : str++;
4003 160 : continue;
4004 : }
4005 :
4006 544 : fieldstart = str;
4007 544 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4008 544 : if (dterr)
4009 184 : return dterr;
4010 :
4011 : /*
4012 : * Note: we could step off the end of the string here. Code below
4013 : * *must* exit the loop if unit == '\0'.
4014 : */
4015 540 : unit = *str++;
4016 :
4017 540 : if (datepart)
4018 : {
4019 312 : switch (unit) /* before T: Y M W D */
4020 : {
4021 56 : case 'Y':
4022 56 : if (!AdjustYears(val, 1, itm_in) ||
4023 56 : !AdjustFractYears(fval, 1, itm_in))
4024 8 : return DTERR_FIELD_OVERFLOW;
4025 48 : break;
4026 72 : case 'M':
4027 72 : if (!AdjustMonths(val, itm_in) ||
4028 64 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
4029 16 : return DTERR_FIELD_OVERFLOW;
4030 56 : break;
4031 36 : case 'W':
4032 36 : if (!AdjustDays(val, 7, itm_in) ||
4033 28 : !AdjustFractDays(fval, 7, itm_in))
4034 16 : return DTERR_FIELD_OVERFLOW;
4035 20 : break;
4036 88 : case 'D':
4037 88 : if (!AdjustDays(val, 1, itm_in) ||
4038 64 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
4039 24 : return DTERR_FIELD_OVERFLOW;
4040 64 : break;
4041 20 : case 'T': /* ISO 8601 4.4.3.3 Alternative Format / Basic */
4042 : case '\0':
4043 20 : if (ISO8601IntegerWidth(fieldstart) == 8 && !havefield)
4044 : {
4045 4 : if (!AdjustYears(val / 10000, 1, itm_in) ||
4046 4 : !AdjustMonths((val / 100) % 100, itm_in) ||
4047 4 : !AdjustDays(val % 100, 1, itm_in) ||
4048 4 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
4049 0 : return DTERR_FIELD_OVERFLOW;
4050 4 : if (unit == '\0')
4051 0 : return 0;
4052 4 : datepart = false;
4053 4 : havefield = false;
4054 4 : continue;
4055 : }
4056 : /* Else fall through to extended alternative format */
4057 : pg_fallthrough;
4058 : case '-': /* ISO 8601 4.4.3.3 Alternative Format,
4059 : * Extended */
4060 56 : if (havefield)
4061 0 : return DTERR_BAD_FORMAT;
4062 :
4063 56 : if (!AdjustYears(val, 1, itm_in) ||
4064 48 : !AdjustFractYears(fval, 1, itm_in))
4065 8 : return DTERR_FIELD_OVERFLOW;
4066 48 : if (unit == '\0')
4067 4 : return 0;
4068 44 : if (unit == 'T')
4069 : {
4070 4 : datepart = false;
4071 4 : havefield = false;
4072 4 : continue;
4073 : }
4074 :
4075 40 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4076 40 : if (dterr)
4077 0 : return dterr;
4078 40 : if (!AdjustMonths(val, itm_in) ||
4079 36 : !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in))
4080 4 : return DTERR_FIELD_OVERFLOW;
4081 36 : if (*str == '\0')
4082 4 : return 0;
4083 32 : if (*str == 'T')
4084 : {
4085 4 : datepart = false;
4086 4 : havefield = false;
4087 4 : continue;
4088 : }
4089 28 : if (*str != '-')
4090 0 : return DTERR_BAD_FORMAT;
4091 28 : str++;
4092 :
4093 28 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4094 28 : if (dterr)
4095 0 : return dterr;
4096 28 : if (!AdjustDays(val, 1, itm_in) ||
4097 24 : !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in))
4098 4 : return DTERR_FIELD_OVERFLOW;
4099 24 : if (*str == '\0')
4100 8 : return 0;
4101 16 : if (*str == 'T')
4102 : {
4103 16 : datepart = false;
4104 16 : havefield = false;
4105 16 : continue;
4106 : }
4107 0 : return DTERR_BAD_FORMAT;
4108 0 : default:
4109 : /* not a valid date unit suffix */
4110 0 : return DTERR_BAD_FORMAT;
4111 : }
4112 : }
4113 : else
4114 : {
4115 228 : switch (unit) /* after T: H M S */
4116 : {
4117 72 : case 'H':
4118 72 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
4119 24 : return DTERR_FIELD_OVERFLOW;
4120 48 : break;
4121 40 : case 'M':
4122 40 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
4123 0 : return DTERR_FIELD_OVERFLOW;
4124 40 : break;
4125 64 : case 'S':
4126 64 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
4127 8 : return DTERR_FIELD_OVERFLOW;
4128 56 : break;
4129 24 : case '\0': /* ISO 8601 4.4.3.3 Alternative Format */
4130 24 : if (ISO8601IntegerWidth(fieldstart) == 6 && !havefield)
4131 : {
4132 4 : if (!AdjustMicroseconds(val / 10000, 0, USECS_PER_HOUR, itm_in) ||
4133 4 : !AdjustMicroseconds((val / 100) % 100, 0, USECS_PER_MINUTE, itm_in) ||
4134 4 : !AdjustMicroseconds(val % 100, 0, USECS_PER_SEC, itm_in) ||
4135 4 : !AdjustFractMicroseconds(fval, 1, itm_in))
4136 0 : return DTERR_FIELD_OVERFLOW;
4137 4 : return 0;
4138 : }
4139 : /* Else fall through to extended alternative format */
4140 : pg_fallthrough;
4141 : case ':': /* ISO 8601 4.4.3.3 Alternative Format,
4142 : * Extended */
4143 48 : if (havefield)
4144 0 : return DTERR_BAD_FORMAT;
4145 :
4146 48 : if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in))
4147 20 : return DTERR_FIELD_OVERFLOW;
4148 28 : if (unit == '\0')
4149 12 : return 0;
4150 :
4151 16 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4152 16 : if (dterr)
4153 0 : return dterr;
4154 16 : if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in))
4155 4 : return DTERR_FIELD_OVERFLOW;
4156 12 : if (*str == '\0')
4157 4 : return 0;
4158 8 : if (*str != ':')
4159 0 : return DTERR_BAD_FORMAT;
4160 8 : str++;
4161 :
4162 8 : dterr = ParseISO8601Number(str, &str, &val, &fval);
4163 8 : if (dterr)
4164 0 : return dterr;
4165 8 : if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in))
4166 0 : return DTERR_FIELD_OVERFLOW;
4167 8 : if (*str == '\0')
4168 8 : return 0;
4169 0 : return DTERR_BAD_FORMAT;
4170 :
4171 0 : default:
4172 : /* not a valid time unit suffix */
4173 0 : return DTERR_BAD_FORMAT;
4174 : }
4175 : }
4176 :
4177 332 : havefield = true;
4178 : }
4179 :
4180 72 : return 0;
4181 : }
4182 :
4183 :
4184 : /*
4185 : * DecodeUnits()
4186 : * Decode text string using lookup table.
4187 : *
4188 : * This routine recognizes keywords associated with time interval units.
4189 : *
4190 : * Given string must be lowercased already.
4191 : *
4192 : * Implement a cache lookup since it is likely that dates
4193 : * will be related in format.
4194 : */
4195 : int
4196 75100 : DecodeUnits(int field, const char *lowtoken, int *val)
4197 : {
4198 : int type;
4199 : const datetkn *tp;
4200 :
4201 75100 : tp = deltacache[field];
4202 : /* use strncmp so that we match truncated tokens */
4203 75100 : if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
4204 : {
4205 31923 : tp = datebsearch(lowtoken, deltatktbl, szdeltatktbl);
4206 : }
4207 75100 : if (tp == NULL)
4208 : {
4209 18975 : type = UNKNOWN_FIELD;
4210 18975 : *val = 0;
4211 : }
4212 : else
4213 : {
4214 56125 : deltacache[field] = tp;
4215 56125 : type = tp->type;
4216 56125 : *val = tp->value;
4217 : }
4218 :
4219 75100 : return type;
4220 : } /* DecodeUnits() */
4221 :
4222 : /*
4223 : * Report an error detected by one of the datetime input processing routines.
4224 : *
4225 : * dterr is the error code, and *extra contains any auxiliary info we need
4226 : * for the error report. extra can be NULL if not needed for the particular
4227 : * dterr value.
4228 : *
4229 : * str is the original input string, and datatype is the name of the datatype
4230 : * we were trying to accept. (For some DTERR codes, these are not used and
4231 : * can be NULL.)
4232 : *
4233 : * If escontext points to an ErrorSaveContext node, that is filled instead
4234 : * of throwing an error.
4235 : *
4236 : * Note: it might seem useless to distinguish DTERR_INTERVAL_OVERFLOW and
4237 : * DTERR_TZDISP_OVERFLOW from DTERR_FIELD_OVERFLOW, but SQL99 mandates three
4238 : * separate SQLSTATE codes, so ...
4239 : */
4240 : void
4241 1168 : DateTimeParseError(int dterr, DateTimeErrorExtra *extra,
4242 : const char *str, const char *datatype,
4243 : Node *escontext)
4244 : {
4245 1168 : switch (dterr)
4246 : {
4247 144 : case DTERR_FIELD_OVERFLOW:
4248 144 : errsave(escontext,
4249 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4250 : errmsg("date/time field value out of range: \"%s\"",
4251 : str)));
4252 16 : break;
4253 120 : case DTERR_MD_FIELD_OVERFLOW:
4254 : /* <nanny>same as above, but add hint about DateStyle</nanny> */
4255 120 : errsave(escontext,
4256 : (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
4257 : errmsg("date/time field value out of range: \"%s\"",
4258 : str),
4259 : errhint("Perhaps you need a different \"DateStyle\" setting.")));
4260 0 : break;
4261 480 : case DTERR_INTERVAL_OVERFLOW:
4262 480 : errsave(escontext,
4263 : (errcode(ERRCODE_INTERVAL_FIELD_OVERFLOW),
4264 : errmsg("interval field value out of range: \"%s\"",
4265 : str)));
4266 0 : break;
4267 8 : case DTERR_TZDISP_OVERFLOW:
4268 8 : errsave(escontext,
4269 : (errcode(ERRCODE_INVALID_TIME_ZONE_DISPLACEMENT_VALUE),
4270 : errmsg("time zone displacement out of range: \"%s\"",
4271 : str)));
4272 0 : break;
4273 24 : case DTERR_BAD_TIMEZONE:
4274 24 : errsave(escontext,
4275 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
4276 : errmsg("time zone \"%s\" not recognized",
4277 : extra->dtee_timezone)));
4278 16 : break;
4279 0 : case DTERR_BAD_ZONE_ABBREV:
4280 0 : errsave(escontext,
4281 : (errcode(ERRCODE_CONFIG_FILE_ERROR),
4282 : errmsg("time zone \"%s\" not recognized",
4283 : extra->dtee_timezone),
4284 : errdetail("This time zone name appears in the configuration file for time zone abbreviation \"%s\".",
4285 : extra->dtee_abbrev)));
4286 0 : break;
4287 392 : case DTERR_BAD_FORMAT:
4288 : default:
4289 392 : errsave(escontext,
4290 : (errcode(ERRCODE_INVALID_DATETIME_FORMAT),
4291 : errmsg("invalid input syntax for type %s: \"%s\"",
4292 : datatype, str)));
4293 56 : break;
4294 : }
4295 88 : }
4296 :
4297 : /*
4298 : * datebsearch()
4299 : * Binary search -- from Knuth (6.2.1) Algorithm B. Special case like this
4300 : * is WAY faster than the generic bsearch().
4301 : */
4302 : static const datetkn *
4303 44384 : datebsearch(const char *key, const datetkn *base, int nel)
4304 : {
4305 44384 : if (nel > 0)
4306 : {
4307 44384 : const datetkn *last = base + nel - 1,
4308 : *position;
4309 : int result;
4310 :
4311 277347 : while (last >= base)
4312 : {
4313 252842 : position = base + ((last - base) >> 1);
4314 : /* precheck the first character for a bit of extra speed */
4315 252842 : result = (int) key[0] - (int) position->token[0];
4316 252842 : if (result == 0)
4317 : {
4318 : /* use strncmp so that we match truncated tokens */
4319 69229 : result = strncmp(key, position->token, TOKMAXLEN);
4320 69229 : if (result == 0)
4321 19879 : return position;
4322 : }
4323 232963 : if (result < 0)
4324 114979 : last = position - 1;
4325 : else
4326 117984 : base = position + 1;
4327 : }
4328 : }
4329 24505 : return NULL;
4330 : }
4331 :
4332 : /*
4333 : * EncodeTimezone()
4334 : * Copies representation of a numeric timezone offset to str.
4335 : *
4336 : * Returns a pointer to the new end of string. No NUL terminator is put
4337 : * there; callers are responsible for NUL terminating str themselves.
4338 : */
4339 : static char *
4340 32529 : EncodeTimezone(char *str, int tz, int style)
4341 : {
4342 : int hour,
4343 : min,
4344 : sec;
4345 :
4346 32529 : sec = abs(tz);
4347 32529 : min = sec / SECS_PER_MINUTE;
4348 32529 : sec -= min * SECS_PER_MINUTE;
4349 32529 : hour = min / MINS_PER_HOUR;
4350 32529 : min -= hour * MINS_PER_HOUR;
4351 :
4352 : /* TZ is negated compared to sign we wish to display ... */
4353 32529 : *str++ = (tz <= 0 ? '+' : '-');
4354 :
4355 32529 : if (sec != 0)
4356 : {
4357 0 : str = pg_ultostr_zeropad(str, hour, 2);
4358 0 : *str++ = ':';
4359 0 : str = pg_ultostr_zeropad(str, min, 2);
4360 0 : *str++ = ':';
4361 0 : str = pg_ultostr_zeropad(str, sec, 2);
4362 : }
4363 32529 : else if (min != 0 || style == USE_XSD_DATES)
4364 : {
4365 588 : str = pg_ultostr_zeropad(str, hour, 2);
4366 588 : *str++ = ':';
4367 588 : str = pg_ultostr_zeropad(str, min, 2);
4368 : }
4369 : else
4370 31941 : str = pg_ultostr_zeropad(str, hour, 2);
4371 32529 : return str;
4372 : }
4373 :
4374 : /*
4375 : * EncodeDateOnly()
4376 : * Encode date as local time.
4377 : */
4378 : void
4379 13978 : EncodeDateOnly(struct pg_tm *tm, int style, char *str)
4380 : {
4381 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4382 :
4383 13978 : switch (style)
4384 : {
4385 9161 : case USE_ISO_DATES:
4386 : case USE_XSD_DATES:
4387 : /* compatible with ISO date formats */
4388 9161 : str = pg_ultostr_zeropad(str,
4389 9161 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4390 9161 : *str++ = '-';
4391 9161 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4392 9161 : *str++ = '-';
4393 9161 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4394 9161 : break;
4395 :
4396 0 : case USE_SQL_DATES:
4397 : /* compatible with Oracle/Ingres date formats */
4398 0 : 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 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4407 0 : *str++ = '/';
4408 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4409 : }
4410 0 : *str++ = '/';
4411 0 : str = pg_ultostr_zeropad(str,
4412 0 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4413 0 : break;
4414 :
4415 4 : case USE_GERMAN_DATES:
4416 : /* German-style date format */
4417 4 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4418 4 : *str++ = '.';
4419 4 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4420 4 : *str++ = '.';
4421 4 : str = pg_ultostr_zeropad(str,
4422 4 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4423 4 : break;
4424 :
4425 4813 : case USE_POSTGRES_DATES:
4426 : default:
4427 : /* traditional date-only style for Postgres */
4428 4813 : if (DateOrder == DATEORDER_DMY)
4429 : {
4430 0 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4431 0 : *str++ = '-';
4432 0 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4433 : }
4434 : else
4435 : {
4436 4813 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4437 4813 : *str++ = '-';
4438 4813 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4439 : }
4440 4813 : *str++ = '-';
4441 4813 : str = pg_ultostr_zeropad(str,
4442 4813 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4443 4813 : break;
4444 : }
4445 :
4446 13978 : if (tm->tm_year <= 0)
4447 : {
4448 52 : memcpy(str, " BC", 3); /* Don't copy NUL */
4449 52 : str += 3;
4450 : }
4451 13978 : *str = '\0';
4452 13978 : }
4453 :
4454 :
4455 : /*
4456 : * EncodeTimeOnly()
4457 : * Encode time fields only.
4458 : *
4459 : * tm and fsec are the value to encode, print_tz determines whether to include
4460 : * a time zone (the difference between time and timetz types), tz is the
4461 : * numeric time zone offset, style is the date style, str is where to write the
4462 : * output.
4463 : */
4464 : void
4465 8490 : EncodeTimeOnly(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, int style, char *str)
4466 : {
4467 8490 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4468 8490 : *str++ = ':';
4469 8490 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4470 8490 : *str++ = ':';
4471 8490 : str = AppendSeconds(str, tm->tm_sec, fsec, MAX_TIME_PRECISION, true);
4472 8490 : if (print_tz)
4473 4435 : str = EncodeTimezone(str, tz, style);
4474 8490 : *str = '\0';
4475 8490 : }
4476 :
4477 :
4478 : /*
4479 : * EncodeDateTime()
4480 : * Encode date and time interpreted as local time.
4481 : *
4482 : * tm and fsec are the value to encode, print_tz determines whether to include
4483 : * a time zone (the difference between timestamp and timestamptz types), tz is
4484 : * the numeric time zone offset, tzn is the textual time zone, which if
4485 : * specified will be used instead of tz by some styles, style is the date
4486 : * style, str is where to write the output.
4487 : *
4488 : * Supported date styles:
4489 : * Postgres - day mon hh:mm:ss yyyy tz
4490 : * SQL - mm/dd/yyyy hh:mm:ss.ss tz
4491 : * ISO - yyyy-mm-dd hh:mm:ss+/-tz
4492 : * German - dd.mm.yyyy hh:mm:ss tz
4493 : * XSD - yyyy-mm-ddThh:mm:ss.ss+/-tz
4494 : */
4495 : void
4496 64154 : EncodeDateTime(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, const char *tzn, int style, char *str)
4497 : {
4498 : int day;
4499 :
4500 : Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
4501 :
4502 : /*
4503 : * Negative tm_isdst means we have no valid time zone translation.
4504 : */
4505 64154 : if (tm->tm_isdst < 0)
4506 23990 : print_tz = false;
4507 :
4508 64154 : switch (style)
4509 : {
4510 44642 : case USE_ISO_DATES:
4511 : case USE_XSD_DATES:
4512 : /* Compatible with ISO-8601 date formats */
4513 44642 : str = pg_ultostr_zeropad(str,
4514 44642 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4515 44642 : *str++ = '-';
4516 44642 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4517 44642 : *str++ = '-';
4518 44642 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4519 44642 : *str++ = (style == USE_ISO_DATES) ? ' ' : 'T';
4520 44642 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4521 44642 : *str++ = ':';
4522 44642 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4523 44642 : *str++ = ':';
4524 44642 : str = AppendTimestampSeconds(str, tm, fsec);
4525 44642 : if (print_tz)
4526 28094 : str = EncodeTimezone(str, tz, style);
4527 44642 : break;
4528 :
4529 520 : case USE_SQL_DATES:
4530 : /* Compatible with Oracle/Ingres date formats */
4531 520 : if (DateOrder == DATEORDER_DMY)
4532 : {
4533 256 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4534 256 : *str++ = '/';
4535 256 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4536 : }
4537 : else
4538 : {
4539 264 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4540 264 : *str++ = '/';
4541 264 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4542 : }
4543 520 : *str++ = '/';
4544 520 : str = pg_ultostr_zeropad(str,
4545 520 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4546 520 : *str++ = ' ';
4547 520 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4548 520 : *str++ = ':';
4549 520 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4550 520 : *str++ = ':';
4551 520 : str = AppendTimestampSeconds(str, tm, fsec);
4552 :
4553 : /*
4554 : * Note: the uses of %.*s in this function would be risky if the
4555 : * timezone names ever contain non-ASCII characters, since we are
4556 : * not being careful to do encoding-aware clipping. However, all
4557 : * TZ abbreviations in the IANA database are plain ASCII.
4558 : */
4559 520 : if (print_tz)
4560 : {
4561 12 : if (tzn)
4562 : {
4563 12 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4564 12 : str += strlen(str);
4565 : }
4566 : else
4567 0 : str = EncodeTimezone(str, tz, style);
4568 : }
4569 520 : break;
4570 :
4571 16 : case USE_GERMAN_DATES:
4572 : /* German variant on European style */
4573 16 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4574 16 : *str++ = '.';
4575 16 : str = pg_ultostr_zeropad(str, tm->tm_mon, 2);
4576 16 : *str++ = '.';
4577 16 : str = pg_ultostr_zeropad(str,
4578 16 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4579 16 : *str++ = ' ';
4580 16 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4581 16 : *str++ = ':';
4582 16 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4583 16 : *str++ = ':';
4584 16 : str = AppendTimestampSeconds(str, tm, fsec);
4585 :
4586 16 : if (print_tz)
4587 : {
4588 16 : if (tzn)
4589 : {
4590 16 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4591 16 : str += strlen(str);
4592 : }
4593 : else
4594 0 : str = EncodeTimezone(str, tz, style);
4595 : }
4596 16 : break;
4597 :
4598 18976 : case USE_POSTGRES_DATES:
4599 : default:
4600 : /* Backward-compatible with traditional Postgres abstime dates */
4601 18976 : day = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday);
4602 18976 : tm->tm_wday = j2day(day);
4603 18976 : memcpy(str, days[tm->tm_wday], 3);
4604 18976 : str += 3;
4605 18976 : *str++ = ' ';
4606 18976 : if (DateOrder == DATEORDER_DMY)
4607 : {
4608 264 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4609 264 : *str++ = ' ';
4610 264 : memcpy(str, months[tm->tm_mon - 1], 3);
4611 264 : str += 3;
4612 : }
4613 : else
4614 : {
4615 18712 : memcpy(str, months[tm->tm_mon - 1], 3);
4616 18712 : str += 3;
4617 18712 : *str++ = ' ';
4618 18712 : str = pg_ultostr_zeropad(str, tm->tm_mday, 2);
4619 : }
4620 18976 : *str++ = ' ';
4621 18976 : str = pg_ultostr_zeropad(str, tm->tm_hour, 2);
4622 18976 : *str++ = ':';
4623 18976 : str = pg_ultostr_zeropad(str, tm->tm_min, 2);
4624 18976 : *str++ = ':';
4625 18976 : str = AppendTimestampSeconds(str, tm, fsec);
4626 18976 : *str++ = ' ';
4627 18976 : str = pg_ultostr_zeropad(str,
4628 18976 : (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4629 :
4630 18976 : if (print_tz)
4631 : {
4632 12042 : if (tzn)
4633 : {
4634 12042 : sprintf(str, " %.*s", MAXTZLEN, tzn);
4635 12042 : str += strlen(str);
4636 : }
4637 : else
4638 : {
4639 : /*
4640 : * We have a time zone, but no string version. Use the
4641 : * numeric form, but be sure to include a leading space to
4642 : * avoid formatting something which would be rejected by
4643 : * the date/time parser later. - thomas 2001-10-19
4644 : */
4645 0 : *str++ = ' ';
4646 0 : str = EncodeTimezone(str, tz, style);
4647 : }
4648 : }
4649 18976 : break;
4650 : }
4651 :
4652 64154 : if (tm->tm_year <= 0)
4653 : {
4654 186 : memcpy(str, " BC", 3); /* Don't copy NUL */
4655 186 : str += 3;
4656 : }
4657 64154 : *str = '\0';
4658 64154 : }
4659 :
4660 :
4661 : /*
4662 : * Helper functions to avoid duplicated code in EncodeInterval.
4663 : */
4664 :
4665 : /* Append an ISO-8601-style interval field, but only if value isn't zero */
4666 : static char *
4667 140 : AddISO8601IntPart(char *cp, int64 value, char units)
4668 : {
4669 140 : if (value == 0)
4670 36 : return cp;
4671 104 : sprintf(cp, "%" PRId64 "%c", value, units);
4672 104 : return cp + strlen(cp);
4673 : }
4674 :
4675 : /* Append a postgres-style interval field, but only if value isn't zero */
4676 : static char *
4677 10056 : AddPostgresIntPart(char *cp, int64 value, const char *units,
4678 : bool *is_zero, bool *is_before)
4679 : {
4680 10056 : if (value == 0)
4681 5501 : return cp;
4682 13665 : sprintf(cp, "%s%s%" PRId64 " %s%s",
4683 4555 : (!*is_zero) ? " " : "",
4684 4555 : (*is_before && value > 0) ? "+" : "",
4685 : value,
4686 : units,
4687 : (value != 1) ? "s" : "");
4688 :
4689 : /*
4690 : * Each nonzero field sets is_before for (only) the next one. This is a
4691 : * tad bizarre but it's how it worked before...
4692 : */
4693 4555 : *is_before = (value < 0);
4694 4555 : *is_zero = false;
4695 4555 : return cp + strlen(cp);
4696 : }
4697 :
4698 : /* Append a verbose-style interval field, but only if value isn't zero */
4699 : static char *
4700 28215 : AddVerboseIntPart(char *cp, int64 value, const char *units,
4701 : bool *is_zero, bool *is_before)
4702 : {
4703 28215 : if (value == 0)
4704 18730 : return cp;
4705 : /* first nonzero value sets is_before */
4706 9485 : if (*is_zero)
4707 : {
4708 5230 : *is_before = (value < 0);
4709 5230 : value = i64abs(value);
4710 : }
4711 4255 : else if (*is_before)
4712 904 : value = -value;
4713 9485 : sprintf(cp, " %" PRId64 " %s%s", value, units, (value == 1) ? "" : "s");
4714 9485 : *is_zero = false;
4715 9485 : return cp + strlen(cp);
4716 : }
4717 :
4718 :
4719 : /*
4720 : * EncodeInterval()
4721 : * Interpret time structure as a delta time and convert to string.
4722 : *
4723 : * Support "traditional Postgres" and ISO-8601 styles.
4724 : * Actually, afaik ISO does not address time interval formatting,
4725 : * but this looks similar to the spec for absolute date/time.
4726 : * - thomas 1998-04-30
4727 : *
4728 : * Actually, afaik, ISO 8601 does specify formats for "time
4729 : * intervals...[of the]...format with time-unit designators", which
4730 : * are pretty ugly. The format looks something like
4731 : * P1Y1M1DT1H1M1.12345S
4732 : * but useful for exchanging data with computers instead of humans.
4733 : * - ron 2003-07-14
4734 : *
4735 : * And ISO's SQL 2008 standard specifies standards for
4736 : * "year-month literal"s (that look like '2-3') and
4737 : * "day-time literal"s (that look like ('4 5:6:7')
4738 : */
4739 : void
4740 9111 : EncodeInterval(struct pg_itm *itm, int style, char *str)
4741 : {
4742 9111 : char *cp = str;
4743 9111 : int year = itm->tm_year;
4744 9111 : int mon = itm->tm_mon;
4745 9111 : int64 mday = itm->tm_mday; /* tm_mday could be INT_MIN */
4746 9111 : int64 hour = itm->tm_hour;
4747 9111 : int min = itm->tm_min;
4748 9111 : int sec = itm->tm_sec;
4749 9111 : int fsec = itm->tm_usec;
4750 9111 : bool is_before = false;
4751 9111 : bool is_zero = true;
4752 :
4753 : /*
4754 : * The sign of year and month are guaranteed to match, since they are
4755 : * stored internally as "month". But we'll need to check for is_before and
4756 : * is_zero when determining the signs of day and hour/minute/seconds
4757 : * fields.
4758 : */
4759 9111 : switch (style)
4760 : {
4761 : /* SQL Standard interval format */
4762 84 : case INTSTYLE_SQL_STANDARD:
4763 : {
4764 64 : bool has_negative = year < 0 || mon < 0 ||
4765 44 : mday < 0 || hour < 0 ||
4766 148 : min < 0 || sec < 0 || fsec < 0;
4767 68 : bool has_positive = year > 0 || mon > 0 ||
4768 44 : mday > 0 || hour > 0 ||
4769 152 : min > 0 || sec > 0 || fsec > 0;
4770 84 : bool has_year_month = year != 0 || mon != 0;
4771 28 : bool has_day_time = mday != 0 || hour != 0 ||
4772 112 : min != 0 || sec != 0 || fsec != 0;
4773 84 : bool has_day = mday != 0;
4774 148 : bool sql_standard_value = !(has_negative && has_positive) &&
4775 64 : !(has_year_month && has_day_time);
4776 :
4777 : /*
4778 : * SQL Standard wants only 1 "<sign>" preceding the whole
4779 : * interval ... but can't do that if mixed signs.
4780 : */
4781 84 : if (has_negative && sql_standard_value)
4782 : {
4783 20 : *cp++ = '-';
4784 20 : year = -year;
4785 20 : mon = -mon;
4786 20 : mday = -mday;
4787 20 : hour = -hour;
4788 20 : min = -min;
4789 20 : sec = -sec;
4790 20 : fsec = -fsec;
4791 : }
4792 :
4793 84 : if (!has_negative && !has_positive)
4794 : {
4795 8 : sprintf(cp, "0");
4796 : }
4797 76 : else if (!sql_standard_value)
4798 : {
4799 : /*
4800 : * For non sql-standard interval values, force outputting
4801 : * the signs to avoid ambiguities with intervals with
4802 : * mixed sign components.
4803 : */
4804 36 : char year_sign = (year < 0 || mon < 0) ? '-' : '+';
4805 36 : char day_sign = (mday < 0) ? '-' : '+';
4806 52 : char sec_sign = (hour < 0 || min < 0 ||
4807 16 : sec < 0 || fsec < 0) ? '-' : '+';
4808 :
4809 36 : sprintf(cp, "%c%d-%d %c%" PRId64 " %c%" PRId64 ":%02d:",
4810 : year_sign, abs(year), abs(mon),
4811 : day_sign, i64abs(mday),
4812 : sec_sign, i64abs(hour), abs(min));
4813 36 : cp += strlen(cp);
4814 36 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4815 36 : *cp = '\0';
4816 : }
4817 40 : else if (has_year_month)
4818 : {
4819 12 : sprintf(cp, "%d-%d", year, mon);
4820 : }
4821 28 : else if (has_day)
4822 : {
4823 20 : sprintf(cp, "%" PRId64 " %" PRId64 ":%02d:",
4824 : mday, hour, min);
4825 20 : cp += strlen(cp);
4826 20 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4827 20 : *cp = '\0';
4828 : }
4829 : else
4830 : {
4831 8 : sprintf(cp, "%" PRId64 ":%02d:", hour, min);
4832 8 : cp += strlen(cp);
4833 8 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4834 8 : *cp = '\0';
4835 : }
4836 : }
4837 84 : break;
4838 :
4839 : /* ISO 8601 "time-intervals by duration only" */
4840 32 : case INTSTYLE_ISO_8601:
4841 : /* special-case zero to avoid printing nothing */
4842 32 : if (year == 0 && mon == 0 && mday == 0 &&
4843 4 : hour == 0 && min == 0 && sec == 0 && fsec == 0)
4844 : {
4845 4 : sprintf(cp, "PT0S");
4846 4 : break;
4847 : }
4848 28 : *cp++ = 'P';
4849 28 : cp = AddISO8601IntPart(cp, year, 'Y');
4850 28 : cp = AddISO8601IntPart(cp, mon, 'M');
4851 28 : cp = AddISO8601IntPart(cp, mday, 'D');
4852 28 : if (hour != 0 || min != 0 || sec != 0 || fsec != 0)
4853 24 : *cp++ = 'T';
4854 28 : cp = AddISO8601IntPart(cp, hour, 'H');
4855 28 : cp = AddISO8601IntPart(cp, min, 'M');
4856 28 : if (sec != 0 || fsec != 0)
4857 : {
4858 24 : if (sec < 0 || fsec < 0)
4859 8 : *cp++ = '-';
4860 24 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4861 24 : *cp++ = 'S';
4862 24 : *cp++ = '\0';
4863 : }
4864 28 : break;
4865 :
4866 : /* Compatible with postgresql < 8.4 when DateStyle = 'iso' */
4867 3352 : case INTSTYLE_POSTGRES:
4868 3352 : cp = AddPostgresIntPart(cp, year, "year", &is_zero, &is_before);
4869 :
4870 : /*
4871 : * Ideally we should spell out "month" like we do for "year" and
4872 : * "day". However, for backward compatibility, we can't easily
4873 : * fix this. bjm 2011-05-24
4874 : */
4875 3352 : cp = AddPostgresIntPart(cp, mon, "mon", &is_zero, &is_before);
4876 3352 : cp = AddPostgresIntPart(cp, mday, "day", &is_zero, &is_before);
4877 3352 : if (is_zero || hour != 0 || min != 0 || sec != 0 || fsec != 0)
4878 : {
4879 2738 : bool minus = (hour < 0 || min < 0 || sec < 0 || fsec < 0);
4880 :
4881 8073 : sprintf(cp, "%s%s%02" PRId64 ":%02d:",
4882 2738 : is_zero ? "" : " ",
4883 2597 : (minus ? "-" : (is_before ? "+" : "")),
4884 : i64abs(hour), abs(min));
4885 2738 : cp += strlen(cp);
4886 2738 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4887 2738 : *cp = '\0';
4888 : }
4889 3352 : break;
4890 :
4891 : /* Compatible with postgresql < 8.4 when DateStyle != 'iso' */
4892 5643 : case INTSTYLE_POSTGRES_VERBOSE:
4893 : default:
4894 5643 : strcpy(cp, "@");
4895 5643 : cp++;
4896 5643 : cp = AddVerboseIntPart(cp, year, "year", &is_zero, &is_before);
4897 5643 : cp = AddVerboseIntPart(cp, mon, "mon", &is_zero, &is_before);
4898 5643 : cp = AddVerboseIntPart(cp, mday, "day", &is_zero, &is_before);
4899 5643 : cp = AddVerboseIntPart(cp, hour, "hour", &is_zero, &is_before);
4900 5643 : cp = AddVerboseIntPart(cp, min, "min", &is_zero, &is_before);
4901 5643 : if (sec != 0 || fsec != 0)
4902 : {
4903 2174 : *cp++ = ' ';
4904 2174 : if (sec < 0 || (sec == 0 && fsec < 0))
4905 : {
4906 679 : if (is_zero)
4907 228 : is_before = true;
4908 451 : else if (!is_before)
4909 3 : *cp++ = '-';
4910 : }
4911 1495 : else if (is_before)
4912 8 : *cp++ = '-';
4913 2174 : cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4914 : /* We output "ago", not negatives, so use abs(). */
4915 2174 : sprintf(cp, " sec%s",
4916 2174 : (abs(sec) != 1 || fsec != 0) ? "s" : "");
4917 2174 : is_zero = false;
4918 : }
4919 : /* identically zero? then put in a unitless zero... */
4920 5643 : if (is_zero)
4921 151 : strcat(cp, " 0");
4922 5643 : if (is_before)
4923 942 : strcat(cp, " ago");
4924 5643 : break;
4925 : }
4926 9111 : }
4927 :
4928 :
4929 : /*
4930 : * We've been burnt by stupid errors in the ordering of the datetkn tables
4931 : * once too often. Arrange to check them during postmaster start.
4932 : */
4933 : static bool
4934 2032 : CheckDateTokenTable(const char *tablename, const datetkn *base, int nel)
4935 : {
4936 2032 : bool ok = true;
4937 : int i;
4938 :
4939 137160 : for (i = 0; i < nel; i++)
4940 : {
4941 : /* check for token strings that don't fit */
4942 135128 : if (strlen(base[i].token) > TOKMAXLEN)
4943 : {
4944 : /* %.*s is safe since all our tokens are ASCII */
4945 0 : elog(LOG, "token too long in %s table: \"%.*s\"",
4946 : tablename,
4947 : TOKMAXLEN + 1, base[i].token);
4948 0 : ok = false;
4949 0 : break; /* don't risk applying strcmp */
4950 : }
4951 : /* check for out of order */
4952 135128 : if (i > 0 &&
4953 133096 : strcmp(base[i - 1].token, base[i].token) >= 0)
4954 : {
4955 0 : elog(LOG, "ordering error in %s table: \"%s\" >= \"%s\"",
4956 : tablename,
4957 : base[i - 1].token,
4958 : base[i].token);
4959 0 : ok = false;
4960 : }
4961 : }
4962 2032 : return ok;
4963 : }
4964 :
4965 : bool
4966 1016 : CheckDateTokenTables(void)
4967 : {
4968 1016 : bool ok = true;
4969 :
4970 : Assert(UNIX_EPOCH_JDATE == date2j(1970, 1, 1));
4971 : Assert(POSTGRES_EPOCH_JDATE == date2j(2000, 1, 1));
4972 :
4973 1016 : ok &= CheckDateTokenTable("datetktbl", datetktbl, szdatetktbl);
4974 1016 : ok &= CheckDateTokenTable("deltatktbl", deltatktbl, szdeltatktbl);
4975 1016 : return ok;
4976 : }
4977 :
4978 : /*
4979 : * Common code for temporal prosupport functions: simplify, if possible,
4980 : * a call to a temporal type's length-coercion function.
4981 : *
4982 : * Types time, timetz, timestamp and timestamptz each have a range of allowed
4983 : * precisions. An unspecified precision is rigorously equivalent to the
4984 : * highest specifiable precision. We can replace the function call with a
4985 : * no-op RelabelType if it is coercing to the same or higher precision as the
4986 : * input is known to have.
4987 : *
4988 : * The input Node is always a FuncExpr, but to reduce the #include footprint
4989 : * of datetime.h, we declare it as Node *.
4990 : *
4991 : * Note: timestamp_scale throws an error when the typmod is out of range, but
4992 : * we can't get there from a cast: our typmodin will have caught it already.
4993 : */
4994 : Node *
4995 16 : TemporalSimplify(int32 max_precis, Node *node)
4996 : {
4997 16 : FuncExpr *expr = castNode(FuncExpr, node);
4998 16 : Node *ret = NULL;
4999 : Node *typmod;
5000 :
5001 : Assert(list_length(expr->args) >= 2);
5002 :
5003 16 : typmod = (Node *) lsecond(expr->args);
5004 :
5005 16 : if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
5006 : {
5007 16 : Node *source = (Node *) linitial(expr->args);
5008 16 : int32 old_precis = exprTypmod(source);
5009 16 : int32 new_precis = DatumGetInt32(((Const *) typmod)->constvalue);
5010 :
5011 16 : if (new_precis < 0 || new_precis == max_precis ||
5012 0 : (old_precis >= 0 && new_precis >= old_precis))
5013 0 : ret = relabel_to_typmod(source, new_precis);
5014 : }
5015 :
5016 16 : return ret;
5017 : }
5018 :
5019 : /*
5020 : * This function gets called during timezone config file load or reload
5021 : * to create the final array of timezone tokens. The argument array
5022 : * is already sorted in name order.
5023 : *
5024 : * The result is a TimeZoneAbbrevTable (which must be a single guc_malloc'd
5025 : * chunk) or NULL on alloc failure. No other error conditions are defined.
5026 : */
5027 : TimeZoneAbbrevTable *
5028 8715 : ConvertTimeZoneAbbrevs(struct tzEntry *abbrevs, int n)
5029 : {
5030 : TimeZoneAbbrevTable *tbl;
5031 : Size tbl_size;
5032 : int i;
5033 :
5034 : /* Space for fixed fields and datetkn array */
5035 8715 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
5036 8715 : n * sizeof(datetkn);
5037 8715 : tbl_size = MAXALIGN(tbl_size);
5038 : /* Count up space for dynamic abbreviations */
5039 1708148 : for (i = 0; i < n; i++)
5040 : {
5041 1699433 : struct tzEntry *abbr = abbrevs + i;
5042 :
5043 1699433 : if (abbr->zone != NULL)
5044 : {
5045 : Size dsize;
5046 :
5047 435746 : dsize = offsetof(DynamicZoneAbbrev, zone) +
5048 435746 : strlen(abbr->zone) + 1;
5049 435746 : tbl_size += MAXALIGN(dsize);
5050 : }
5051 : }
5052 :
5053 : /* Alloc the result ... */
5054 8715 : tbl = guc_malloc(LOG, tbl_size);
5055 8715 : if (!tbl)
5056 0 : return NULL;
5057 :
5058 : /* ... and fill it in */
5059 8715 : tbl->tblsize = tbl_size;
5060 8715 : tbl->numabbrevs = n;
5061 : /* in this loop, tbl_size reprises the space calculation above */
5062 8715 : tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
5063 8715 : n * sizeof(datetkn);
5064 8715 : tbl_size = MAXALIGN(tbl_size);
5065 1708148 : for (i = 0; i < n; i++)
5066 : {
5067 1699433 : struct tzEntry *abbr = abbrevs + i;
5068 1699433 : datetkn *dtoken = tbl->abbrevs + i;
5069 :
5070 : /* use strlcpy to truncate name if necessary */
5071 1699433 : strlcpy(dtoken->token, abbr->abbrev, TOKMAXLEN + 1);
5072 1699433 : if (abbr->zone != NULL)
5073 : {
5074 : /* Allocate a DynamicZoneAbbrev for this abbreviation */
5075 : DynamicZoneAbbrev *dtza;
5076 : Size dsize;
5077 :
5078 435746 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tbl_size);
5079 435746 : dtza->tz = NULL;
5080 435746 : strcpy(dtza->zone, abbr->zone);
5081 :
5082 435746 : dtoken->type = DYNTZ;
5083 : /* value is offset from table start to DynamicZoneAbbrev */
5084 435746 : dtoken->value = (int32) tbl_size;
5085 :
5086 435746 : dsize = offsetof(DynamicZoneAbbrev, zone) +
5087 435746 : strlen(abbr->zone) + 1;
5088 435746 : tbl_size += MAXALIGN(dsize);
5089 : }
5090 : else
5091 : {
5092 1263687 : dtoken->type = abbr->is_dst ? DTZ : TZ;
5093 1263687 : dtoken->value = abbr->offset;
5094 : }
5095 : }
5096 :
5097 : /* Assert the two loops above agreed on size calculations */
5098 : Assert(tbl->tblsize == tbl_size);
5099 :
5100 : /* Check the ordering, if testing */
5101 : Assert(CheckDateTokenTable("timezone abbreviations", tbl->abbrevs, n));
5102 :
5103 8715 : return tbl;
5104 : }
5105 :
5106 : /*
5107 : * Install a TimeZoneAbbrevTable as the active table.
5108 : *
5109 : * Caller is responsible that the passed table doesn't go away while in use.
5110 : */
5111 : void
5112 8605 : InstallTimeZoneAbbrevs(TimeZoneAbbrevTable *tbl)
5113 : {
5114 8605 : zoneabbrevtbl = tbl;
5115 : /* reset tzabbrevcache, which may contain results from old table */
5116 8605 : memset(tzabbrevcache, 0, sizeof(tzabbrevcache));
5117 8605 : }
5118 :
5119 : /*
5120 : * Helper subroutine to locate pg_tz timezone for a dynamic abbreviation.
5121 : *
5122 : * On failure, returns NULL and fills *extra for a DTERR_BAD_ZONE_ABBREV error.
5123 : */
5124 : static pg_tz *
5125 820 : FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp,
5126 : DateTimeErrorExtra *extra)
5127 : {
5128 : DynamicZoneAbbrev *dtza;
5129 :
5130 : /* Just some sanity checks to prevent indexing off into nowhere */
5131 : Assert(tp->type == DYNTZ);
5132 : Assert(tp->value > 0 && tp->value < tbl->tblsize);
5133 :
5134 820 : dtza = (DynamicZoneAbbrev *) ((char *) tbl + tp->value);
5135 :
5136 : /* Look up the underlying zone if we haven't already */
5137 820 : if (dtza->tz == NULL)
5138 : {
5139 612 : dtza->tz = pg_tzset(dtza->zone);
5140 612 : if (dtza->tz == NULL)
5141 : {
5142 : /* Ooops, bogus zone name in config file entry */
5143 0 : extra->dtee_timezone = dtza->zone;
5144 0 : extra->dtee_abbrev = tp->token;
5145 : }
5146 : }
5147 820 : return dtza->tz;
5148 : }
5149 :
5150 :
5151 : /*
5152 : * This set-returning function reads all the time zone abbreviations
5153 : * defined by the IANA data for the current timezone setting,
5154 : * and returns a set of (abbrev, utc_offset, is_dst).
5155 : */
5156 : Datum
5157 168 : pg_timezone_abbrevs_zone(PG_FUNCTION_ARGS)
5158 : {
5159 : FuncCallContext *funcctx;
5160 : int *pindex;
5161 : Datum result;
5162 : HeapTuple tuple;
5163 : Datum values[3];
5164 168 : bool nulls[3] = {0};
5165 168 : TimestampTz now = GetCurrentTransactionStartTimestamp();
5166 168 : pg_time_t t = timestamptz_to_time_t(now);
5167 : const char *abbrev;
5168 : long int gmtoff;
5169 : int isdst;
5170 : struct pg_itm_in itm_in;
5171 : Interval *resInterval;
5172 :
5173 : /* stuff done only on the first call of the function */
5174 168 : if (SRF_IS_FIRSTCALL())
5175 : {
5176 : TupleDesc tupdesc;
5177 : MemoryContext oldcontext;
5178 :
5179 : /* create a function context for cross-call persistence */
5180 28 : funcctx = SRF_FIRSTCALL_INIT();
5181 :
5182 : /*
5183 : * switch to memory context appropriate for multiple function calls
5184 : */
5185 28 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
5186 :
5187 : /* allocate memory for user context */
5188 28 : pindex = palloc_object(int);
5189 28 : *pindex = 0;
5190 28 : funcctx->user_fctx = pindex;
5191 :
5192 28 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
5193 0 : elog(ERROR, "return type must be a row type");
5194 28 : funcctx->tuple_desc = tupdesc;
5195 :
5196 28 : MemoryContextSwitchTo(oldcontext);
5197 : }
5198 :
5199 : /* stuff done on every call of the function */
5200 168 : funcctx = SRF_PERCALL_SETUP();
5201 168 : pindex = (int *) funcctx->user_fctx;
5202 :
5203 168 : while ((abbrev = pg_get_next_timezone_abbrev(pindex,
5204 168 : session_timezone)) != NULL)
5205 : {
5206 : /* Ignore abbreviations that aren't all-alphabetic */
5207 140 : if (strspn(abbrev, "ABCDEFGHIJKLMNOPQRSTUVWXYZ") != strlen(abbrev))
5208 0 : continue;
5209 :
5210 : /* Determine the current meaning of the abbrev */
5211 140 : if (!pg_interpret_timezone_abbrev(abbrev,
5212 : &t,
5213 : &gmtoff,
5214 : &isdst,
5215 : session_timezone))
5216 0 : continue; /* hm, not actually used in this zone? */
5217 :
5218 140 : values[0] = CStringGetTextDatum(abbrev);
5219 :
5220 : /* Convert offset (in seconds) to an interval; can't overflow */
5221 560 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5222 140 : itm_in.tm_usec = (int64) gmtoff * USECS_PER_SEC;
5223 140 : resInterval = palloc_object(Interval);
5224 140 : (void) itmin2interval(&itm_in, resInterval);
5225 140 : values[1] = IntervalPGetDatum(resInterval);
5226 :
5227 140 : values[2] = BoolGetDatum(isdst);
5228 :
5229 140 : tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
5230 140 : result = HeapTupleGetDatum(tuple);
5231 :
5232 140 : SRF_RETURN_NEXT(funcctx, result);
5233 : }
5234 :
5235 28 : SRF_RETURN_DONE(funcctx);
5236 : }
5237 :
5238 : /*
5239 : * This set-returning function reads all the time zone abbreviations
5240 : * defined by the timezone_abbreviations setting,
5241 : * and returns a set of (abbrev, utc_offset, is_dst).
5242 : */
5243 : Datum
5244 3144 : pg_timezone_abbrevs_abbrevs(PG_FUNCTION_ARGS)
5245 : {
5246 : FuncCallContext *funcctx;
5247 : int *pindex;
5248 : Datum result;
5249 : HeapTuple tuple;
5250 : Datum values[3];
5251 3144 : bool nulls[3] = {0};
5252 : const datetkn *tp;
5253 : char buffer[TOKMAXLEN + 1];
5254 : int gmtoffset;
5255 : bool is_dst;
5256 : unsigned char *p;
5257 : struct pg_itm_in itm_in;
5258 : Interval *resInterval;
5259 :
5260 : /* stuff done only on the first call of the function */
5261 3144 : if (SRF_IS_FIRSTCALL())
5262 : {
5263 : TupleDesc tupdesc;
5264 : MemoryContext oldcontext;
5265 :
5266 : /* create a function context for cross-call persistence */
5267 16 : funcctx = SRF_FIRSTCALL_INIT();
5268 :
5269 : /*
5270 : * switch to memory context appropriate for multiple function calls
5271 : */
5272 16 : oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
5273 :
5274 : /* allocate memory for user context */
5275 16 : pindex = palloc_object(int);
5276 16 : *pindex = 0;
5277 16 : funcctx->user_fctx = pindex;
5278 :
5279 16 : if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
5280 0 : elog(ERROR, "return type must be a row type");
5281 16 : funcctx->tuple_desc = tupdesc;
5282 :
5283 16 : MemoryContextSwitchTo(oldcontext);
5284 : }
5285 :
5286 : /* stuff done on every call of the function */
5287 3144 : funcctx = SRF_PERCALL_SETUP();
5288 3144 : pindex = (int *) funcctx->user_fctx;
5289 :
5290 3144 : if (zoneabbrevtbl == NULL ||
5291 3144 : *pindex >= zoneabbrevtbl->numabbrevs)
5292 16 : SRF_RETURN_DONE(funcctx);
5293 :
5294 3128 : tp = zoneabbrevtbl->abbrevs + *pindex;
5295 :
5296 3128 : switch (tp->type)
5297 : {
5298 1564 : case TZ:
5299 1564 : gmtoffset = tp->value;
5300 1564 : is_dst = false;
5301 1564 : break;
5302 768 : case DTZ:
5303 768 : gmtoffset = tp->value;
5304 768 : is_dst = true;
5305 768 : break;
5306 796 : case DYNTZ:
5307 : {
5308 : /* Determine the current meaning of the abbrev */
5309 : pg_tz *tzp;
5310 : DateTimeErrorExtra extra;
5311 : TimestampTz now;
5312 : int isdst;
5313 :
5314 796 : tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp, &extra);
5315 796 : if (tzp == NULL)
5316 0 : DateTimeParseError(DTERR_BAD_ZONE_ABBREV, &extra,
5317 : NULL, NULL, NULL);
5318 796 : now = GetCurrentTransactionStartTimestamp();
5319 1592 : gmtoffset = -DetermineTimeZoneAbbrevOffsetTS(now,
5320 796 : tp->token,
5321 : tzp,
5322 : &isdst);
5323 796 : is_dst = (bool) isdst;
5324 796 : break;
5325 : }
5326 0 : default:
5327 0 : elog(ERROR, "unrecognized timezone type %d", (int) tp->type);
5328 : gmtoffset = 0; /* keep compiler quiet */
5329 : is_dst = false;
5330 : break;
5331 : }
5332 :
5333 : /*
5334 : * Convert name to text, using upcasing conversion that is the inverse of
5335 : * what ParseDateTime() uses.
5336 : */
5337 3128 : strlcpy(buffer, tp->token, sizeof(buffer));
5338 14496 : for (p = (unsigned char *) buffer; *p; p++)
5339 11368 : *p = pg_toupper(*p);
5340 :
5341 3128 : values[0] = CStringGetTextDatum(buffer);
5342 :
5343 : /* Convert offset (in seconds) to an interval; can't overflow */
5344 12512 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5345 3128 : itm_in.tm_usec = (int64) gmtoffset * USECS_PER_SEC;
5346 3128 : resInterval = palloc_object(Interval);
5347 3128 : (void) itmin2interval(&itm_in, resInterval);
5348 3128 : values[1] = IntervalPGetDatum(resInterval);
5349 :
5350 3128 : values[2] = BoolGetDatum(is_dst);
5351 :
5352 3128 : (*pindex)++;
5353 :
5354 3128 : tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
5355 3128 : result = HeapTupleGetDatum(tuple);
5356 :
5357 3128 : SRF_RETURN_NEXT(funcctx, result);
5358 : }
5359 :
5360 : /*
5361 : * This set-returning function reads all the available full time zones
5362 : * and returns a set of (name, abbrev, utc_offset, is_dst).
5363 : */
5364 : Datum
5365 10 : pg_timezone_names(PG_FUNCTION_ARGS)
5366 : {
5367 10 : ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
5368 : pg_tzenum *tzenum;
5369 : pg_tz *tz;
5370 : Datum values[4];
5371 10 : bool nulls[4] = {0};
5372 : int tzoff;
5373 : struct pg_tm tm;
5374 : fsec_t fsec;
5375 : const char *tzn;
5376 : Interval *resInterval;
5377 : struct pg_itm_in itm_in;
5378 :
5379 10 : InitMaterializedSRF(fcinfo, 0);
5380 :
5381 : /* initialize timezone scanning code */
5382 10 : tzenum = pg_tzenumerate_start();
5383 :
5384 : /* search for another zone to display */
5385 : for (;;)
5386 : {
5387 5990 : tz = pg_tzenumerate_next(tzenum);
5388 5990 : if (!tz)
5389 10 : break;
5390 :
5391 : /* Convert now() to local time in this zone */
5392 5980 : if (timestamp2tm(GetCurrentTransactionStartTimestamp(),
5393 : &tzoff, &tm, &fsec, &tzn, tz) != 0)
5394 0 : continue; /* ignore if conversion fails */
5395 :
5396 : /*
5397 : * IANA's rather silly "Factory" time zone used to emit ridiculously
5398 : * long "abbreviations" such as "Local time zone must be set--see zic
5399 : * manual page" or "Local time zone must be set--use tzsetup". While
5400 : * modern versions of tzdb emit the much saner "-00", it seems some
5401 : * benighted packagers are hacking the IANA data so that it continues
5402 : * to produce these strings. To prevent producing a weirdly wide
5403 : * abbrev column, reject ridiculously long abbreviations.
5404 : */
5405 5980 : if (tzn && strlen(tzn) > 31)
5406 0 : continue;
5407 :
5408 5980 : values[0] = CStringGetTextDatum(pg_get_timezone_name(tz));
5409 5980 : values[1] = CStringGetTextDatum(tzn ? tzn : "");
5410 :
5411 : /* Convert tzoff to an interval; can't overflow */
5412 23920 : MemSet(&itm_in, 0, sizeof(struct pg_itm_in));
5413 5980 : itm_in.tm_usec = (int64) -tzoff * USECS_PER_SEC;
5414 5980 : resInterval = palloc_object(Interval);
5415 5980 : (void) itmin2interval(&itm_in, resInterval);
5416 5980 : values[2] = IntervalPGetDatum(resInterval);
5417 :
5418 5980 : values[3] = BoolGetDatum(tm.tm_isdst > 0);
5419 :
5420 5980 : tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls);
5421 : }
5422 :
5423 10 : pg_tzenumerate_end(tzenum);
5424 10 : return (Datum) 0;
5425 : }
|
