1 /*-------------------------------------------------------------------------
2 *
3 * datetime.c
4 * Support functions for date/time types.
5 *
6 * Portions Copyright (c) 1996-2019, 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 "catalog/pg_type.h"
24 #include "common/string.h"
25 #include "funcapi.h"
26 #include "miscadmin.h"
27 #include "nodes/nodeFuncs.h"
28 #include "utils/builtins.h"
29 #include "utils/date.h"
30 #include "utils/datetime.h"
31 #include "utils/memutils.h"
32 #include "utils/tzparser.h"
33
34
35 static int DecodeNumber(int flen, char *field, bool haveTextMonth,
36 int fmask, int *tmask,
37 struct pg_tm *tm, fsec_t *fsec, bool *is2digits);
38 static int DecodeNumberField(int len, char *str,
39 int fmask, int *tmask,
40 struct pg_tm *tm, fsec_t *fsec, bool *is2digits);
41 static int DecodeTime(char *str, int fmask, int range,
42 int *tmask, struct pg_tm *tm, fsec_t *fsec);
43 static const datetkn *datebsearch(const char *key, const datetkn *base, int nel);
44 static int DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
45 struct pg_tm *tm);
46 static char *AppendSeconds(char *cp, int sec, fsec_t fsec,
47 int precision, bool fillzeros);
48 static void AdjustFractSeconds(double frac, struct pg_tm *tm, fsec_t *fsec,
49 int scale);
50 static void AdjustFractDays(double frac, struct pg_tm *tm, fsec_t *fsec,
51 int scale);
52 static int DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp,
53 pg_time_t *tp);
54 static bool DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t,
55 const char *abbr, pg_tz *tzp,
56 int *offset, int *isdst);
57 static pg_tz *FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp);
58
59
60 const int day_tab[2][13] =
61 {
62 {31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0},
63 {31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0}
64 };
65
66 const char *const months[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun",
67 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", NULL};
68
69 const char *const days[] = {"Sunday", "Monday", "Tuesday", "Wednesday",
70 "Thursday", "Friday", "Saturday", NULL};
71
72
73 /*****************************************************************************
74 * PRIVATE ROUTINES *
75 *****************************************************************************/
76
77 /*
78 * datetktbl holds date/time keywords.
79 *
80 * Note that this table must be strictly alphabetically ordered to allow an
81 * O(ln(N)) search algorithm to be used.
82 *
83 * The token field must be NUL-terminated; we truncate entries to TOKMAXLEN
84 * characters to fit.
85 *
86 * The static table contains no TZ, DTZ, or DYNTZ entries; rather those
87 * are loaded from configuration files and stored in zoneabbrevtbl, whose
88 * abbrevs[] field has the same format as the static datetktbl.
89 */
90 static const datetkn datetktbl[] = {
91 /* token, type, value */
92 {EARLY, RESERV, DTK_EARLY}, /* "-infinity" reserved for "early time" */
93 {DA_D, ADBC, AD}, /* "ad" for years > 0 */
94 {"allballs", RESERV, DTK_ZULU}, /* 00:00:00 */
95 {"am", AMPM, AM},
96 {"apr", MONTH, 4},
97 {"april", MONTH, 4},
98 {"at", IGNORE_DTF, 0}, /* "at" (throwaway) */
99 {"aug", MONTH, 8},
100 {"august", MONTH, 8},
101 {DB_C, ADBC, BC}, /* "bc" for years <= 0 */
102 {"d", UNITS, DTK_DAY}, /* "day of month" for ISO input */
103 {"dec", MONTH, 12},
104 {"december", MONTH, 12},
105 {"dow", UNITS, DTK_DOW}, /* day of week */
106 {"doy", UNITS, DTK_DOY}, /* day of year */
107 {"dst", DTZMOD, SECS_PER_HOUR},
108 {EPOCH, RESERV, DTK_EPOCH}, /* "epoch" reserved for system epoch time */
109 {"feb", MONTH, 2},
110 {"february", MONTH, 2},
111 {"fri", DOW, 5},
112 {"friday", DOW, 5},
113 {"h", UNITS, DTK_HOUR}, /* "hour" */
114 {LATE, RESERV, DTK_LATE}, /* "infinity" reserved for "late time" */
115 {"isodow", UNITS, DTK_ISODOW}, /* ISO day of week, Sunday == 7 */
116 {"isoyear", UNITS, DTK_ISOYEAR}, /* year in terms of the ISO week date */
117 {"j", UNITS, DTK_JULIAN},
118 {"jan", MONTH, 1},
119 {"january", MONTH, 1},
120 {"jd", UNITS, DTK_JULIAN},
121 {"jul", MONTH, 7},
122 {"julian", UNITS, DTK_JULIAN},
123 {"july", MONTH, 7},
124 {"jun", MONTH, 6},
125 {"june", MONTH, 6},
126 {"m", UNITS, DTK_MONTH}, /* "month" for ISO input */
127 {"mar", MONTH, 3},
128 {"march", MONTH, 3},
129 {"may", MONTH, 5},
130 {"mm", UNITS, DTK_MINUTE}, /* "minute" for ISO input */
131 {"mon", DOW, 1},
132 {"monday", DOW, 1},
133 {"nov", MONTH, 11},
134 {"november", MONTH, 11},
135 {NOW, RESERV, DTK_NOW}, /* current transaction time */
136 {"oct", MONTH, 10},
137 {"october", MONTH, 10},
138 {"on", IGNORE_DTF, 0}, /* "on" (throwaway) */
139 {"pm", AMPM, PM},
140 {"s", UNITS, DTK_SECOND}, /* "seconds" for ISO input */
141 {"sat", DOW, 6},
142 {"saturday", DOW, 6},
143 {"sep", MONTH, 9},
144 {"sept", MONTH, 9},
145 {"september", MONTH, 9},
146 {"sun", DOW, 0},
147 {"sunday", DOW, 0},
148 {"t", ISOTIME, DTK_TIME}, /* Filler for ISO time fields */
149 {"thu", DOW, 4},
150 {"thur", DOW, 4},
151 {"thurs", DOW, 4},
152 {"thursday", DOW, 4},
153 {TODAY, RESERV, DTK_TODAY}, /* midnight */
154 {TOMORROW, RESERV, DTK_TOMORROW}, /* tomorrow midnight */
155 {"tue", DOW, 2},
156 {"tues", DOW, 2},
157 {"tuesday", DOW, 2},
158 {"wed", DOW, 3},
159 {"wednesday", DOW, 3},
160 {"weds", DOW, 3},
161 {"y", UNITS, DTK_YEAR}, /* "year" for ISO input */
162 {YESTERDAY, RESERV, DTK_YESTERDAY} /* yesterday midnight */
163 };
164
165 static const int szdatetktbl = sizeof datetktbl / sizeof datetktbl[0];
166
167 /*
168 * deltatktbl: same format as datetktbl, but holds keywords used to represent
169 * time units (eg, for intervals, and for EXTRACT).
170 */
171 static const datetkn deltatktbl[] = {
172 /* token, type, value */
173 {"@", IGNORE_DTF, 0}, /* postgres relative prefix */
174 {DAGO, AGO, 0}, /* "ago" indicates negative time offset */
175 {"c", UNITS, DTK_CENTURY}, /* "century" relative */
176 {"cent", UNITS, DTK_CENTURY}, /* "century" relative */
177 {"centuries", UNITS, DTK_CENTURY}, /* "centuries" relative */
178 {DCENTURY, UNITS, DTK_CENTURY}, /* "century" relative */
179 {"d", UNITS, DTK_DAY}, /* "day" relative */
180 {DDAY, UNITS, DTK_DAY}, /* "day" relative */
181 {"days", UNITS, DTK_DAY}, /* "days" relative */
182 {"dec", UNITS, DTK_DECADE}, /* "decade" relative */
183 {DDECADE, UNITS, DTK_DECADE}, /* "decade" relative */
184 {"decades", UNITS, DTK_DECADE}, /* "decades" relative */
185 {"decs", UNITS, DTK_DECADE}, /* "decades" relative */
186 {"h", UNITS, DTK_HOUR}, /* "hour" relative */
187 {DHOUR, UNITS, DTK_HOUR}, /* "hour" relative */
188 {"hours", UNITS, DTK_HOUR}, /* "hours" relative */
189 {"hr", UNITS, DTK_HOUR}, /* "hour" relative */
190 {"hrs", UNITS, DTK_HOUR}, /* "hours" relative */
191 {"m", UNITS, DTK_MINUTE}, /* "minute" relative */
192 {"microsecon", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
193 {"mil", UNITS, DTK_MILLENNIUM}, /* "millennium" relative */
194 {"millennia", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */
195 {DMILLENNIUM, UNITS, DTK_MILLENNIUM}, /* "millennium" relative */
196 {"millisecon", UNITS, DTK_MILLISEC}, /* relative */
197 {"mils", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */
198 {"min", UNITS, DTK_MINUTE}, /* "minute" relative */
199 {"mins", UNITS, DTK_MINUTE}, /* "minutes" relative */
200 {DMINUTE, UNITS, DTK_MINUTE}, /* "minute" relative */
201 {"minutes", UNITS, DTK_MINUTE}, /* "minutes" relative */
202 {"mon", UNITS, DTK_MONTH}, /* "months" relative */
203 {"mons", UNITS, DTK_MONTH}, /* "months" relative */
204 {DMONTH, UNITS, DTK_MONTH}, /* "month" relative */
205 {"months", UNITS, DTK_MONTH},
206 {"ms", UNITS, DTK_MILLISEC},
207 {"msec", UNITS, DTK_MILLISEC},
208 {DMILLISEC, UNITS, DTK_MILLISEC},
209 {"mseconds", UNITS, DTK_MILLISEC},
210 {"msecs", UNITS, DTK_MILLISEC},
211 {"qtr", UNITS, DTK_QUARTER}, /* "quarter" relative */
212 {DQUARTER, UNITS, DTK_QUARTER}, /* "quarter" relative */
213 {"s", UNITS, DTK_SECOND},
214 {"sec", UNITS, DTK_SECOND},
215 {DSECOND, UNITS, DTK_SECOND},
216 {"seconds", UNITS, DTK_SECOND},
217 {"secs", UNITS, DTK_SECOND},
218 {DTIMEZONE, UNITS, DTK_TZ}, /* "timezone" time offset */
219 {"timezone_h", UNITS, DTK_TZ_HOUR}, /* timezone hour units */
220 {"timezone_m", UNITS, DTK_TZ_MINUTE}, /* timezone minutes units */
221 {"us", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
222 {"usec", UNITS, DTK_MICROSEC}, /* "microsecond" relative */
223 {DMICROSEC, UNITS, DTK_MICROSEC}, /* "microsecond" relative */
224 {"useconds", UNITS, DTK_MICROSEC}, /* "microseconds" relative */
225 {"usecs", UNITS, DTK_MICROSEC}, /* "microseconds" relative */
226 {"w", UNITS, DTK_WEEK}, /* "week" relative */
227 {DWEEK, UNITS, DTK_WEEK}, /* "week" relative */
228 {"weeks", UNITS, DTK_WEEK}, /* "weeks" relative */
229 {"y", UNITS, DTK_YEAR}, /* "year" relative */
230 {DYEAR, UNITS, DTK_YEAR}, /* "year" relative */
231 {"years", UNITS, DTK_YEAR}, /* "years" relative */
232 {"yr", UNITS, DTK_YEAR}, /* "year" relative */
233 {"yrs", UNITS, DTK_YEAR} /* "years" relative */
234 };
235
236 static const int szdeltatktbl = sizeof deltatktbl / sizeof deltatktbl[0];
237
238 static TimeZoneAbbrevTable *zoneabbrevtbl = NULL;
239
240 /* Caches of recent lookup results in the above tables */
241
242 static const datetkn *datecache[MAXDATEFIELDS] = {NULL};
243
244 static const datetkn *deltacache[MAXDATEFIELDS] = {NULL};
245
246 static const datetkn *abbrevcache[MAXDATEFIELDS] = {NULL};
247
248
249 /*
250 * Calendar time to Julian date conversions.
251 * Julian date is commonly used in astronomical applications,
252 * since it is numerically accurate and computationally simple.
253 * The algorithms here will accurately convert between Julian day
254 * and calendar date for all non-negative Julian days
255 * (i.e. from Nov 24, -4713 on).
256 *
257 * Rewritten to eliminate overflow problems. This now allows the
258 * routines to work correctly for all Julian day counts from
259 * 0 to 2147483647 (Nov 24, -4713 to Jun 3, 5874898) assuming
260 * a 32-bit integer. Longer types should also work to the limits
261 * of their precision.
262 *
263 * Actually, date2j() will work sanely, in the sense of producing
264 * valid negative Julian dates, significantly before Nov 24, -4713.
265 * We rely on it to do so back to Nov 1, -4713; see IS_VALID_JULIAN()
266 * and associated commentary in timestamp.h.
267 */
268
269 int
date2j(int y,int m,int d)270 date2j(int y, int m, int d)
271 {
272 int julian;
273 int century;
274
275 if (m > 2)
276 {
277 m += 1;
278 y += 4800;
279 }
280 else
281 {
282 m += 13;
283 y += 4799;
284 }
285
286 century = y / 100;
287 julian = y * 365 - 32167;
288 julian += y / 4 - century + century / 4;
289 julian += 7834 * m / 256 + d;
290
291 return julian;
292 } /* date2j() */
293
294 void
j2date(int jd,int * year,int * month,int * day)295 j2date(int jd, int *year, int *month, int *day)
296 {
297 unsigned int julian;
298 unsigned int quad;
299 unsigned int extra;
300 int y;
301
302 julian = jd;
303 julian += 32044;
304 quad = julian / 146097;
305 extra = (julian - quad * 146097) * 4 + 3;
306 julian += 60 + quad * 3 + extra / 146097;
307 quad = julian / 1461;
308 julian -= quad * 1461;
309 y = julian * 4 / 1461;
310 julian = ((y != 0) ? ((julian + 305) % 365) : ((julian + 306) % 366))
311 + 123;
312 y += quad * 4;
313 *year = y - 4800;
314 quad = julian * 2141 / 65536;
315 *day = julian - 7834 * quad / 256;
316 *month = (quad + 10) % MONTHS_PER_YEAR + 1;
317
318 return;
319 } /* j2date() */
320
321
322 /*
323 * j2day - convert Julian date to day-of-week (0..6 == Sun..Sat)
324 *
325 * Note: various places use the locution j2day(date - 1) to produce a
326 * result according to the convention 0..6 = Mon..Sun. This is a bit of
327 * a crock, but will work as long as the computation here is just a modulo.
328 */
329 int
j2day(int date)330 j2day(int date)
331 {
332 date += 1;
333 date %= 7;
334 /* Cope if division truncates towards zero, as it probably does */
335 if (date < 0)
336 date += 7;
337
338 return date;
339 } /* j2day() */
340
341
342 /*
343 * GetCurrentDateTime()
344 *
345 * Get the transaction start time ("now()") broken down as a struct pg_tm.
346 */
347 void
GetCurrentDateTime(struct pg_tm * tm)348 GetCurrentDateTime(struct pg_tm *tm)
349 {
350 int tz;
351 fsec_t fsec;
352
353 timestamp2tm(GetCurrentTransactionStartTimestamp(), &tz, tm, &fsec,
354 NULL, NULL);
355 /* Note: don't pass NULL tzp to timestamp2tm; affects behavior */
356 }
357
358 /*
359 * GetCurrentTimeUsec()
360 *
361 * Get the transaction start time ("now()") broken down as a struct pg_tm,
362 * including fractional seconds and timezone offset.
363 */
364 void
GetCurrentTimeUsec(struct pg_tm * tm,fsec_t * fsec,int * tzp)365 GetCurrentTimeUsec(struct pg_tm *tm, fsec_t *fsec, int *tzp)
366 {
367 int tz;
368
369 timestamp2tm(GetCurrentTransactionStartTimestamp(), &tz, tm, fsec,
370 NULL, NULL);
371 /* Note: don't pass NULL tzp to timestamp2tm; affects behavior */
372 if (tzp != NULL)
373 *tzp = tz;
374 }
375
376
377 /*
378 * Append seconds and fractional seconds (if any) at *cp.
379 *
380 * precision is the max number of fraction digits, fillzeros says to
381 * pad to two integral-seconds digits.
382 *
383 * Returns a pointer to the new end of string. No NUL terminator is put
384 * there; callers are responsible for NUL terminating str themselves.
385 *
386 * Note that any sign is stripped from the input seconds values.
387 */
388 static char *
AppendSeconds(char * cp,int sec,fsec_t fsec,int precision,bool fillzeros)389 AppendSeconds(char *cp, int sec, fsec_t fsec, int precision, bool fillzeros)
390 {
391 Assert(precision >= 0);
392
393 if (fillzeros)
394 cp = pg_ltostr_zeropad(cp, Abs(sec), 2);
395 else
396 cp = pg_ltostr(cp, Abs(sec));
397
398 /* fsec_t is just an int32 */
399 if (fsec != 0)
400 {
401 int32 value = Abs(fsec);
402 char *end = &cp[precision + 1];
403 bool gotnonzero = false;
404
405 *cp++ = '.';
406
407 /*
408 * Append the fractional seconds part. Note that we don't want any
409 * trailing zeros here, so since we're building the number in reverse
410 * we'll skip appending zeros until we've output a non-zero digit.
411 */
412 while (precision--)
413 {
414 int32 oldval = value;
415 int32 remainder;
416
417 value /= 10;
418 remainder = oldval - value * 10;
419
420 /* check if we got a non-zero */
421 if (remainder)
422 gotnonzero = true;
423
424 if (gotnonzero)
425 cp[precision] = '0' + remainder;
426 else
427 end = &cp[precision];
428 }
429
430 /*
431 * If we still have a non-zero value then precision must have not been
432 * enough to print the number. We punt the problem to pg_ltostr(),
433 * which will generate a correct answer in the minimum valid width.
434 */
435 if (value)
436 return pg_ltostr(cp, Abs(fsec));
437
438 return end;
439 }
440 else
441 return cp;
442 }
443
444
445 /*
446 * Variant of above that's specialized to timestamp case.
447 *
448 * Returns a pointer to the new end of string. No NUL terminator is put
449 * there; callers are responsible for NUL terminating str themselves.
450 */
451 static char *
AppendTimestampSeconds(char * cp,struct pg_tm * tm,fsec_t fsec)452 AppendTimestampSeconds(char *cp, struct pg_tm *tm, fsec_t fsec)
453 {
454 return AppendSeconds(cp, tm->tm_sec, fsec, MAX_TIMESTAMP_PRECISION, true);
455 }
456
457 /*
458 * Multiply frac by scale (to produce seconds) and add to *tm & *fsec.
459 * We assume the input frac is less than 1 so overflow is not an issue.
460 */
461 static void
AdjustFractSeconds(double frac,struct pg_tm * tm,fsec_t * fsec,int scale)462 AdjustFractSeconds(double frac, struct pg_tm *tm, fsec_t *fsec, int scale)
463 {
464 int sec;
465
466 if (frac == 0)
467 return;
468 frac *= scale;
469 sec = (int) frac;
470 tm->tm_sec += sec;
471 frac -= sec;
472 *fsec += rint(frac * 1000000);
473 }
474
475 /* As above, but initial scale produces days */
476 static void
AdjustFractDays(double frac,struct pg_tm * tm,fsec_t * fsec,int scale)477 AdjustFractDays(double frac, struct pg_tm *tm, fsec_t *fsec, int scale)
478 {
479 int extra_days;
480
481 if (frac == 0)
482 return;
483 frac *= scale;
484 extra_days = (int) frac;
485 tm->tm_mday += extra_days;
486 frac -= extra_days;
487 AdjustFractSeconds(frac, tm, fsec, SECS_PER_DAY);
488 }
489
490 /* Fetch a fractional-second value with suitable error checking */
491 static int
ParseFractionalSecond(char * cp,fsec_t * fsec)492 ParseFractionalSecond(char *cp, fsec_t *fsec)
493 {
494 double frac;
495
496 /* Caller should always pass the start of the fraction part */
497 Assert(*cp == '.');
498 errno = 0;
499 frac = strtod(cp, &cp);
500 /* check for parse failure */
501 if (*cp != '\0' || errno != 0)
502 return DTERR_BAD_FORMAT;
503 *fsec = rint(frac * 1000000);
504 return 0;
505 }
506
507
508 /* ParseDateTime()
509 * Break string into tokens based on a date/time context.
510 * Returns 0 if successful, DTERR code if bogus input detected.
511 *
512 * timestr - the input string
513 * workbuf - workspace for field string storage. This must be
514 * larger than the largest legal input for this datetime type --
515 * some additional space will be needed to NUL terminate fields.
516 * buflen - the size of workbuf
517 * field[] - pointers to field strings are returned in this array
518 * ftype[] - field type indicators are returned in this array
519 * maxfields - dimensions of the above two arrays
520 * *numfields - set to the actual number of fields detected
521 *
522 * The fields extracted from the input are stored as separate,
523 * null-terminated strings in the workspace at workbuf. Any text is
524 * converted to lower case.
525 *
526 * Several field types are assigned:
527 * DTK_NUMBER - digits and (possibly) a decimal point
528 * DTK_DATE - digits and two delimiters, or digits and text
529 * DTK_TIME - digits, colon delimiters, and possibly a decimal point
530 * DTK_STRING - text (no digits or punctuation)
531 * DTK_SPECIAL - leading "+" or "-" followed by text
532 * DTK_TZ - leading "+" or "-" followed by digits (also eats ':', '.', '-')
533 *
534 * Note that some field types can hold unexpected items:
535 * DTK_NUMBER can hold date fields (yy.ddd)
536 * DTK_STRING can hold months (January) and time zones (PST)
537 * DTK_DATE can hold time zone names (America/New_York, GMT-8)
538 */
539 int
ParseDateTime(const char * timestr,char * workbuf,size_t buflen,char ** field,int * ftype,int maxfields,int * numfields)540 ParseDateTime(const char *timestr, char *workbuf, size_t buflen,
541 char **field, int *ftype, int maxfields, int *numfields)
542 {
543 int nf = 0;
544 const char *cp = timestr;
545 char *bufp = workbuf;
546 const char *bufend = workbuf + buflen;
547
548 /*
549 * Set the character pointed-to by "bufptr" to "newchar", and increment
550 * "bufptr". "end" gives the end of the buffer -- we return an error if
551 * there is no space left to append a character to the buffer. Note that
552 * "bufptr" is evaluated twice.
553 */
554 #define APPEND_CHAR(bufptr, end, newchar) \
555 do \
556 { \
557 if (((bufptr) + 1) >= (end)) \
558 return DTERR_BAD_FORMAT; \
559 *(bufptr)++ = newchar; \
560 } while (0)
561
562 /* outer loop through fields */
563 while (*cp != '\0')
564 {
565 /* Ignore spaces between fields */
566 if (isspace((unsigned char) *cp))
567 {
568 cp++;
569 continue;
570 }
571
572 /* Record start of current field */
573 if (nf >= maxfields)
574 return DTERR_BAD_FORMAT;
575 field[nf] = bufp;
576
577 /* leading digit? then date or time */
578 if (isdigit((unsigned char) *cp))
579 {
580 APPEND_CHAR(bufp, bufend, *cp++);
581 while (isdigit((unsigned char) *cp))
582 APPEND_CHAR(bufp, bufend, *cp++);
583
584 /* time field? */
585 if (*cp == ':')
586 {
587 ftype[nf] = DTK_TIME;
588 APPEND_CHAR(bufp, bufend, *cp++);
589 while (isdigit((unsigned char) *cp) ||
590 (*cp == ':') || (*cp == '.'))
591 APPEND_CHAR(bufp, bufend, *cp++);
592 }
593 /* date field? allow embedded text month */
594 else if (*cp == '-' || *cp == '/' || *cp == '.')
595 {
596 /* save delimiting character to use later */
597 char delim = *cp;
598
599 APPEND_CHAR(bufp, bufend, *cp++);
600 /* second field is all digits? then no embedded text month */
601 if (isdigit((unsigned char) *cp))
602 {
603 ftype[nf] = ((delim == '.') ? DTK_NUMBER : DTK_DATE);
604 while (isdigit((unsigned char) *cp))
605 APPEND_CHAR(bufp, bufend, *cp++);
606
607 /*
608 * insist that the delimiters match to get a three-field
609 * date.
610 */
611 if (*cp == delim)
612 {
613 ftype[nf] = DTK_DATE;
614 APPEND_CHAR(bufp, bufend, *cp++);
615 while (isdigit((unsigned char) *cp) || *cp == delim)
616 APPEND_CHAR(bufp, bufend, *cp++);
617 }
618 }
619 else
620 {
621 ftype[nf] = DTK_DATE;
622 while (isalnum((unsigned char) *cp) || *cp == delim)
623 APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
624 }
625 }
626
627 /*
628 * otherwise, number only and will determine year, month, day, or
629 * concatenated fields later...
630 */
631 else
632 ftype[nf] = DTK_NUMBER;
633 }
634 /* Leading decimal point? Then fractional seconds... */
635 else if (*cp == '.')
636 {
637 APPEND_CHAR(bufp, bufend, *cp++);
638 while (isdigit((unsigned char) *cp))
639 APPEND_CHAR(bufp, bufend, *cp++);
640
641 ftype[nf] = DTK_NUMBER;
642 }
643
644 /*
645 * text? then date string, month, day of week, special, or timezone
646 */
647 else if (isalpha((unsigned char) *cp))
648 {
649 bool is_date;
650
651 ftype[nf] = DTK_STRING;
652 APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
653 while (isalpha((unsigned char) *cp))
654 APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
655
656 /*
657 * Dates can have embedded '-', '/', or '.' separators. It could
658 * also be a timezone name containing embedded '/', '+', '-', '_',
659 * or ':' (but '_' or ':' can't be the first punctuation). If the
660 * next character is a digit or '+', we need to check whether what
661 * we have so far is a recognized non-timezone keyword --- if so,
662 * don't believe that this is the start of a timezone.
663 */
664 is_date = false;
665 if (*cp == '-' || *cp == '/' || *cp == '.')
666 is_date = true;
667 else if (*cp == '+' || isdigit((unsigned char) *cp))
668 {
669 *bufp = '\0'; /* null-terminate current field value */
670 /* we need search only the core token table, not TZ names */
671 if (datebsearch(field[nf], datetktbl, szdatetktbl) == NULL)
672 is_date = true;
673 }
674 if (is_date)
675 {
676 ftype[nf] = DTK_DATE;
677 do
678 {
679 APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
680 } while (*cp == '+' || *cp == '-' ||
681 *cp == '/' || *cp == '_' ||
682 *cp == '.' || *cp == ':' ||
683 isalnum((unsigned char) *cp));
684 }
685 }
686 /* sign? then special or numeric timezone */
687 else if (*cp == '+' || *cp == '-')
688 {
689 APPEND_CHAR(bufp, bufend, *cp++);
690 /* soak up leading whitespace */
691 while (isspace((unsigned char) *cp))
692 cp++;
693 /* numeric timezone? */
694 /* note that "DTK_TZ" could also be a signed float or yyyy-mm */
695 if (isdigit((unsigned char) *cp))
696 {
697 ftype[nf] = DTK_TZ;
698 APPEND_CHAR(bufp, bufend, *cp++);
699 while (isdigit((unsigned char) *cp) ||
700 *cp == ':' || *cp == '.' || *cp == '-')
701 APPEND_CHAR(bufp, bufend, *cp++);
702 }
703 /* special? */
704 else if (isalpha((unsigned char) *cp))
705 {
706 ftype[nf] = DTK_SPECIAL;
707 APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
708 while (isalpha((unsigned char) *cp))
709 APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++));
710 }
711 /* otherwise something wrong... */
712 else
713 return DTERR_BAD_FORMAT;
714 }
715 /* ignore other punctuation but use as delimiter */
716 else if (ispunct((unsigned char) *cp))
717 {
718 cp++;
719 continue;
720 }
721 /* otherwise, something is not right... */
722 else
723 return DTERR_BAD_FORMAT;
724
725 /* force in a delimiter after each field */
726 *bufp++ = '\0';
727 nf++;
728 }
729
730 *numfields = nf;
731
732 return 0;
733 }
734
735
736 /* DecodeDateTime()
737 * Interpret previously parsed fields for general date and time.
738 * Return 0 if full date, 1 if only time, and negative DTERR code if problems.
739 * (Currently, all callers treat 1 as an error return too.)
740 *
741 * External format(s):
742 * "<weekday> <month>-<day>-<year> <hour>:<minute>:<second>"
743 * "Fri Feb-7-1997 15:23:27"
744 * "Feb-7-1997 15:23:27"
745 * "2-7-1997 15:23:27"
746 * "1997-2-7 15:23:27"
747 * "1997.038 15:23:27" (day of year 1-366)
748 * Also supports input in compact time:
749 * "970207 152327"
750 * "97038 152327"
751 * "20011225T040506.789-07"
752 *
753 * Use the system-provided functions to get the current time zone
754 * if not specified in the input string.
755 *
756 * If the date is outside the range of pg_time_t (in practice that could only
757 * happen if pg_time_t is just 32 bits), then assume UTC time zone - thomas
758 * 1997-05-27
759 */
760 int
DecodeDateTime(char ** field,int * ftype,int nf,int * dtype,struct pg_tm * tm,fsec_t * fsec,int * tzp)761 DecodeDateTime(char **field, int *ftype, int nf,
762 int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp)
763 {
764 int fmask = 0,
765 tmask,
766 type;
767 int ptype = 0; /* "prefix type" for ISO y2001m02d04 format */
768 int i;
769 int val;
770 int dterr;
771 int mer = HR24;
772 bool haveTextMonth = false;
773 bool isjulian = false;
774 bool is2digits = false;
775 bool bc = false;
776 pg_tz *namedTz = NULL;
777 pg_tz *abbrevTz = NULL;
778 pg_tz *valtz;
779 char *abbrev = NULL;
780 struct pg_tm cur_tm;
781
782 /*
783 * We'll insist on at least all of the date fields, but initialize the
784 * remaining fields in case they are not set later...
785 */
786 *dtype = DTK_DATE;
787 tm->tm_hour = 0;
788 tm->tm_min = 0;
789 tm->tm_sec = 0;
790 *fsec = 0;
791 /* don't know daylight savings time status apriori */
792 tm->tm_isdst = -1;
793 if (tzp != NULL)
794 *tzp = 0;
795
796 for (i = 0; i < nf; i++)
797 {
798 switch (ftype[i])
799 {
800 case DTK_DATE:
801
802 /*
803 * Integral julian day with attached time zone? All other
804 * forms with JD will be separated into distinct fields, so we
805 * handle just this case here.
806 */
807 if (ptype == DTK_JULIAN)
808 {
809 char *cp;
810 int val;
811
812 if (tzp == NULL)
813 return DTERR_BAD_FORMAT;
814
815 errno = 0;
816 val = strtoint(field[i], &cp, 10);
817 if (errno == ERANGE || val < 0)
818 return DTERR_FIELD_OVERFLOW;
819
820 j2date(val, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
821 isjulian = true;
822
823 /* Get the time zone from the end of the string */
824 dterr = DecodeTimezone(cp, tzp);
825 if (dterr)
826 return dterr;
827
828 tmask = DTK_DATE_M | DTK_TIME_M | DTK_M(TZ);
829 ptype = 0;
830 break;
831 }
832
833 /*
834 * Already have a date? Then this might be a time zone name
835 * with embedded punctuation (e.g. "America/New_York") or a
836 * run-together time with trailing time zone (e.g. hhmmss-zz).
837 * - thomas 2001-12-25
838 *
839 * We consider it a time zone if we already have month & day.
840 * This is to allow the form "mmm dd hhmmss tz year", which
841 * we've historically accepted.
842 */
843 else if (ptype != 0 ||
844 ((fmask & (DTK_M(MONTH) | DTK_M(DAY))) ==
845 (DTK_M(MONTH) | DTK_M(DAY))))
846 {
847 /* No time zone accepted? Then quit... */
848 if (tzp == NULL)
849 return DTERR_BAD_FORMAT;
850
851 if (isdigit((unsigned char) *field[i]) || ptype != 0)
852 {
853 char *cp;
854
855 if (ptype != 0)
856 {
857 /* Sanity check; should not fail this test */
858 if (ptype != DTK_TIME)
859 return DTERR_BAD_FORMAT;
860 ptype = 0;
861 }
862
863 /*
864 * Starts with a digit but we already have a time
865 * field? Then we are in trouble with a date and time
866 * already...
867 */
868 if ((fmask & DTK_TIME_M) == DTK_TIME_M)
869 return DTERR_BAD_FORMAT;
870
871 if ((cp = strchr(field[i], '-')) == NULL)
872 return DTERR_BAD_FORMAT;
873
874 /* Get the time zone from the end of the string */
875 dterr = DecodeTimezone(cp, tzp);
876 if (dterr)
877 return dterr;
878 *cp = '\0';
879
880 /*
881 * Then read the rest of the field as a concatenated
882 * time
883 */
884 dterr = DecodeNumberField(strlen(field[i]), field[i],
885 fmask,
886 &tmask, tm,
887 fsec, &is2digits);
888 if (dterr < 0)
889 return dterr;
890
891 /*
892 * modify tmask after returning from
893 * DecodeNumberField()
894 */
895 tmask |= DTK_M(TZ);
896 }
897 else
898 {
899 namedTz = pg_tzset(field[i]);
900 if (!namedTz)
901 {
902 /*
903 * We should return an error code instead of
904 * ereport'ing directly, but then there is no way
905 * to report the bad time zone name.
906 */
907 ereport(ERROR,
908 (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
909 errmsg("time zone \"%s\" not recognized",
910 field[i])));
911 }
912 /* we'll apply the zone setting below */
913 tmask = DTK_M(TZ);
914 }
915 }
916 else
917 {
918 dterr = DecodeDate(field[i], fmask,
919 &tmask, &is2digits, tm);
920 if (dterr)
921 return dterr;
922 }
923 break;
924
925 case DTK_TIME:
926
927 /*
928 * This might be an ISO time following a "t" field.
929 */
930 if (ptype != 0)
931 {
932 /* Sanity check; should not fail this test */
933 if (ptype != DTK_TIME)
934 return DTERR_BAD_FORMAT;
935 ptype = 0;
936 }
937 dterr = DecodeTime(field[i], fmask, INTERVAL_FULL_RANGE,
938 &tmask, tm, fsec);
939 if (dterr)
940 return dterr;
941
942 /* check for time overflow */
943 if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec,
944 *fsec))
945 return DTERR_FIELD_OVERFLOW;
946 break;
947
948 case DTK_TZ:
949 {
950 int tz;
951
952 if (tzp == NULL)
953 return DTERR_BAD_FORMAT;
954
955 dterr = DecodeTimezone(field[i], &tz);
956 if (dterr)
957 return dterr;
958 *tzp = tz;
959 tmask = DTK_M(TZ);
960 }
961 break;
962
963 case DTK_NUMBER:
964
965 /*
966 * Was this an "ISO date" with embedded field labels? An
967 * example is "y2001m02d04" - thomas 2001-02-04
968 */
969 if (ptype != 0)
970 {
971 char *cp;
972 int val;
973
974 errno = 0;
975 val = strtoint(field[i], &cp, 10);
976 if (errno == ERANGE)
977 return DTERR_FIELD_OVERFLOW;
978
979 /*
980 * only a few kinds are allowed to have an embedded
981 * decimal
982 */
983 if (*cp == '.')
984 switch (ptype)
985 {
986 case DTK_JULIAN:
987 case DTK_TIME:
988 case DTK_SECOND:
989 break;
990 default:
991 return DTERR_BAD_FORMAT;
992 break;
993 }
994 else if (*cp != '\0')
995 return DTERR_BAD_FORMAT;
996
997 switch (ptype)
998 {
999 case DTK_YEAR:
1000 tm->tm_year = val;
1001 tmask = DTK_M(YEAR);
1002 break;
1003
1004 case DTK_MONTH:
1005
1006 /*
1007 * already have a month and hour? then assume
1008 * minutes
1009 */
1010 if ((fmask & DTK_M(MONTH)) != 0 &&
1011 (fmask & DTK_M(HOUR)) != 0)
1012 {
1013 tm->tm_min = val;
1014 tmask = DTK_M(MINUTE);
1015 }
1016 else
1017 {
1018 tm->tm_mon = val;
1019 tmask = DTK_M(MONTH);
1020 }
1021 break;
1022
1023 case DTK_DAY:
1024 tm->tm_mday = val;
1025 tmask = DTK_M(DAY);
1026 break;
1027
1028 case DTK_HOUR:
1029 tm->tm_hour = val;
1030 tmask = DTK_M(HOUR);
1031 break;
1032
1033 case DTK_MINUTE:
1034 tm->tm_min = val;
1035 tmask = DTK_M(MINUTE);
1036 break;
1037
1038 case DTK_SECOND:
1039 tm->tm_sec = val;
1040 tmask = DTK_M(SECOND);
1041 if (*cp == '.')
1042 {
1043 dterr = ParseFractionalSecond(cp, fsec);
1044 if (dterr)
1045 return dterr;
1046 tmask = DTK_ALL_SECS_M;
1047 }
1048 break;
1049
1050 case DTK_TZ:
1051 tmask = DTK_M(TZ);
1052 dterr = DecodeTimezone(field[i], tzp);
1053 if (dterr)
1054 return dterr;
1055 break;
1056
1057 case DTK_JULIAN:
1058 /* previous field was a label for "julian date" */
1059 if (val < 0)
1060 return DTERR_FIELD_OVERFLOW;
1061 tmask = DTK_DATE_M;
1062 j2date(val, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1063 isjulian = true;
1064
1065 /* fractional Julian Day? */
1066 if (*cp == '.')
1067 {
1068 double time;
1069
1070 errno = 0;
1071 time = strtod(cp, &cp);
1072 if (*cp != '\0' || errno != 0)
1073 return DTERR_BAD_FORMAT;
1074 time *= USECS_PER_DAY;
1075 dt2time(time,
1076 &tm->tm_hour, &tm->tm_min,
1077 &tm->tm_sec, fsec);
1078 tmask |= DTK_TIME_M;
1079 }
1080 break;
1081
1082 case DTK_TIME:
1083 /* previous field was "t" for ISO time */
1084 dterr = DecodeNumberField(strlen(field[i]), field[i],
1085 (fmask | DTK_DATE_M),
1086 &tmask, tm,
1087 fsec, &is2digits);
1088 if (dterr < 0)
1089 return dterr;
1090 if (tmask != DTK_TIME_M)
1091 return DTERR_BAD_FORMAT;
1092 break;
1093
1094 default:
1095 return DTERR_BAD_FORMAT;
1096 break;
1097 }
1098
1099 ptype = 0;
1100 *dtype = DTK_DATE;
1101 }
1102 else
1103 {
1104 char *cp;
1105 int flen;
1106
1107 flen = strlen(field[i]);
1108 cp = strchr(field[i], '.');
1109
1110 /* Embedded decimal and no date yet? */
1111 if (cp != NULL && !(fmask & DTK_DATE_M))
1112 {
1113 dterr = DecodeDate(field[i], fmask,
1114 &tmask, &is2digits, tm);
1115 if (dterr)
1116 return dterr;
1117 }
1118 /* embedded decimal and several digits before? */
1119 else if (cp != NULL && flen - strlen(cp) > 2)
1120 {
1121 /*
1122 * Interpret as a concatenated date or time Set the
1123 * type field to allow decoding other fields later.
1124 * Example: 20011223 or 040506
1125 */
1126 dterr = DecodeNumberField(flen, field[i], fmask,
1127 &tmask, tm,
1128 fsec, &is2digits);
1129 if (dterr < 0)
1130 return dterr;
1131 }
1132
1133 /*
1134 * Is this a YMD or HMS specification, or a year number?
1135 * YMD and HMS are required to be six digits or more, so
1136 * if it is 5 digits, it is a year. If it is six or more
1137 * digits, we assume it is YMD or HMS unless no date and
1138 * no time values have been specified. This forces 6+
1139 * digit years to be at the end of the string, or to use
1140 * the ISO date specification.
1141 */
1142 else if (flen >= 6 && (!(fmask & DTK_DATE_M) ||
1143 !(fmask & DTK_TIME_M)))
1144 {
1145 dterr = DecodeNumberField(flen, field[i], fmask,
1146 &tmask, tm,
1147 fsec, &is2digits);
1148 if (dterr < 0)
1149 return dterr;
1150 }
1151 /* otherwise it is a single date/time field... */
1152 else
1153 {
1154 dterr = DecodeNumber(flen, field[i],
1155 haveTextMonth, fmask,
1156 &tmask, tm,
1157 fsec, &is2digits);
1158 if (dterr)
1159 return dterr;
1160 }
1161 }
1162 break;
1163
1164 case DTK_STRING:
1165 case DTK_SPECIAL:
1166 /* timezone abbrevs take precedence over built-in tokens */
1167 type = DecodeTimezoneAbbrev(i, field[i], &val, &valtz);
1168 if (type == UNKNOWN_FIELD)
1169 type = DecodeSpecial(i, field[i], &val);
1170 if (type == IGNORE_DTF)
1171 continue;
1172
1173 tmask = DTK_M(type);
1174 switch (type)
1175 {
1176 case RESERV:
1177 switch (val)
1178 {
1179 case DTK_NOW:
1180 tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ));
1181 *dtype = DTK_DATE;
1182 GetCurrentTimeUsec(tm, fsec, tzp);
1183 break;
1184
1185 case DTK_YESTERDAY:
1186 tmask = DTK_DATE_M;
1187 *dtype = DTK_DATE;
1188 GetCurrentDateTime(&cur_tm);
1189 j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) - 1,
1190 &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1191 break;
1192
1193 case DTK_TODAY:
1194 tmask = DTK_DATE_M;
1195 *dtype = DTK_DATE;
1196 GetCurrentDateTime(&cur_tm);
1197 tm->tm_year = cur_tm.tm_year;
1198 tm->tm_mon = cur_tm.tm_mon;
1199 tm->tm_mday = cur_tm.tm_mday;
1200 break;
1201
1202 case DTK_TOMORROW:
1203 tmask = DTK_DATE_M;
1204 *dtype = DTK_DATE;
1205 GetCurrentDateTime(&cur_tm);
1206 j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) + 1,
1207 &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1208 break;
1209
1210 case DTK_ZULU:
1211 tmask = (DTK_TIME_M | DTK_M(TZ));
1212 *dtype = DTK_DATE;
1213 tm->tm_hour = 0;
1214 tm->tm_min = 0;
1215 tm->tm_sec = 0;
1216 if (tzp != NULL)
1217 *tzp = 0;
1218 break;
1219
1220 default:
1221 *dtype = val;
1222 }
1223
1224 break;
1225
1226 case MONTH:
1227
1228 /*
1229 * already have a (numeric) month? then see if we can
1230 * substitute...
1231 */
1232 if ((fmask & DTK_M(MONTH)) && !haveTextMonth &&
1233 !(fmask & DTK_M(DAY)) && tm->tm_mon >= 1 &&
1234 tm->tm_mon <= 31)
1235 {
1236 tm->tm_mday = tm->tm_mon;
1237 tmask = DTK_M(DAY);
1238 }
1239 haveTextMonth = true;
1240 tm->tm_mon = val;
1241 break;
1242
1243 case DTZMOD:
1244
1245 /*
1246 * daylight savings time modifier (solves "MET DST"
1247 * syntax)
1248 */
1249 tmask |= DTK_M(DTZ);
1250 tm->tm_isdst = 1;
1251 if (tzp == NULL)
1252 return DTERR_BAD_FORMAT;
1253 *tzp -= val;
1254 break;
1255
1256 case DTZ:
1257
1258 /*
1259 * set mask for TZ here _or_ check for DTZ later when
1260 * getting default timezone
1261 */
1262 tmask |= DTK_M(TZ);
1263 tm->tm_isdst = 1;
1264 if (tzp == NULL)
1265 return DTERR_BAD_FORMAT;
1266 *tzp = -val;
1267 break;
1268
1269 case TZ:
1270 tm->tm_isdst = 0;
1271 if (tzp == NULL)
1272 return DTERR_BAD_FORMAT;
1273 *tzp = -val;
1274 break;
1275
1276 case DYNTZ:
1277 tmask |= DTK_M(TZ);
1278 if (tzp == NULL)
1279 return DTERR_BAD_FORMAT;
1280 /* we'll determine the actual offset later */
1281 abbrevTz = valtz;
1282 abbrev = field[i];
1283 break;
1284
1285 case AMPM:
1286 mer = val;
1287 break;
1288
1289 case ADBC:
1290 bc = (val == BC);
1291 break;
1292
1293 case DOW:
1294 tm->tm_wday = val;
1295 break;
1296
1297 case UNITS:
1298 tmask = 0;
1299 ptype = val;
1300 break;
1301
1302 case ISOTIME:
1303
1304 /*
1305 * This is a filler field "t" indicating that the next
1306 * field is time. Try to verify that this is sensible.
1307 */
1308 tmask = 0;
1309
1310 /* No preceding date? Then quit... */
1311 if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1312 return DTERR_BAD_FORMAT;
1313
1314 /***
1315 * We will need one of the following fields:
1316 * DTK_NUMBER should be hhmmss.fff
1317 * DTK_TIME should be hh:mm:ss.fff
1318 * DTK_DATE should be hhmmss-zz
1319 ***/
1320 if (i >= nf - 1 ||
1321 (ftype[i + 1] != DTK_NUMBER &&
1322 ftype[i + 1] != DTK_TIME &&
1323 ftype[i + 1] != DTK_DATE))
1324 return DTERR_BAD_FORMAT;
1325
1326 ptype = val;
1327 break;
1328
1329 case UNKNOWN_FIELD:
1330
1331 /*
1332 * Before giving up and declaring error, check to see
1333 * if it is an all-alpha timezone name.
1334 */
1335 namedTz = pg_tzset(field[i]);
1336 if (!namedTz)
1337 return DTERR_BAD_FORMAT;
1338 /* we'll apply the zone setting below */
1339 tmask = DTK_M(TZ);
1340 break;
1341
1342 default:
1343 return DTERR_BAD_FORMAT;
1344 }
1345 break;
1346
1347 default:
1348 return DTERR_BAD_FORMAT;
1349 }
1350
1351 if (tmask & fmask)
1352 return DTERR_BAD_FORMAT;
1353 fmask |= tmask;
1354 } /* end loop over fields */
1355
1356 /* do final checking/adjustment of Y/M/D fields */
1357 dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
1358 if (dterr)
1359 return dterr;
1360
1361 /* handle AM/PM */
1362 if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
1363 return DTERR_FIELD_OVERFLOW;
1364 if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
1365 tm->tm_hour = 0;
1366 else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
1367 tm->tm_hour += HOURS_PER_DAY / 2;
1368
1369 /* do additional checking for full date specs... */
1370 if (*dtype == DTK_DATE)
1371 {
1372 if ((fmask & DTK_DATE_M) != DTK_DATE_M)
1373 {
1374 if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1375 return 1;
1376 return DTERR_BAD_FORMAT;
1377 }
1378
1379 /*
1380 * If we had a full timezone spec, compute the offset (we could not do
1381 * it before, because we need the date to resolve DST status).
1382 */
1383 if (namedTz != NULL)
1384 {
1385 /* daylight savings time modifier disallowed with full TZ */
1386 if (fmask & DTK_M(DTZMOD))
1387 return DTERR_BAD_FORMAT;
1388
1389 *tzp = DetermineTimeZoneOffset(tm, namedTz);
1390 }
1391
1392 /*
1393 * Likewise, if we had a dynamic timezone abbreviation, resolve it
1394 * now.
1395 */
1396 if (abbrevTz != NULL)
1397 {
1398 /* daylight savings time modifier disallowed with dynamic TZ */
1399 if (fmask & DTK_M(DTZMOD))
1400 return DTERR_BAD_FORMAT;
1401
1402 *tzp = DetermineTimeZoneAbbrevOffset(tm, abbrev, abbrevTz);
1403 }
1404
1405 /* timezone not specified? then use session timezone */
1406 if (tzp != NULL && !(fmask & DTK_M(TZ)))
1407 {
1408 /*
1409 * daylight savings time modifier but no standard timezone? then
1410 * error
1411 */
1412 if (fmask & DTK_M(DTZMOD))
1413 return DTERR_BAD_FORMAT;
1414
1415 *tzp = DetermineTimeZoneOffset(tm, session_timezone);
1416 }
1417 }
1418
1419 return 0;
1420 }
1421
1422
1423 /* DetermineTimeZoneOffset()
1424 *
1425 * Given a struct pg_tm in which tm_year, tm_mon, tm_mday, tm_hour, tm_min,
1426 * and tm_sec fields are set, and a zic-style time zone definition, determine
1427 * the applicable GMT offset and daylight-savings status at that time.
1428 * Set the struct pg_tm's tm_isdst field accordingly, and return the GMT
1429 * offset as the function result.
1430 *
1431 * Note: if the date is out of the range we can deal with, we return zero
1432 * as the GMT offset and set tm_isdst = 0. We don't throw an error here,
1433 * though probably some higher-level code will.
1434 */
1435 int
DetermineTimeZoneOffset(struct pg_tm * tm,pg_tz * tzp)1436 DetermineTimeZoneOffset(struct pg_tm *tm, pg_tz *tzp)
1437 {
1438 pg_time_t t;
1439
1440 return DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1441 }
1442
1443
1444 /* DetermineTimeZoneOffsetInternal()
1445 *
1446 * As above, but also return the actual UTC time imputed to the date/time
1447 * into *tp.
1448 *
1449 * In event of an out-of-range date, we punt by returning zero into *tp.
1450 * This is okay for the immediate callers but is a good reason for not
1451 * exposing this worker function globally.
1452 *
1453 * Note: it might seem that we should use mktime() for this, but bitter
1454 * experience teaches otherwise. This code is much faster than most versions
1455 * of mktime(), anyway.
1456 */
1457 static int
DetermineTimeZoneOffsetInternal(struct pg_tm * tm,pg_tz * tzp,pg_time_t * tp)1458 DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp, pg_time_t *tp)
1459 {
1460 int date,
1461 sec;
1462 pg_time_t day,
1463 mytime,
1464 prevtime,
1465 boundary,
1466 beforetime,
1467 aftertime;
1468 long int before_gmtoff,
1469 after_gmtoff;
1470 int before_isdst,
1471 after_isdst;
1472 int res;
1473
1474 /*
1475 * First, generate the pg_time_t value corresponding to the given
1476 * y/m/d/h/m/s taken as GMT time. If this overflows, punt and decide the
1477 * timezone is GMT. (For a valid Julian date, integer overflow should be
1478 * impossible with 64-bit pg_time_t, but let's check for safety.)
1479 */
1480 if (!IS_VALID_JULIAN(tm->tm_year, tm->tm_mon, tm->tm_mday))
1481 goto overflow;
1482 date = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday) - UNIX_EPOCH_JDATE;
1483
1484 day = ((pg_time_t) date) * SECS_PER_DAY;
1485 if (day / SECS_PER_DAY != date)
1486 goto overflow;
1487 sec = tm->tm_sec + (tm->tm_min + tm->tm_hour * MINS_PER_HOUR) * SECS_PER_MINUTE;
1488 mytime = day + sec;
1489 /* since sec >= 0, overflow could only be from +day to -mytime */
1490 if (mytime < 0 && day > 0)
1491 goto overflow;
1492
1493 /*
1494 * Find the DST time boundary just before or following the target time. We
1495 * assume that all zones have GMT offsets less than 24 hours, and that DST
1496 * boundaries can't be closer together than 48 hours, so backing up 24
1497 * hours and finding the "next" boundary will work.
1498 */
1499 prevtime = mytime - SECS_PER_DAY;
1500 if (mytime < 0 && prevtime > 0)
1501 goto overflow;
1502
1503 res = pg_next_dst_boundary(&prevtime,
1504 &before_gmtoff, &before_isdst,
1505 &boundary,
1506 &after_gmtoff, &after_isdst,
1507 tzp);
1508 if (res < 0)
1509 goto overflow; /* failure? */
1510
1511 if (res == 0)
1512 {
1513 /* Non-DST zone, life is simple */
1514 tm->tm_isdst = before_isdst;
1515 *tp = mytime - before_gmtoff;
1516 return -(int) before_gmtoff;
1517 }
1518
1519 /*
1520 * Form the candidate pg_time_t values with local-time adjustment
1521 */
1522 beforetime = mytime - before_gmtoff;
1523 if ((before_gmtoff > 0 &&
1524 mytime < 0 && beforetime > 0) ||
1525 (before_gmtoff <= 0 &&
1526 mytime > 0 && beforetime < 0))
1527 goto overflow;
1528 aftertime = mytime - after_gmtoff;
1529 if ((after_gmtoff > 0 &&
1530 mytime < 0 && aftertime > 0) ||
1531 (after_gmtoff <= 0 &&
1532 mytime > 0 && aftertime < 0))
1533 goto overflow;
1534
1535 /*
1536 * If both before or both after the boundary time, we know what to do. The
1537 * boundary time itself is considered to be after the transition, which
1538 * means we can accept aftertime == boundary in the second case.
1539 */
1540 if (beforetime < boundary && aftertime < boundary)
1541 {
1542 tm->tm_isdst = before_isdst;
1543 *tp = beforetime;
1544 return -(int) before_gmtoff;
1545 }
1546 if (beforetime > boundary && aftertime >= boundary)
1547 {
1548 tm->tm_isdst = after_isdst;
1549 *tp = aftertime;
1550 return -(int) after_gmtoff;
1551 }
1552
1553 /*
1554 * It's an invalid or ambiguous time due to timezone transition. In a
1555 * spring-forward transition, prefer the "before" interpretation; in a
1556 * fall-back transition, prefer "after". (We used to define and implement
1557 * this test as "prefer the standard-time interpretation", but that rule
1558 * does not help to resolve the behavior when both times are reported as
1559 * standard time; which does happen, eg Europe/Moscow in Oct 2014. Also,
1560 * in some zones such as Europe/Dublin, there is widespread confusion
1561 * about which time offset is "standard" time, so it's fortunate that our
1562 * behavior doesn't depend on that.)
1563 */
1564 if (beforetime > aftertime)
1565 {
1566 tm->tm_isdst = before_isdst;
1567 *tp = beforetime;
1568 return -(int) before_gmtoff;
1569 }
1570 tm->tm_isdst = after_isdst;
1571 *tp = aftertime;
1572 return -(int) after_gmtoff;
1573
1574 overflow:
1575 /* Given date is out of range, so assume UTC */
1576 tm->tm_isdst = 0;
1577 *tp = 0;
1578 return 0;
1579 }
1580
1581
1582 /* DetermineTimeZoneAbbrevOffset()
1583 *
1584 * Determine the GMT offset and DST flag to be attributed to a dynamic
1585 * time zone abbreviation, that is one whose meaning has changed over time.
1586 * *tm contains the local time at which the meaning should be determined,
1587 * and tm->tm_isdst receives the DST flag.
1588 *
1589 * This differs from the behavior of DetermineTimeZoneOffset() in that a
1590 * standard-time or daylight-time abbreviation forces use of the corresponding
1591 * GMT offset even when the zone was then in DS or standard time respectively.
1592 * (However, that happens only if we can match the given abbreviation to some
1593 * abbreviation that appears in the IANA timezone data. Otherwise, we fall
1594 * back to doing DetermineTimeZoneOffset().)
1595 */
1596 int
DetermineTimeZoneAbbrevOffset(struct pg_tm * tm,const char * abbr,pg_tz * tzp)1597 DetermineTimeZoneAbbrevOffset(struct pg_tm *tm, const char *abbr, pg_tz *tzp)
1598 {
1599 pg_time_t t;
1600 int zone_offset;
1601 int abbr_offset;
1602 int abbr_isdst;
1603
1604 /*
1605 * Compute the UTC time we want to probe at. (In event of overflow, we'll
1606 * probe at the epoch, which is a bit random but probably doesn't matter.)
1607 */
1608 zone_offset = DetermineTimeZoneOffsetInternal(tm, tzp, &t);
1609
1610 /*
1611 * Try to match the abbreviation to something in the zone definition.
1612 */
1613 if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1614 &abbr_offset, &abbr_isdst))
1615 {
1616 /* Success, so use the abbrev-specific answers. */
1617 tm->tm_isdst = abbr_isdst;
1618 return abbr_offset;
1619 }
1620
1621 /*
1622 * No match, so use the answers we already got from
1623 * DetermineTimeZoneOffsetInternal.
1624 */
1625 return zone_offset;
1626 }
1627
1628
1629 /* DetermineTimeZoneAbbrevOffsetTS()
1630 *
1631 * As above but the probe time is specified as a TimestampTz (hence, UTC time),
1632 * and DST status is returned into *isdst rather than into tm->tm_isdst.
1633 */
1634 int
DetermineTimeZoneAbbrevOffsetTS(TimestampTz ts,const char * abbr,pg_tz * tzp,int * isdst)1635 DetermineTimeZoneAbbrevOffsetTS(TimestampTz ts, const char *abbr,
1636 pg_tz *tzp, int *isdst)
1637 {
1638 pg_time_t t = timestamptz_to_time_t(ts);
1639 int zone_offset;
1640 int abbr_offset;
1641 int tz;
1642 struct pg_tm tm;
1643 fsec_t fsec;
1644
1645 /*
1646 * If the abbrev matches anything in the zone data, this is pretty easy.
1647 */
1648 if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp,
1649 &abbr_offset, isdst))
1650 return abbr_offset;
1651
1652 /*
1653 * Else, break down the timestamp so we can use DetermineTimeZoneOffset.
1654 */
1655 if (timestamp2tm(ts, &tz, &tm, &fsec, NULL, tzp) != 0)
1656 ereport(ERROR,
1657 (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE),
1658 errmsg("timestamp out of range")));
1659
1660 zone_offset = DetermineTimeZoneOffset(&tm, tzp);
1661 *isdst = tm.tm_isdst;
1662 return zone_offset;
1663 }
1664
1665
1666 /* DetermineTimeZoneAbbrevOffsetInternal()
1667 *
1668 * Workhorse for above two functions: work from a pg_time_t probe instant.
1669 * On success, return GMT offset and DST status into *offset and *isdst.
1670 */
1671 static bool
DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t,const char * abbr,pg_tz * tzp,int * offset,int * isdst)1672 DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t, const char *abbr, pg_tz *tzp,
1673 int *offset, int *isdst)
1674 {
1675 char upabbr[TZ_STRLEN_MAX + 1];
1676 unsigned char *p;
1677 long int gmtoff;
1678
1679 /* We need to force the abbrev to upper case */
1680 strlcpy(upabbr, abbr, sizeof(upabbr));
1681 for (p = (unsigned char *) upabbr; *p; p++)
1682 *p = pg_toupper(*p);
1683
1684 /* Look up the abbrev's meaning at this time in this zone */
1685 if (pg_interpret_timezone_abbrev(upabbr,
1686 &t,
1687 &gmtoff,
1688 isdst,
1689 tzp))
1690 {
1691 /* Change sign to agree with DetermineTimeZoneOffset() */
1692 *offset = (int) -gmtoff;
1693 return true;
1694 }
1695 return false;
1696 }
1697
1698
1699 /* DecodeTimeOnly()
1700 * Interpret parsed string as time fields only.
1701 * Returns 0 if successful, DTERR code if bogus input detected.
1702 *
1703 * Note that support for time zone is here for
1704 * SQL TIME WITH TIME ZONE, but it reveals
1705 * bogosity with SQL date/time standards, since
1706 * we must infer a time zone from current time.
1707 * - thomas 2000-03-10
1708 * Allow specifying date to get a better time zone,
1709 * if time zones are allowed. - thomas 2001-12-26
1710 */
1711 int
DecodeTimeOnly(char ** field,int * ftype,int nf,int * dtype,struct pg_tm * tm,fsec_t * fsec,int * tzp)1712 DecodeTimeOnly(char **field, int *ftype, int nf,
1713 int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp)
1714 {
1715 int fmask = 0,
1716 tmask,
1717 type;
1718 int ptype = 0; /* "prefix type" for ISO h04mm05s06 format */
1719 int i;
1720 int val;
1721 int dterr;
1722 bool isjulian = false;
1723 bool is2digits = false;
1724 bool bc = false;
1725 int mer = HR24;
1726 pg_tz *namedTz = NULL;
1727 pg_tz *abbrevTz = NULL;
1728 char *abbrev = NULL;
1729 pg_tz *valtz;
1730
1731 *dtype = DTK_TIME;
1732 tm->tm_hour = 0;
1733 tm->tm_min = 0;
1734 tm->tm_sec = 0;
1735 *fsec = 0;
1736 /* don't know daylight savings time status apriori */
1737 tm->tm_isdst = -1;
1738
1739 if (tzp != NULL)
1740 *tzp = 0;
1741
1742 for (i = 0; i < nf; i++)
1743 {
1744 switch (ftype[i])
1745 {
1746 case DTK_DATE:
1747
1748 /*
1749 * Time zone not allowed? Then should not accept dates or time
1750 * zones no matter what else!
1751 */
1752 if (tzp == NULL)
1753 return DTERR_BAD_FORMAT;
1754
1755 /* Under limited circumstances, we will accept a date... */
1756 if (i == 0 && nf >= 2 &&
1757 (ftype[nf - 1] == DTK_DATE || ftype[1] == DTK_TIME))
1758 {
1759 dterr = DecodeDate(field[i], fmask,
1760 &tmask, &is2digits, tm);
1761 if (dterr)
1762 return dterr;
1763 }
1764 /* otherwise, this is a time and/or time zone */
1765 else
1766 {
1767 if (isdigit((unsigned char) *field[i]))
1768 {
1769 char *cp;
1770
1771 /*
1772 * Starts with a digit but we already have a time
1773 * field? Then we are in trouble with time already...
1774 */
1775 if ((fmask & DTK_TIME_M) == DTK_TIME_M)
1776 return DTERR_BAD_FORMAT;
1777
1778 /*
1779 * Should not get here and fail. Sanity check only...
1780 */
1781 if ((cp = strchr(field[i], '-')) == NULL)
1782 return DTERR_BAD_FORMAT;
1783
1784 /* Get the time zone from the end of the string */
1785 dterr = DecodeTimezone(cp, tzp);
1786 if (dterr)
1787 return dterr;
1788 *cp = '\0';
1789
1790 /*
1791 * Then read the rest of the field as a concatenated
1792 * time
1793 */
1794 dterr = DecodeNumberField(strlen(field[i]), field[i],
1795 (fmask | DTK_DATE_M),
1796 &tmask, tm,
1797 fsec, &is2digits);
1798 if (dterr < 0)
1799 return dterr;
1800 ftype[i] = dterr;
1801
1802 tmask |= DTK_M(TZ);
1803 }
1804 else
1805 {
1806 namedTz = pg_tzset(field[i]);
1807 if (!namedTz)
1808 {
1809 /*
1810 * We should return an error code instead of
1811 * ereport'ing directly, but then there is no way
1812 * to report the bad time zone name.
1813 */
1814 ereport(ERROR,
1815 (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
1816 errmsg("time zone \"%s\" not recognized",
1817 field[i])));
1818 }
1819 /* we'll apply the zone setting below */
1820 ftype[i] = DTK_TZ;
1821 tmask = DTK_M(TZ);
1822 }
1823 }
1824 break;
1825
1826 case DTK_TIME:
1827 dterr = DecodeTime(field[i], (fmask | DTK_DATE_M),
1828 INTERVAL_FULL_RANGE,
1829 &tmask, tm, fsec);
1830 if (dterr)
1831 return dterr;
1832 break;
1833
1834 case DTK_TZ:
1835 {
1836 int tz;
1837
1838 if (tzp == NULL)
1839 return DTERR_BAD_FORMAT;
1840
1841 dterr = DecodeTimezone(field[i], &tz);
1842 if (dterr)
1843 return dterr;
1844 *tzp = tz;
1845 tmask = DTK_M(TZ);
1846 }
1847 break;
1848
1849 case DTK_NUMBER:
1850
1851 /*
1852 * Was this an "ISO time" with embedded field labels? An
1853 * example is "h04m05s06" - thomas 2001-02-04
1854 */
1855 if (ptype != 0)
1856 {
1857 char *cp;
1858 int val;
1859
1860 /* Only accept a date under limited circumstances */
1861 switch (ptype)
1862 {
1863 case DTK_JULIAN:
1864 case DTK_YEAR:
1865 case DTK_MONTH:
1866 case DTK_DAY:
1867 if (tzp == NULL)
1868 return DTERR_BAD_FORMAT;
1869 default:
1870 break;
1871 }
1872
1873 errno = 0;
1874 val = strtoint(field[i], &cp, 10);
1875 if (errno == ERANGE)
1876 return DTERR_FIELD_OVERFLOW;
1877
1878 /*
1879 * only a few kinds are allowed to have an embedded
1880 * decimal
1881 */
1882 if (*cp == '.')
1883 switch (ptype)
1884 {
1885 case DTK_JULIAN:
1886 case DTK_TIME:
1887 case DTK_SECOND:
1888 break;
1889 default:
1890 return DTERR_BAD_FORMAT;
1891 break;
1892 }
1893 else if (*cp != '\0')
1894 return DTERR_BAD_FORMAT;
1895
1896 switch (ptype)
1897 {
1898 case DTK_YEAR:
1899 tm->tm_year = val;
1900 tmask = DTK_M(YEAR);
1901 break;
1902
1903 case DTK_MONTH:
1904
1905 /*
1906 * already have a month and hour? then assume
1907 * minutes
1908 */
1909 if ((fmask & DTK_M(MONTH)) != 0 &&
1910 (fmask & DTK_M(HOUR)) != 0)
1911 {
1912 tm->tm_min = val;
1913 tmask = DTK_M(MINUTE);
1914 }
1915 else
1916 {
1917 tm->tm_mon = val;
1918 tmask = DTK_M(MONTH);
1919 }
1920 break;
1921
1922 case DTK_DAY:
1923 tm->tm_mday = val;
1924 tmask = DTK_M(DAY);
1925 break;
1926
1927 case DTK_HOUR:
1928 tm->tm_hour = val;
1929 tmask = DTK_M(HOUR);
1930 break;
1931
1932 case DTK_MINUTE:
1933 tm->tm_min = val;
1934 tmask = DTK_M(MINUTE);
1935 break;
1936
1937 case DTK_SECOND:
1938 tm->tm_sec = val;
1939 tmask = DTK_M(SECOND);
1940 if (*cp == '.')
1941 {
1942 dterr = ParseFractionalSecond(cp, fsec);
1943 if (dterr)
1944 return dterr;
1945 tmask = DTK_ALL_SECS_M;
1946 }
1947 break;
1948
1949 case DTK_TZ:
1950 tmask = DTK_M(TZ);
1951 dterr = DecodeTimezone(field[i], tzp);
1952 if (dterr)
1953 return dterr;
1954 break;
1955
1956 case DTK_JULIAN:
1957 /* previous field was a label for "julian date" */
1958 if (val < 0)
1959 return DTERR_FIELD_OVERFLOW;
1960 tmask = DTK_DATE_M;
1961 j2date(val, &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
1962 isjulian = true;
1963
1964 if (*cp == '.')
1965 {
1966 double time;
1967
1968 errno = 0;
1969 time = strtod(cp, &cp);
1970 if (*cp != '\0' || errno != 0)
1971 return DTERR_BAD_FORMAT;
1972 time *= USECS_PER_DAY;
1973 dt2time(time,
1974 &tm->tm_hour, &tm->tm_min,
1975 &tm->tm_sec, fsec);
1976 tmask |= DTK_TIME_M;
1977 }
1978 break;
1979
1980 case DTK_TIME:
1981 /* previous field was "t" for ISO time */
1982 dterr = DecodeNumberField(strlen(field[i]), field[i],
1983 (fmask | DTK_DATE_M),
1984 &tmask, tm,
1985 fsec, &is2digits);
1986 if (dterr < 0)
1987 return dterr;
1988 ftype[i] = dterr;
1989
1990 if (tmask != DTK_TIME_M)
1991 return DTERR_BAD_FORMAT;
1992 break;
1993
1994 default:
1995 return DTERR_BAD_FORMAT;
1996 break;
1997 }
1998
1999 ptype = 0;
2000 *dtype = DTK_DATE;
2001 }
2002 else
2003 {
2004 char *cp;
2005 int flen;
2006
2007 flen = strlen(field[i]);
2008 cp = strchr(field[i], '.');
2009
2010 /* Embedded decimal? */
2011 if (cp != NULL)
2012 {
2013 /*
2014 * Under limited circumstances, we will accept a
2015 * date...
2016 */
2017 if (i == 0 && nf >= 2 && ftype[nf - 1] == DTK_DATE)
2018 {
2019 dterr = DecodeDate(field[i], fmask,
2020 &tmask, &is2digits, tm);
2021 if (dterr)
2022 return dterr;
2023 }
2024 /* embedded decimal and several digits before? */
2025 else if (flen - strlen(cp) > 2)
2026 {
2027 /*
2028 * Interpret as a concatenated date or time Set
2029 * the type field to allow decoding other fields
2030 * later. Example: 20011223 or 040506
2031 */
2032 dterr = DecodeNumberField(flen, field[i],
2033 (fmask | DTK_DATE_M),
2034 &tmask, tm,
2035 fsec, &is2digits);
2036 if (dterr < 0)
2037 return dterr;
2038 ftype[i] = dterr;
2039 }
2040 else
2041 return DTERR_BAD_FORMAT;
2042 }
2043 else if (flen > 4)
2044 {
2045 dterr = DecodeNumberField(flen, field[i],
2046 (fmask | DTK_DATE_M),
2047 &tmask, tm,
2048 fsec, &is2digits);
2049 if (dterr < 0)
2050 return dterr;
2051 ftype[i] = dterr;
2052 }
2053 /* otherwise it is a single date/time field... */
2054 else
2055 {
2056 dterr = DecodeNumber(flen, field[i],
2057 false,
2058 (fmask | DTK_DATE_M),
2059 &tmask, tm,
2060 fsec, &is2digits);
2061 if (dterr)
2062 return dterr;
2063 }
2064 }
2065 break;
2066
2067 case DTK_STRING:
2068 case DTK_SPECIAL:
2069 /* timezone abbrevs take precedence over built-in tokens */
2070 type = DecodeTimezoneAbbrev(i, field[i], &val, &valtz);
2071 if (type == UNKNOWN_FIELD)
2072 type = DecodeSpecial(i, field[i], &val);
2073 if (type == IGNORE_DTF)
2074 continue;
2075
2076 tmask = DTK_M(type);
2077 switch (type)
2078 {
2079 case RESERV:
2080 switch (val)
2081 {
2082 case DTK_NOW:
2083 tmask = DTK_TIME_M;
2084 *dtype = DTK_TIME;
2085 GetCurrentTimeUsec(tm, fsec, NULL);
2086 break;
2087
2088 case DTK_ZULU:
2089 tmask = (DTK_TIME_M | DTK_M(TZ));
2090 *dtype = DTK_TIME;
2091 tm->tm_hour = 0;
2092 tm->tm_min = 0;
2093 tm->tm_sec = 0;
2094 tm->tm_isdst = 0;
2095 break;
2096
2097 default:
2098 return DTERR_BAD_FORMAT;
2099 }
2100
2101 break;
2102
2103 case DTZMOD:
2104
2105 /*
2106 * daylight savings time modifier (solves "MET DST"
2107 * syntax)
2108 */
2109 tmask |= DTK_M(DTZ);
2110 tm->tm_isdst = 1;
2111 if (tzp == NULL)
2112 return DTERR_BAD_FORMAT;
2113 *tzp -= val;
2114 break;
2115
2116 case DTZ:
2117
2118 /*
2119 * set mask for TZ here _or_ check for DTZ later when
2120 * getting default timezone
2121 */
2122 tmask |= DTK_M(TZ);
2123 tm->tm_isdst = 1;
2124 if (tzp == NULL)
2125 return DTERR_BAD_FORMAT;
2126 *tzp = -val;
2127 ftype[i] = DTK_TZ;
2128 break;
2129
2130 case TZ:
2131 tm->tm_isdst = 0;
2132 if (tzp == NULL)
2133 return DTERR_BAD_FORMAT;
2134 *tzp = -val;
2135 ftype[i] = DTK_TZ;
2136 break;
2137
2138 case DYNTZ:
2139 tmask |= DTK_M(TZ);
2140 if (tzp == NULL)
2141 return DTERR_BAD_FORMAT;
2142 /* we'll determine the actual offset later */
2143 abbrevTz = valtz;
2144 abbrev = field[i];
2145 ftype[i] = DTK_TZ;
2146 break;
2147
2148 case AMPM:
2149 mer = val;
2150 break;
2151
2152 case ADBC:
2153 bc = (val == BC);
2154 break;
2155
2156 case UNITS:
2157 tmask = 0;
2158 ptype = val;
2159 break;
2160
2161 case ISOTIME:
2162 tmask = 0;
2163
2164 /***
2165 * We will need one of the following fields:
2166 * DTK_NUMBER should be hhmmss.fff
2167 * DTK_TIME should be hh:mm:ss.fff
2168 * DTK_DATE should be hhmmss-zz
2169 ***/
2170 if (i >= nf - 1 ||
2171 (ftype[i + 1] != DTK_NUMBER &&
2172 ftype[i + 1] != DTK_TIME &&
2173 ftype[i + 1] != DTK_DATE))
2174 return DTERR_BAD_FORMAT;
2175
2176 ptype = val;
2177 break;
2178
2179 case UNKNOWN_FIELD:
2180
2181 /*
2182 * Before giving up and declaring error, check to see
2183 * if it is an all-alpha timezone name.
2184 */
2185 namedTz = pg_tzset(field[i]);
2186 if (!namedTz)
2187 return DTERR_BAD_FORMAT;
2188 /* we'll apply the zone setting below */
2189 tmask = DTK_M(TZ);
2190 break;
2191
2192 default:
2193 return DTERR_BAD_FORMAT;
2194 }
2195 break;
2196
2197 default:
2198 return DTERR_BAD_FORMAT;
2199 }
2200
2201 if (tmask & fmask)
2202 return DTERR_BAD_FORMAT;
2203 fmask |= tmask;
2204 } /* end loop over fields */
2205
2206 /* do final checking/adjustment of Y/M/D fields */
2207 dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm);
2208 if (dterr)
2209 return dterr;
2210
2211 /* handle AM/PM */
2212 if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2)
2213 return DTERR_FIELD_OVERFLOW;
2214 if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2)
2215 tm->tm_hour = 0;
2216 else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2)
2217 tm->tm_hour += HOURS_PER_DAY / 2;
2218
2219 /* check for time overflow */
2220 if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec, *fsec))
2221 return DTERR_FIELD_OVERFLOW;
2222
2223 if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2224 return DTERR_BAD_FORMAT;
2225
2226 /*
2227 * If we had a full timezone spec, compute the offset (we could not do it
2228 * before, because we may need the date to resolve DST status).
2229 */
2230 if (namedTz != NULL)
2231 {
2232 long int gmtoff;
2233
2234 /* daylight savings time modifier disallowed with full TZ */
2235 if (fmask & DTK_M(DTZMOD))
2236 return DTERR_BAD_FORMAT;
2237
2238 /* if non-DST zone, we do not need to know the date */
2239 if (pg_get_timezone_offset(namedTz, &gmtoff))
2240 {
2241 *tzp = -(int) gmtoff;
2242 }
2243 else
2244 {
2245 /* a date has to be specified */
2246 if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2247 return DTERR_BAD_FORMAT;
2248 *tzp = DetermineTimeZoneOffset(tm, namedTz);
2249 }
2250 }
2251
2252 /*
2253 * Likewise, if we had a dynamic timezone abbreviation, resolve it now.
2254 */
2255 if (abbrevTz != NULL)
2256 {
2257 struct pg_tm tt,
2258 *tmp = &tt;
2259
2260 /*
2261 * daylight savings time modifier but no standard timezone? then error
2262 */
2263 if (fmask & DTK_M(DTZMOD))
2264 return DTERR_BAD_FORMAT;
2265
2266 if ((fmask & DTK_DATE_M) == 0)
2267 GetCurrentDateTime(tmp);
2268 else
2269 {
2270 /* a date has to be specified */
2271 if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2272 return DTERR_BAD_FORMAT;
2273 tmp->tm_year = tm->tm_year;
2274 tmp->tm_mon = tm->tm_mon;
2275 tmp->tm_mday = tm->tm_mday;
2276 }
2277 tmp->tm_hour = tm->tm_hour;
2278 tmp->tm_min = tm->tm_min;
2279 tmp->tm_sec = tm->tm_sec;
2280 *tzp = DetermineTimeZoneAbbrevOffset(tmp, abbrev, abbrevTz);
2281 tm->tm_isdst = tmp->tm_isdst;
2282 }
2283
2284 /* timezone not specified? then use session timezone */
2285 if (tzp != NULL && !(fmask & DTK_M(TZ)))
2286 {
2287 struct pg_tm tt,
2288 *tmp = &tt;
2289
2290 /*
2291 * daylight savings time modifier but no standard timezone? then error
2292 */
2293 if (fmask & DTK_M(DTZMOD))
2294 return DTERR_BAD_FORMAT;
2295
2296 if ((fmask & DTK_DATE_M) == 0)
2297 GetCurrentDateTime(tmp);
2298 else
2299 {
2300 /* a date has to be specified */
2301 if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2302 return DTERR_BAD_FORMAT;
2303 tmp->tm_year = tm->tm_year;
2304 tmp->tm_mon = tm->tm_mon;
2305 tmp->tm_mday = tm->tm_mday;
2306 }
2307 tmp->tm_hour = tm->tm_hour;
2308 tmp->tm_min = tm->tm_min;
2309 tmp->tm_sec = tm->tm_sec;
2310 *tzp = DetermineTimeZoneOffset(tmp, session_timezone);
2311 tm->tm_isdst = tmp->tm_isdst;
2312 }
2313
2314 return 0;
2315 }
2316
2317 /* DecodeDate()
2318 * Decode date string which includes delimiters.
2319 * Return 0 if okay, a DTERR code if not.
2320 *
2321 * str: field to be parsed
2322 * fmask: bitmask for field types already seen
2323 * *tmask: receives bitmask for fields found here
2324 * *is2digits: set to true if we find 2-digit year
2325 * *tm: field values are stored into appropriate members of this struct
2326 */
2327 static int
DecodeDate(char * str,int fmask,int * tmask,bool * is2digits,struct pg_tm * tm)2328 DecodeDate(char *str, int fmask, int *tmask, bool *is2digits,
2329 struct pg_tm *tm)
2330 {
2331 fsec_t fsec;
2332 int nf = 0;
2333 int i,
2334 len;
2335 int dterr;
2336 bool haveTextMonth = false;
2337 int type,
2338 val,
2339 dmask = 0;
2340 char *field[MAXDATEFIELDS];
2341
2342 *tmask = 0;
2343
2344 /* parse this string... */
2345 while (*str != '\0' && nf < MAXDATEFIELDS)
2346 {
2347 /* skip field separators */
2348 while (*str != '\0' && !isalnum((unsigned char) *str))
2349 str++;
2350
2351 if (*str == '\0')
2352 return DTERR_BAD_FORMAT; /* end of string after separator */
2353
2354 field[nf] = str;
2355 if (isdigit((unsigned char) *str))
2356 {
2357 while (isdigit((unsigned char) *str))
2358 str++;
2359 }
2360 else if (isalpha((unsigned char) *str))
2361 {
2362 while (isalpha((unsigned char) *str))
2363 str++;
2364 }
2365
2366 /* Just get rid of any non-digit, non-alpha characters... */
2367 if (*str != '\0')
2368 *str++ = '\0';
2369 nf++;
2370 }
2371
2372 /* look first for text fields, since that will be unambiguous month */
2373 for (i = 0; i < nf; i++)
2374 {
2375 if (isalpha((unsigned char) *field[i]))
2376 {
2377 type = DecodeSpecial(i, field[i], &val);
2378 if (type == IGNORE_DTF)
2379 continue;
2380
2381 dmask = DTK_M(type);
2382 switch (type)
2383 {
2384 case MONTH:
2385 tm->tm_mon = val;
2386 haveTextMonth = true;
2387 break;
2388
2389 default:
2390 return DTERR_BAD_FORMAT;
2391 }
2392 if (fmask & dmask)
2393 return DTERR_BAD_FORMAT;
2394
2395 fmask |= dmask;
2396 *tmask |= dmask;
2397
2398 /* mark this field as being completed */
2399 field[i] = NULL;
2400 }
2401 }
2402
2403 /* now pick up remaining numeric fields */
2404 for (i = 0; i < nf; i++)
2405 {
2406 if (field[i] == NULL)
2407 continue;
2408
2409 if ((len = strlen(field[i])) <= 0)
2410 return DTERR_BAD_FORMAT;
2411
2412 dterr = DecodeNumber(len, field[i], haveTextMonth, fmask,
2413 &dmask, tm,
2414 &fsec, is2digits);
2415 if (dterr)
2416 return dterr;
2417
2418 if (fmask & dmask)
2419 return DTERR_BAD_FORMAT;
2420
2421 fmask |= dmask;
2422 *tmask |= dmask;
2423 }
2424
2425 if ((fmask & ~(DTK_M(DOY) | DTK_M(TZ))) != DTK_DATE_M)
2426 return DTERR_BAD_FORMAT;
2427
2428 /* validation of the field values must wait until ValidateDate() */
2429
2430 return 0;
2431 }
2432
2433 /* ValidateDate()
2434 * Check valid year/month/day values, handle BC and DOY cases
2435 * Return 0 if okay, a DTERR code if not.
2436 */
2437 int
ValidateDate(int fmask,bool isjulian,bool is2digits,bool bc,struct pg_tm * tm)2438 ValidateDate(int fmask, bool isjulian, bool is2digits, bool bc,
2439 struct pg_tm *tm)
2440 {
2441 if (fmask & DTK_M(YEAR))
2442 {
2443 if (isjulian)
2444 {
2445 /* tm_year is correct and should not be touched */
2446 }
2447 else if (bc)
2448 {
2449 /* there is no year zero in AD/BC notation */
2450 if (tm->tm_year <= 0)
2451 return DTERR_FIELD_OVERFLOW;
2452 /* internally, we represent 1 BC as year zero, 2 BC as -1, etc */
2453 tm->tm_year = -(tm->tm_year - 1);
2454 }
2455 else if (is2digits)
2456 {
2457 /* process 1 or 2-digit input as 1970-2069 AD, allow '0' and '00' */
2458 if (tm->tm_year < 0) /* just paranoia */
2459 return DTERR_FIELD_OVERFLOW;
2460 if (tm->tm_year < 70)
2461 tm->tm_year += 2000;
2462 else if (tm->tm_year < 100)
2463 tm->tm_year += 1900;
2464 }
2465 else
2466 {
2467 /* there is no year zero in AD/BC notation */
2468 if (tm->tm_year <= 0)
2469 return DTERR_FIELD_OVERFLOW;
2470 }
2471 }
2472
2473 /* now that we have correct year, decode DOY */
2474 if (fmask & DTK_M(DOY))
2475 {
2476 j2date(date2j(tm->tm_year, 1, 1) + tm->tm_yday - 1,
2477 &tm->tm_year, &tm->tm_mon, &tm->tm_mday);
2478 }
2479
2480 /* check for valid month */
2481 if (fmask & DTK_M(MONTH))
2482 {
2483 if (tm->tm_mon < 1 || tm->tm_mon > MONTHS_PER_YEAR)
2484 return DTERR_MD_FIELD_OVERFLOW;
2485 }
2486
2487 /* minimal check for valid day */
2488 if (fmask & DTK_M(DAY))
2489 {
2490 if (tm->tm_mday < 1 || tm->tm_mday > 31)
2491 return DTERR_MD_FIELD_OVERFLOW;
2492 }
2493
2494 if ((fmask & DTK_DATE_M) == DTK_DATE_M)
2495 {
2496 /*
2497 * Check for valid day of month, now that we know for sure the month
2498 * and year. Note we don't use MD_FIELD_OVERFLOW here, since it seems
2499 * unlikely that "Feb 29" is a YMD-order error.
2500 */
2501 if (tm->tm_mday > day_tab[isleap(tm->tm_year)][tm->tm_mon - 1])
2502 return DTERR_FIELD_OVERFLOW;
2503 }
2504
2505 return 0;
2506 }
2507
2508
2509 /* DecodeTime()
2510 * Decode time string which includes delimiters.
2511 * Return 0 if okay, a DTERR code if not.
2512 *
2513 * Only check the lower limit on hours, since this same code can be
2514 * used to represent time spans.
2515 */
2516 static int
DecodeTime(char * str,int fmask,int range,int * tmask,struct pg_tm * tm,fsec_t * fsec)2517 DecodeTime(char *str, int fmask, int range,
2518 int *tmask, struct pg_tm *tm, fsec_t *fsec)
2519 {
2520 char *cp;
2521 int dterr;
2522
2523 *tmask = DTK_TIME_M;
2524
2525 errno = 0;
2526 tm->tm_hour = strtoint(str, &cp, 10);
2527 if (errno == ERANGE)
2528 return DTERR_FIELD_OVERFLOW;
2529 if (*cp != ':')
2530 return DTERR_BAD_FORMAT;
2531 errno = 0;
2532 tm->tm_min = strtoint(cp + 1, &cp, 10);
2533 if (errno == ERANGE)
2534 return DTERR_FIELD_OVERFLOW;
2535 if (*cp == '\0')
2536 {
2537 tm->tm_sec = 0;
2538 *fsec = 0;
2539 /* If it's a MINUTE TO SECOND interval, take 2 fields as being mm:ss */
2540 if (range == (INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND)))
2541 {
2542 tm->tm_sec = tm->tm_min;
2543 tm->tm_min = tm->tm_hour;
2544 tm->tm_hour = 0;
2545 }
2546 }
2547 else if (*cp == '.')
2548 {
2549 /* always assume mm:ss.sss is MINUTE TO SECOND */
2550 dterr = ParseFractionalSecond(cp, fsec);
2551 if (dterr)
2552 return dterr;
2553 tm->tm_sec = tm->tm_min;
2554 tm->tm_min = tm->tm_hour;
2555 tm->tm_hour = 0;
2556 }
2557 else if (*cp == ':')
2558 {
2559 errno = 0;
2560 tm->tm_sec = strtoint(cp + 1, &cp, 10);
2561 if (errno == ERANGE)
2562 return DTERR_FIELD_OVERFLOW;
2563 if (*cp == '\0')
2564 *fsec = 0;
2565 else if (*cp == '.')
2566 {
2567 dterr = ParseFractionalSecond(cp, fsec);
2568 if (dterr)
2569 return dterr;
2570 }
2571 else
2572 return DTERR_BAD_FORMAT;
2573 }
2574 else
2575 return DTERR_BAD_FORMAT;
2576
2577 /* do a sanity check */
2578 if (tm->tm_hour < 0 || tm->tm_min < 0 || tm->tm_min > MINS_PER_HOUR - 1 ||
2579 tm->tm_sec < 0 || tm->tm_sec > SECS_PER_MINUTE ||
2580 *fsec < INT64CONST(0) ||
2581 *fsec > USECS_PER_SEC)
2582 return DTERR_FIELD_OVERFLOW;
2583
2584 return 0;
2585 }
2586
2587
2588 /* DecodeNumber()
2589 * Interpret plain numeric field as a date value in context.
2590 * Return 0 if okay, a DTERR code if not.
2591 */
2592 static int
DecodeNumber(int flen,char * str,bool haveTextMonth,int fmask,int * tmask,struct pg_tm * tm,fsec_t * fsec,bool * is2digits)2593 DecodeNumber(int flen, char *str, bool haveTextMonth, int fmask,
2594 int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2595 {
2596 int val;
2597 char *cp;
2598 int dterr;
2599
2600 *tmask = 0;
2601
2602 errno = 0;
2603 val = strtoint(str, &cp, 10);
2604 if (errno == ERANGE)
2605 return DTERR_FIELD_OVERFLOW;
2606 if (cp == str)
2607 return DTERR_BAD_FORMAT;
2608
2609 if (*cp == '.')
2610 {
2611 /*
2612 * More than two digits before decimal point? Then could be a date or
2613 * a run-together time: 2001.360 20011225 040506.789
2614 */
2615 if (cp - str > 2)
2616 {
2617 dterr = DecodeNumberField(flen, str,
2618 (fmask | DTK_DATE_M),
2619 tmask, tm,
2620 fsec, is2digits);
2621 if (dterr < 0)
2622 return dterr;
2623 return 0;
2624 }
2625
2626 dterr = ParseFractionalSecond(cp, fsec);
2627 if (dterr)
2628 return dterr;
2629 }
2630 else if (*cp != '\0')
2631 return DTERR_BAD_FORMAT;
2632
2633 /* Special case for day of year */
2634 if (flen == 3 && (fmask & DTK_DATE_M) == DTK_M(YEAR) && val >= 1 &&
2635 val <= 366)
2636 {
2637 *tmask = (DTK_M(DOY) | DTK_M(MONTH) | DTK_M(DAY));
2638 tm->tm_yday = val;
2639 /* tm_mon and tm_mday can't actually be set yet ... */
2640 return 0;
2641 }
2642
2643 /* Switch based on what we have so far */
2644 switch (fmask & DTK_DATE_M)
2645 {
2646 case 0:
2647
2648 /*
2649 * Nothing so far; make a decision about what we think the input
2650 * is. There used to be lots of heuristics here, but the
2651 * consensus now is to be paranoid. It *must* be either
2652 * YYYY-MM-DD (with a more-than-two-digit year field), or the
2653 * field order defined by DateOrder.
2654 */
2655 if (flen >= 3 || DateOrder == DATEORDER_YMD)
2656 {
2657 *tmask = DTK_M(YEAR);
2658 tm->tm_year = val;
2659 }
2660 else if (DateOrder == DATEORDER_DMY)
2661 {
2662 *tmask = DTK_M(DAY);
2663 tm->tm_mday = val;
2664 }
2665 else
2666 {
2667 *tmask = DTK_M(MONTH);
2668 tm->tm_mon = val;
2669 }
2670 break;
2671
2672 case (DTK_M(YEAR)):
2673 /* Must be at second field of YY-MM-DD */
2674 *tmask = DTK_M(MONTH);
2675 tm->tm_mon = val;
2676 break;
2677
2678 case (DTK_M(MONTH)):
2679 if (haveTextMonth)
2680 {
2681 /*
2682 * We are at the first numeric field of a date that included a
2683 * textual month name. We want to support the variants
2684 * MON-DD-YYYY, DD-MON-YYYY, and YYYY-MON-DD as unambiguous
2685 * inputs. We will also accept MON-DD-YY or DD-MON-YY in
2686 * either DMY or MDY modes, as well as YY-MON-DD in YMD mode.
2687 */
2688 if (flen >= 3 || DateOrder == DATEORDER_YMD)
2689 {
2690 *tmask = DTK_M(YEAR);
2691 tm->tm_year = val;
2692 }
2693 else
2694 {
2695 *tmask = DTK_M(DAY);
2696 tm->tm_mday = val;
2697 }
2698 }
2699 else
2700 {
2701 /* Must be at second field of MM-DD-YY */
2702 *tmask = DTK_M(DAY);
2703 tm->tm_mday = val;
2704 }
2705 break;
2706
2707 case (DTK_M(YEAR) | DTK_M(MONTH)):
2708 if (haveTextMonth)
2709 {
2710 /* Need to accept DD-MON-YYYY even in YMD mode */
2711 if (flen >= 3 && *is2digits)
2712 {
2713 /* Guess that first numeric field is day was wrong */
2714 *tmask = DTK_M(DAY); /* YEAR is already set */
2715 tm->tm_mday = tm->tm_year;
2716 tm->tm_year = val;
2717 *is2digits = false;
2718 }
2719 else
2720 {
2721 *tmask = DTK_M(DAY);
2722 tm->tm_mday = val;
2723 }
2724 }
2725 else
2726 {
2727 /* Must be at third field of YY-MM-DD */
2728 *tmask = DTK_M(DAY);
2729 tm->tm_mday = val;
2730 }
2731 break;
2732
2733 case (DTK_M(DAY)):
2734 /* Must be at second field of DD-MM-YY */
2735 *tmask = DTK_M(MONTH);
2736 tm->tm_mon = val;
2737 break;
2738
2739 case (DTK_M(MONTH) | DTK_M(DAY)):
2740 /* Must be at third field of DD-MM-YY or MM-DD-YY */
2741 *tmask = DTK_M(YEAR);
2742 tm->tm_year = val;
2743 break;
2744
2745 case (DTK_M(YEAR) | DTK_M(MONTH) | DTK_M(DAY)):
2746 /* we have all the date, so it must be a time field */
2747 dterr = DecodeNumberField(flen, str, fmask,
2748 tmask, tm,
2749 fsec, is2digits);
2750 if (dterr < 0)
2751 return dterr;
2752 return 0;
2753
2754 default:
2755 /* Anything else is bogus input */
2756 return DTERR_BAD_FORMAT;
2757 }
2758
2759 /*
2760 * When processing a year field, mark it for adjustment if it's only one
2761 * or two digits.
2762 */
2763 if (*tmask == DTK_M(YEAR))
2764 *is2digits = (flen <= 2);
2765
2766 return 0;
2767 }
2768
2769
2770 /* DecodeNumberField()
2771 * Interpret numeric string as a concatenated date or time field.
2772 * Return a DTK token (>= 0) if successful, a DTERR code (< 0) if not.
2773 *
2774 * Use the context of previously decoded fields to help with
2775 * the interpretation.
2776 */
2777 static int
DecodeNumberField(int len,char * str,int fmask,int * tmask,struct pg_tm * tm,fsec_t * fsec,bool * is2digits)2778 DecodeNumberField(int len, char *str, int fmask,
2779 int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits)
2780 {
2781 char *cp;
2782
2783 /*
2784 * Have a decimal point? Then this is a date or something with a seconds
2785 * field...
2786 */
2787 if ((cp = strchr(str, '.')) != NULL)
2788 {
2789 /*
2790 * Can we use ParseFractionalSecond here? Not clear whether trailing
2791 * junk should be rejected ...
2792 */
2793 double frac;
2794
2795 errno = 0;
2796 frac = strtod(cp, NULL);
2797 if (errno != 0)
2798 return DTERR_BAD_FORMAT;
2799 *fsec = rint(frac * 1000000);
2800 /* Now truncate off the fraction for further processing */
2801 *cp = '\0';
2802 len = strlen(str);
2803 }
2804 /* No decimal point and no complete date yet? */
2805 else if ((fmask & DTK_DATE_M) != DTK_DATE_M)
2806 {
2807 if (len >= 6)
2808 {
2809 *tmask = DTK_DATE_M;
2810
2811 /*
2812 * Start from end and consider first 2 as Day, next 2 as Month,
2813 * and the rest as Year.
2814 */
2815 tm->tm_mday = atoi(str + (len - 2));
2816 *(str + (len - 2)) = '\0';
2817 tm->tm_mon = atoi(str + (len - 4));
2818 *(str + (len - 4)) = '\0';
2819 tm->tm_year = atoi(str);
2820 if ((len - 4) == 2)
2821 *is2digits = true;
2822
2823 return DTK_DATE;
2824 }
2825 }
2826
2827 /* not all time fields are specified? */
2828 if ((fmask & DTK_TIME_M) != DTK_TIME_M)
2829 {
2830 /* hhmmss */
2831 if (len == 6)
2832 {
2833 *tmask = DTK_TIME_M;
2834 tm->tm_sec = atoi(str + 4);
2835 *(str + 4) = '\0';
2836 tm->tm_min = atoi(str + 2);
2837 *(str + 2) = '\0';
2838 tm->tm_hour = atoi(str);
2839
2840 return DTK_TIME;
2841 }
2842 /* hhmm? */
2843 else if (len == 4)
2844 {
2845 *tmask = DTK_TIME_M;
2846 tm->tm_sec = 0;
2847 tm->tm_min = atoi(str + 2);
2848 *(str + 2) = '\0';
2849 tm->tm_hour = atoi(str);
2850
2851 return DTK_TIME;
2852 }
2853 }
2854
2855 return DTERR_BAD_FORMAT;
2856 }
2857
2858
2859 /* DecodeTimezone()
2860 * Interpret string as a numeric timezone.
2861 *
2862 * Return 0 if okay (and set *tzp), a DTERR code if not okay.
2863 */
2864 int
DecodeTimezone(char * str,int * tzp)2865 DecodeTimezone(char *str, int *tzp)
2866 {
2867 int tz;
2868 int hr,
2869 min,
2870 sec = 0;
2871 char *cp;
2872
2873 /* leading character must be "+" or "-" */
2874 if (*str != '+' && *str != '-')
2875 return DTERR_BAD_FORMAT;
2876
2877 errno = 0;
2878 hr = strtoint(str + 1, &cp, 10);
2879 if (errno == ERANGE)
2880 return DTERR_TZDISP_OVERFLOW;
2881
2882 /* explicit delimiter? */
2883 if (*cp == ':')
2884 {
2885 errno = 0;
2886 min = strtoint(cp + 1, &cp, 10);
2887 if (errno == ERANGE)
2888 return DTERR_TZDISP_OVERFLOW;
2889 if (*cp == ':')
2890 {
2891 errno = 0;
2892 sec = strtoint(cp + 1, &cp, 10);
2893 if (errno == ERANGE)
2894 return DTERR_TZDISP_OVERFLOW;
2895 }
2896 }
2897 /* otherwise, might have run things together... */
2898 else if (*cp == '\0' && strlen(str) > 3)
2899 {
2900 min = hr % 100;
2901 hr = hr / 100;
2902 /* we could, but don't, support a run-together hhmmss format */
2903 }
2904 else
2905 min = 0;
2906
2907 /* Range-check the values; see notes in datatype/timestamp.h */
2908 if (hr < 0 || hr > MAX_TZDISP_HOUR)
2909 return DTERR_TZDISP_OVERFLOW;
2910 if (min < 0 || min >= MINS_PER_HOUR)
2911 return DTERR_TZDISP_OVERFLOW;
2912 if (sec < 0 || sec >= SECS_PER_MINUTE)
2913 return DTERR_TZDISP_OVERFLOW;
2914
2915 tz = (hr * MINS_PER_HOUR + min) * SECS_PER_MINUTE + sec;
2916 if (*str == '-')
2917 tz = -tz;
2918
2919 *tzp = -tz;
2920
2921 if (*cp != '\0')
2922 return DTERR_BAD_FORMAT;
2923
2924 return 0;
2925 }
2926
2927
2928 /* DecodeTimezoneAbbrev()
2929 * Interpret string as a timezone abbreviation, if possible.
2930 *
2931 * Returns an abbreviation type (TZ, DTZ, or DYNTZ), or UNKNOWN_FIELD if
2932 * string is not any known abbreviation. On success, set *offset and *tz to
2933 * represent the UTC offset (for TZ or DTZ) or underlying zone (for DYNTZ).
2934 * Note that full timezone names (such as America/New_York) are not handled
2935 * here, mostly for historical reasons.
2936 *
2937 * Given string must be lowercased already.
2938 *
2939 * Implement a cache lookup since it is likely that dates
2940 * will be related in format.
2941 */
2942 int
DecodeTimezoneAbbrev(int field,char * lowtoken,int * offset,pg_tz ** tz)2943 DecodeTimezoneAbbrev(int field, char *lowtoken,
2944 int *offset, pg_tz **tz)
2945 {
2946 int type;
2947 const datetkn *tp;
2948
2949 tp = abbrevcache[field];
2950 /* use strncmp so that we match truncated tokens */
2951 if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
2952 {
2953 if (zoneabbrevtbl)
2954 tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs,
2955 zoneabbrevtbl->numabbrevs);
2956 else
2957 tp = NULL;
2958 }
2959 if (tp == NULL)
2960 {
2961 type = UNKNOWN_FIELD;
2962 *offset = 0;
2963 *tz = NULL;
2964 }
2965 else
2966 {
2967 abbrevcache[field] = tp;
2968 type = tp->type;
2969 if (type == DYNTZ)
2970 {
2971 *offset = 0;
2972 *tz = FetchDynamicTimeZone(zoneabbrevtbl, tp);
2973 }
2974 else
2975 {
2976 *offset = tp->value;
2977 *tz = NULL;
2978 }
2979 }
2980
2981 return type;
2982 }
2983
2984
2985 /* DecodeSpecial()
2986 * Decode text string using lookup table.
2987 *
2988 * Recognizes the keywords listed in datetktbl.
2989 * Note: at one time this would also recognize timezone abbreviations,
2990 * but no more; use DecodeTimezoneAbbrev for that.
2991 *
2992 * Given string must be lowercased already.
2993 *
2994 * Implement a cache lookup since it is likely that dates
2995 * will be related in format.
2996 */
2997 int
DecodeSpecial(int field,char * lowtoken,int * val)2998 DecodeSpecial(int field, char *lowtoken, int *val)
2999 {
3000 int type;
3001 const datetkn *tp;
3002
3003 tp = datecache[field];
3004 /* use strncmp so that we match truncated tokens */
3005 if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3006 {
3007 tp = datebsearch(lowtoken, datetktbl, szdatetktbl);
3008 }
3009 if (tp == NULL)
3010 {
3011 type = UNKNOWN_FIELD;
3012 *val = 0;
3013 }
3014 else
3015 {
3016 datecache[field] = tp;
3017 type = tp->type;
3018 *val = tp->value;
3019 }
3020
3021 return type;
3022 }
3023
3024
3025 /* ClearPgTm
3026 *
3027 * Zero out a pg_tm and associated fsec_t
3028 */
3029 static inline void
ClearPgTm(struct pg_tm * tm,fsec_t * fsec)3030 ClearPgTm(struct pg_tm *tm, fsec_t *fsec)
3031 {
3032 tm->tm_year = 0;
3033 tm->tm_mon = 0;
3034 tm->tm_mday = 0;
3035 tm->tm_hour = 0;
3036 tm->tm_min = 0;
3037 tm->tm_sec = 0;
3038 *fsec = 0;
3039 }
3040
3041
3042 /* DecodeInterval()
3043 * Interpret previously parsed fields for general time interval.
3044 * Returns 0 if successful, DTERR code if bogus input detected.
3045 * dtype, tm, fsec are output parameters.
3046 *
3047 * Allow "date" field DTK_DATE since this could be just
3048 * an unsigned floating point number. - thomas 1997-11-16
3049 *
3050 * Allow ISO-style time span, with implicit units on number of days
3051 * preceding an hh:mm:ss field. - thomas 1998-04-30
3052 */
3053 int
DecodeInterval(char ** field,int * ftype,int nf,int range,int * dtype,struct pg_tm * tm,fsec_t * fsec)3054 DecodeInterval(char **field, int *ftype, int nf, int range,
3055 int *dtype, struct pg_tm *tm, fsec_t *fsec)
3056 {
3057 bool is_before = false;
3058 char *cp;
3059 int fmask = 0,
3060 tmask,
3061 type;
3062 int i;
3063 int dterr;
3064 int val;
3065 double fval;
3066
3067 *dtype = DTK_DELTA;
3068 type = IGNORE_DTF;
3069 ClearPgTm(tm, fsec);
3070
3071 /* read through list backwards to pick up units before values */
3072 for (i = nf - 1; i >= 0; i--)
3073 {
3074 switch (ftype[i])
3075 {
3076 case DTK_TIME:
3077 dterr = DecodeTime(field[i], fmask, range,
3078 &tmask, tm, fsec);
3079 if (dterr)
3080 return dterr;
3081 type = DTK_DAY;
3082 break;
3083
3084 case DTK_TZ:
3085
3086 /*
3087 * Timezone means a token with a leading sign character and at
3088 * least one digit; there could be ':', '.', '-' embedded in
3089 * it as well.
3090 */
3091 Assert(*field[i] == '-' || *field[i] == '+');
3092
3093 /*
3094 * Check for signed hh:mm or hh:mm:ss. If so, process exactly
3095 * like DTK_TIME case above, plus handling the sign.
3096 */
3097 if (strchr(field[i] + 1, ':') != NULL &&
3098 DecodeTime(field[i] + 1, fmask, range,
3099 &tmask, tm, fsec) == 0)
3100 {
3101 if (*field[i] == '-')
3102 {
3103 /* flip the sign on all fields */
3104 tm->tm_hour = -tm->tm_hour;
3105 tm->tm_min = -tm->tm_min;
3106 tm->tm_sec = -tm->tm_sec;
3107 *fsec = -(*fsec);
3108 }
3109
3110 /*
3111 * Set the next type to be a day, if units are not
3112 * specified. This handles the case of '1 +02:03' since we
3113 * are reading right to left.
3114 */
3115 type = DTK_DAY;
3116 break;
3117 }
3118
3119 /*
3120 * Otherwise, fall through to DTK_NUMBER case, which can
3121 * handle signed float numbers and signed year-month values.
3122 */
3123
3124 /* FALLTHROUGH */
3125
3126 case DTK_DATE:
3127 case DTK_NUMBER:
3128 if (type == IGNORE_DTF)
3129 {
3130 /* use typmod to decide what rightmost field is */
3131 switch (range)
3132 {
3133 case INTERVAL_MASK(YEAR):
3134 type = DTK_YEAR;
3135 break;
3136 case INTERVAL_MASK(MONTH):
3137 case INTERVAL_MASK(YEAR) | INTERVAL_MASK(MONTH):
3138 type = DTK_MONTH;
3139 break;
3140 case INTERVAL_MASK(DAY):
3141 type = DTK_DAY;
3142 break;
3143 case INTERVAL_MASK(HOUR):
3144 case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR):
3145 type = DTK_HOUR;
3146 break;
3147 case INTERVAL_MASK(MINUTE):
3148 case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3149 case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE):
3150 type = DTK_MINUTE;
3151 break;
3152 case INTERVAL_MASK(SECOND):
3153 case INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3154 case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3155 case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND):
3156 type = DTK_SECOND;
3157 break;
3158 default:
3159 type = DTK_SECOND;
3160 break;
3161 }
3162 }
3163
3164 errno = 0;
3165 val = strtoint(field[i], &cp, 10);
3166 if (errno == ERANGE)
3167 return DTERR_FIELD_OVERFLOW;
3168
3169 if (*cp == '-')
3170 {
3171 /* SQL "years-months" syntax */
3172 int val2;
3173
3174 val2 = strtoint(cp + 1, &cp, 10);
3175 if (errno == ERANGE || val2 < 0 || val2 >= MONTHS_PER_YEAR)
3176 return DTERR_FIELD_OVERFLOW;
3177 if (*cp != '\0')
3178 return DTERR_BAD_FORMAT;
3179 type = DTK_MONTH;
3180 if (*field[i] == '-')
3181 val2 = -val2;
3182 if (((double) val * MONTHS_PER_YEAR + val2) > INT_MAX ||
3183 ((double) val * MONTHS_PER_YEAR + val2) < INT_MIN)
3184 return DTERR_FIELD_OVERFLOW;
3185 val = val * MONTHS_PER_YEAR + val2;
3186 fval = 0;
3187 }
3188 else if (*cp == '.')
3189 {
3190 errno = 0;
3191 fval = strtod(cp, &cp);
3192 if (*cp != '\0' || errno != 0)
3193 return DTERR_BAD_FORMAT;
3194
3195 if (*field[i] == '-')
3196 fval = -fval;
3197 }
3198 else if (*cp == '\0')
3199 fval = 0;
3200 else
3201 return DTERR_BAD_FORMAT;
3202
3203 tmask = 0; /* DTK_M(type); */
3204
3205 switch (type)
3206 {
3207 case DTK_MICROSEC:
3208 *fsec += rint(val + fval);
3209 tmask = DTK_M(MICROSECOND);
3210 break;
3211
3212 case DTK_MILLISEC:
3213 /* avoid overflowing the fsec field */
3214 tm->tm_sec += val / 1000;
3215 val -= (val / 1000) * 1000;
3216 *fsec += rint((val + fval) * 1000);
3217 tmask = DTK_M(MILLISECOND);
3218 break;
3219
3220 case DTK_SECOND:
3221 tm->tm_sec += val;
3222 *fsec += rint(fval * 1000000);
3223
3224 /*
3225 * If any subseconds were specified, consider this
3226 * microsecond and millisecond input as well.
3227 */
3228 if (fval == 0)
3229 tmask = DTK_M(SECOND);
3230 else
3231 tmask = DTK_ALL_SECS_M;
3232 break;
3233
3234 case DTK_MINUTE:
3235 tm->tm_min += val;
3236 AdjustFractSeconds(fval, tm, fsec, SECS_PER_MINUTE);
3237 tmask = DTK_M(MINUTE);
3238 break;
3239
3240 case DTK_HOUR:
3241 tm->tm_hour += val;
3242 AdjustFractSeconds(fval, tm, fsec, SECS_PER_HOUR);
3243 tmask = DTK_M(HOUR);
3244 type = DTK_DAY; /* set for next field */
3245 break;
3246
3247 case DTK_DAY:
3248 tm->tm_mday += val;
3249 AdjustFractSeconds(fval, tm, fsec, SECS_PER_DAY);
3250 tmask = DTK_M(DAY);
3251 break;
3252
3253 case DTK_WEEK:
3254 tm->tm_mday += val * 7;
3255 AdjustFractDays(fval, tm, fsec, 7);
3256 tmask = DTK_M(WEEK);
3257 break;
3258
3259 case DTK_MONTH:
3260 tm->tm_mon += val;
3261 AdjustFractDays(fval, tm, fsec, DAYS_PER_MONTH);
3262 tmask = DTK_M(MONTH);
3263 break;
3264
3265 case DTK_YEAR:
3266 tm->tm_year += val;
3267 if (fval != 0)
3268 tm->tm_mon += fval * MONTHS_PER_YEAR;
3269 tmask = DTK_M(YEAR);
3270 break;
3271
3272 case DTK_DECADE:
3273 tm->tm_year += val * 10;
3274 if (fval != 0)
3275 tm->tm_mon += fval * MONTHS_PER_YEAR * 10;
3276 tmask = DTK_M(DECADE);
3277 break;
3278
3279 case DTK_CENTURY:
3280 tm->tm_year += val * 100;
3281 if (fval != 0)
3282 tm->tm_mon += fval * MONTHS_PER_YEAR * 100;
3283 tmask = DTK_M(CENTURY);
3284 break;
3285
3286 case DTK_MILLENNIUM:
3287 tm->tm_year += val * 1000;
3288 if (fval != 0)
3289 tm->tm_mon += fval * MONTHS_PER_YEAR * 1000;
3290 tmask = DTK_M(MILLENNIUM);
3291 break;
3292
3293 default:
3294 return DTERR_BAD_FORMAT;
3295 }
3296 break;
3297
3298 case DTK_STRING:
3299 case DTK_SPECIAL:
3300 type = DecodeUnits(i, field[i], &val);
3301 if (type == IGNORE_DTF)
3302 continue;
3303
3304 tmask = 0; /* DTK_M(type); */
3305 switch (type)
3306 {
3307 case UNITS:
3308 type = val;
3309 break;
3310
3311 case AGO:
3312 is_before = true;
3313 type = val;
3314 break;
3315
3316 case RESERV:
3317 tmask = (DTK_DATE_M | DTK_TIME_M);
3318 *dtype = val;
3319 break;
3320
3321 default:
3322 return DTERR_BAD_FORMAT;
3323 }
3324 break;
3325
3326 default:
3327 return DTERR_BAD_FORMAT;
3328 }
3329
3330 if (tmask & fmask)
3331 return DTERR_BAD_FORMAT;
3332 fmask |= tmask;
3333 }
3334
3335 /* ensure that at least one time field has been found */
3336 if (fmask == 0)
3337 return DTERR_BAD_FORMAT;
3338
3339 /* ensure fractional seconds are fractional */
3340 if (*fsec != 0)
3341 {
3342 int sec;
3343
3344 sec = *fsec / USECS_PER_SEC;
3345 *fsec -= sec * USECS_PER_SEC;
3346 tm->tm_sec += sec;
3347 }
3348
3349 /*----------
3350 * The SQL standard defines the interval literal
3351 * '-1 1:00:00'
3352 * to mean "negative 1 days and negative 1 hours", while Postgres
3353 * traditionally treats this as meaning "negative 1 days and positive
3354 * 1 hours". In SQL_STANDARD intervalstyle, we apply the leading sign
3355 * to all fields if there are no other explicit signs.
3356 *
3357 * We leave the signs alone if there are additional explicit signs.
3358 * This protects us against misinterpreting postgres-style dump output,
3359 * since the postgres-style output code has always put an explicit sign on
3360 * all fields following a negative field. But note that SQL-spec output
3361 * is ambiguous and can be misinterpreted on load! (So it's best practice
3362 * to dump in postgres style, not SQL style.)
3363 *----------
3364 */
3365 if (IntervalStyle == INTSTYLE_SQL_STANDARD && *field[0] == '-')
3366 {
3367 /* Check for additional explicit signs */
3368 bool more_signs = false;
3369
3370 for (i = 1; i < nf; i++)
3371 {
3372 if (*field[i] == '-' || *field[i] == '+')
3373 {
3374 more_signs = true;
3375 break;
3376 }
3377 }
3378
3379 if (!more_signs)
3380 {
3381 /*
3382 * Rather than re-determining which field was field[0], just force
3383 * 'em all negative.
3384 */
3385 if (*fsec > 0)
3386 *fsec = -(*fsec);
3387 if (tm->tm_sec > 0)
3388 tm->tm_sec = -tm->tm_sec;
3389 if (tm->tm_min > 0)
3390 tm->tm_min = -tm->tm_min;
3391 if (tm->tm_hour > 0)
3392 tm->tm_hour = -tm->tm_hour;
3393 if (tm->tm_mday > 0)
3394 tm->tm_mday = -tm->tm_mday;
3395 if (tm->tm_mon > 0)
3396 tm->tm_mon = -tm->tm_mon;
3397 if (tm->tm_year > 0)
3398 tm->tm_year = -tm->tm_year;
3399 }
3400 }
3401
3402 /* finally, AGO negates everything */
3403 if (is_before)
3404 {
3405 *fsec = -(*fsec);
3406 tm->tm_sec = -tm->tm_sec;
3407 tm->tm_min = -tm->tm_min;
3408 tm->tm_hour = -tm->tm_hour;
3409 tm->tm_mday = -tm->tm_mday;
3410 tm->tm_mon = -tm->tm_mon;
3411 tm->tm_year = -tm->tm_year;
3412 }
3413
3414 return 0;
3415 }
3416
3417
3418 /*
3419 * Helper functions to avoid duplicated code in DecodeISO8601Interval.
3420 *
3421 * Parse a decimal value and break it into integer and fractional parts.
3422 * Returns 0 or DTERR code.
3423 */
3424 static int
ParseISO8601Number(char * str,char ** endptr,int * ipart,double * fpart)3425 ParseISO8601Number(char *str, char **endptr, int *ipart, double *fpart)
3426 {
3427 double val;
3428
3429 if (!(isdigit((unsigned char) *str) || *str == '-' || *str == '.'))
3430 return DTERR_BAD_FORMAT;
3431 errno = 0;
3432 val = strtod(str, endptr);
3433 /* did we not see anything that looks like a double? */
3434 if (*endptr == str || errno != 0)
3435 return DTERR_BAD_FORMAT;
3436 /* watch out for overflow */
3437 if (val < INT_MIN || val > INT_MAX)
3438 return DTERR_FIELD_OVERFLOW;
3439 /* be very sure we truncate towards zero (cf dtrunc()) */
3440 if (val >= 0)
3441 *ipart = (int) floor(val);
3442 else
3443 *ipart = (int) -floor(-val);
3444 *fpart = val - *ipart;
3445 return 0;
3446 }
3447
3448 /*
3449 * Determine number of integral digits in a valid ISO 8601 number field
3450 * (we should ignore sign and any fraction part)
3451 */
3452 static int
ISO8601IntegerWidth(char * fieldstart)3453 ISO8601IntegerWidth(char *fieldstart)
3454 {
3455 /* We might have had a leading '-' */
3456 if (*fieldstart == '-')
3457 fieldstart++;
3458 return strspn(fieldstart, "0123456789");
3459 }
3460
3461
3462 /* DecodeISO8601Interval()
3463 * Decode an ISO 8601 time interval of the "format with designators"
3464 * (section 4.4.3.2) or "alternative format" (section 4.4.3.3)
3465 * Examples: P1D for 1 day
3466 * PT1H for 1 hour
3467 * P2Y6M7DT1H30M for 2 years, 6 months, 7 days 1 hour 30 min
3468 * P0002-06-07T01:30:00 the same value in alternative format
3469 *
3470 * Returns 0 if successful, DTERR code if bogus input detected.
3471 * Note: error code should be DTERR_BAD_FORMAT if input doesn't look like
3472 * ISO8601, otherwise this could cause unexpected error messages.
3473 * dtype, tm, fsec are output parameters.
3474 *
3475 * A couple exceptions from the spec:
3476 * - a week field ('W') may coexist with other units
3477 * - allows decimals in fields other than the least significant unit.
3478 */
3479 int
DecodeISO8601Interval(char * str,int * dtype,struct pg_tm * tm,fsec_t * fsec)3480 DecodeISO8601Interval(char *str,
3481 int *dtype, struct pg_tm *tm, fsec_t *fsec)
3482 {
3483 bool datepart = true;
3484 bool havefield = false;
3485
3486 *dtype = DTK_DELTA;
3487 ClearPgTm(tm, fsec);
3488
3489 if (strlen(str) < 2 || str[0] != 'P')
3490 return DTERR_BAD_FORMAT;
3491
3492 str++;
3493 while (*str)
3494 {
3495 char *fieldstart;
3496 int val;
3497 double fval;
3498 char unit;
3499 int dterr;
3500
3501 if (*str == 'T') /* T indicates the beginning of the time part */
3502 {
3503 datepart = false;
3504 havefield = false;
3505 str++;
3506 continue;
3507 }
3508
3509 fieldstart = str;
3510 dterr = ParseISO8601Number(str, &str, &val, &fval);
3511 if (dterr)
3512 return dterr;
3513
3514 /*
3515 * Note: we could step off the end of the string here. Code below
3516 * *must* exit the loop if unit == '\0'.
3517 */
3518 unit = *str++;
3519
3520 if (datepart)
3521 {
3522 switch (unit) /* before T: Y M W D */
3523 {
3524 case 'Y':
3525 tm->tm_year += val;
3526 tm->tm_mon += (fval * MONTHS_PER_YEAR);
3527 break;
3528 case 'M':
3529 tm->tm_mon += val;
3530 AdjustFractDays(fval, tm, fsec, DAYS_PER_MONTH);
3531 break;
3532 case 'W':
3533 tm->tm_mday += val * 7;
3534 AdjustFractDays(fval, tm, fsec, 7);
3535 break;
3536 case 'D':
3537 tm->tm_mday += val;
3538 AdjustFractSeconds(fval, tm, fsec, SECS_PER_DAY);
3539 break;
3540 case 'T': /* ISO 8601 4.4.3.3 Alternative Format / Basic */
3541 case '\0':
3542 if (ISO8601IntegerWidth(fieldstart) == 8 && !havefield)
3543 {
3544 tm->tm_year += val / 10000;
3545 tm->tm_mon += (val / 100) % 100;
3546 tm->tm_mday += val % 100;
3547 AdjustFractSeconds(fval, tm, fsec, SECS_PER_DAY);
3548 if (unit == '\0')
3549 return 0;
3550 datepart = false;
3551 havefield = false;
3552 continue;
3553 }
3554 /* Else fall through to extended alternative format */
3555 /* FALLTHROUGH */
3556 case '-': /* ISO 8601 4.4.3.3 Alternative Format,
3557 * Extended */
3558 if (havefield)
3559 return DTERR_BAD_FORMAT;
3560
3561 tm->tm_year += val;
3562 tm->tm_mon += (fval * MONTHS_PER_YEAR);
3563 if (unit == '\0')
3564 return 0;
3565 if (unit == 'T')
3566 {
3567 datepart = false;
3568 havefield = false;
3569 continue;
3570 }
3571
3572 dterr = ParseISO8601Number(str, &str, &val, &fval);
3573 if (dterr)
3574 return dterr;
3575 tm->tm_mon += val;
3576 AdjustFractDays(fval, tm, fsec, DAYS_PER_MONTH);
3577 if (*str == '\0')
3578 return 0;
3579 if (*str == 'T')
3580 {
3581 datepart = false;
3582 havefield = false;
3583 continue;
3584 }
3585 if (*str != '-')
3586 return DTERR_BAD_FORMAT;
3587 str++;
3588
3589 dterr = ParseISO8601Number(str, &str, &val, &fval);
3590 if (dterr)
3591 return dterr;
3592 tm->tm_mday += val;
3593 AdjustFractSeconds(fval, tm, fsec, SECS_PER_DAY);
3594 if (*str == '\0')
3595 return 0;
3596 if (*str == 'T')
3597 {
3598 datepart = false;
3599 havefield = false;
3600 continue;
3601 }
3602 return DTERR_BAD_FORMAT;
3603 default:
3604 /* not a valid date unit suffix */
3605 return DTERR_BAD_FORMAT;
3606 }
3607 }
3608 else
3609 {
3610 switch (unit) /* after T: H M S */
3611 {
3612 case 'H':
3613 tm->tm_hour += val;
3614 AdjustFractSeconds(fval, tm, fsec, SECS_PER_HOUR);
3615 break;
3616 case 'M':
3617 tm->tm_min += val;
3618 AdjustFractSeconds(fval, tm, fsec, SECS_PER_MINUTE);
3619 break;
3620 case 'S':
3621 tm->tm_sec += val;
3622 AdjustFractSeconds(fval, tm, fsec, 1);
3623 break;
3624 case '\0': /* ISO 8601 4.4.3.3 Alternative Format */
3625 if (ISO8601IntegerWidth(fieldstart) == 6 && !havefield)
3626 {
3627 tm->tm_hour += val / 10000;
3628 tm->tm_min += (val / 100) % 100;
3629 tm->tm_sec += val % 100;
3630 AdjustFractSeconds(fval, tm, fsec, 1);
3631 return 0;
3632 }
3633 /* Else fall through to extended alternative format */
3634 /* FALLTHROUGH */
3635 case ':': /* ISO 8601 4.4.3.3 Alternative Format,
3636 * Extended */
3637 if (havefield)
3638 return DTERR_BAD_FORMAT;
3639
3640 tm->tm_hour += val;
3641 AdjustFractSeconds(fval, tm, fsec, SECS_PER_HOUR);
3642 if (unit == '\0')
3643 return 0;
3644
3645 dterr = ParseISO8601Number(str, &str, &val, &fval);
3646 if (dterr)
3647 return dterr;
3648 tm->tm_min += val;
3649 AdjustFractSeconds(fval, tm, fsec, SECS_PER_MINUTE);
3650 if (*str == '\0')
3651 return 0;
3652 if (*str != ':')
3653 return DTERR_BAD_FORMAT;
3654 str++;
3655
3656 dterr = ParseISO8601Number(str, &str, &val, &fval);
3657 if (dterr)
3658 return dterr;
3659 tm->tm_sec += val;
3660 AdjustFractSeconds(fval, tm, fsec, 1);
3661 if (*str == '\0')
3662 return 0;
3663 return DTERR_BAD_FORMAT;
3664
3665 default:
3666 /* not a valid time unit suffix */
3667 return DTERR_BAD_FORMAT;
3668 }
3669 }
3670
3671 havefield = true;
3672 }
3673
3674 return 0;
3675 }
3676
3677
3678 /* DecodeUnits()
3679 * Decode text string using lookup table.
3680 *
3681 * This routine recognizes keywords associated with time interval units.
3682 *
3683 * Given string must be lowercased already.
3684 *
3685 * Implement a cache lookup since it is likely that dates
3686 * will be related in format.
3687 */
3688 int
DecodeUnits(int field,char * lowtoken,int * val)3689 DecodeUnits(int field, char *lowtoken, int *val)
3690 {
3691 int type;
3692 const datetkn *tp;
3693
3694 tp = deltacache[field];
3695 /* use strncmp so that we match truncated tokens */
3696 if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0)
3697 {
3698 tp = datebsearch(lowtoken, deltatktbl, szdeltatktbl);
3699 }
3700 if (tp == NULL)
3701 {
3702 type = UNKNOWN_FIELD;
3703 *val = 0;
3704 }
3705 else
3706 {
3707 deltacache[field] = tp;
3708 type = tp->type;
3709 *val = tp->value;
3710 }
3711
3712 return type;
3713 } /* DecodeUnits() */
3714
3715 /*
3716 * Report an error detected by one of the datetime input processing routines.
3717 *
3718 * dterr is the error code, str is the original input string, datatype is
3719 * the name of the datatype we were trying to accept.
3720 *
3721 * Note: it might seem useless to distinguish DTERR_INTERVAL_OVERFLOW and
3722 * DTERR_TZDISP_OVERFLOW from DTERR_FIELD_OVERFLOW, but SQL99 mandates three
3723 * separate SQLSTATE codes, so ...
3724 */
3725 void
DateTimeParseError(int dterr,const char * str,const char * datatype)3726 DateTimeParseError(int dterr, const char *str, const char *datatype)
3727 {
3728 switch (dterr)
3729 {
3730 case DTERR_FIELD_OVERFLOW:
3731 ereport(ERROR,
3732 (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
3733 errmsg("date/time field value out of range: \"%s\"",
3734 str)));
3735 break;
3736 case DTERR_MD_FIELD_OVERFLOW:
3737 /* <nanny>same as above, but add hint about DateStyle</nanny> */
3738 ereport(ERROR,
3739 (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW),
3740 errmsg("date/time field value out of range: \"%s\"",
3741 str),
3742 errhint("Perhaps you need a different \"datestyle\" setting.")));
3743 break;
3744 case DTERR_INTERVAL_OVERFLOW:
3745 ereport(ERROR,
3746 (errcode(ERRCODE_INTERVAL_FIELD_OVERFLOW),
3747 errmsg("interval field value out of range: \"%s\"",
3748 str)));
3749 break;
3750 case DTERR_TZDISP_OVERFLOW:
3751 ereport(ERROR,
3752 (errcode(ERRCODE_INVALID_TIME_ZONE_DISPLACEMENT_VALUE),
3753 errmsg("time zone displacement out of range: \"%s\"",
3754 str)));
3755 break;
3756 case DTERR_BAD_FORMAT:
3757 default:
3758 ereport(ERROR,
3759 (errcode(ERRCODE_INVALID_DATETIME_FORMAT),
3760 errmsg("invalid input syntax for type %s: \"%s\"",
3761 datatype, str)));
3762 break;
3763 }
3764 }
3765
3766 /* datebsearch()
3767 * Binary search -- from Knuth (6.2.1) Algorithm B. Special case like this
3768 * is WAY faster than the generic bsearch().
3769 */
3770 static const datetkn *
datebsearch(const char * key,const datetkn * base,int nel)3771 datebsearch(const char *key, const datetkn *base, int nel)
3772 {
3773 if (nel > 0)
3774 {
3775 const datetkn *last = base + nel - 1,
3776 *position;
3777 int result;
3778
3779 while (last >= base)
3780 {
3781 position = base + ((last - base) >> 1);
3782 /* precheck the first character for a bit of extra speed */
3783 result = (int) key[0] - (int) position->token[0];
3784 if (result == 0)
3785 {
3786 /* use strncmp so that we match truncated tokens */
3787 result = strncmp(key, position->token, TOKMAXLEN);
3788 if (result == 0)
3789 return position;
3790 }
3791 if (result < 0)
3792 last = position - 1;
3793 else
3794 base = position + 1;
3795 }
3796 }
3797 return NULL;
3798 }
3799
3800 /* EncodeTimezone()
3801 * Copies representation of a numeric timezone offset to str.
3802 *
3803 * Returns a pointer to the new end of string. No NUL terminator is put
3804 * there; callers are responsible for NUL terminating str themselves.
3805 */
3806 static char *
EncodeTimezone(char * str,int tz,int style)3807 EncodeTimezone(char *str, int tz, int style)
3808 {
3809 int hour,
3810 min,
3811 sec;
3812
3813 sec = abs(tz);
3814 min = sec / SECS_PER_MINUTE;
3815 sec -= min * SECS_PER_MINUTE;
3816 hour = min / MINS_PER_HOUR;
3817 min -= hour * MINS_PER_HOUR;
3818
3819 /* TZ is negated compared to sign we wish to display ... */
3820 *str++ = (tz <= 0 ? '+' : '-');
3821
3822 if (sec != 0)
3823 {
3824 str = pg_ltostr_zeropad(str, hour, 2);
3825 *str++ = ':';
3826 str = pg_ltostr_zeropad(str, min, 2);
3827 *str++ = ':';
3828 str = pg_ltostr_zeropad(str, sec, 2);
3829 }
3830 else if (min != 0 || style == USE_XSD_DATES)
3831 {
3832 str = pg_ltostr_zeropad(str, hour, 2);
3833 *str++ = ':';
3834 str = pg_ltostr_zeropad(str, min, 2);
3835 }
3836 else
3837 str = pg_ltostr_zeropad(str, hour, 2);
3838 return str;
3839 }
3840
3841 /* EncodeDateOnly()
3842 * Encode date as local time.
3843 */
3844 void
EncodeDateOnly(struct pg_tm * tm,int style,char * str)3845 EncodeDateOnly(struct pg_tm *tm, int style, char *str)
3846 {
3847 Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
3848
3849 switch (style)
3850 {
3851 case USE_ISO_DATES:
3852 case USE_XSD_DATES:
3853 /* compatible with ISO date formats */
3854 str = pg_ltostr_zeropad(str,
3855 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
3856 *str++ = '-';
3857 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
3858 *str++ = '-';
3859 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3860 break;
3861
3862 case USE_SQL_DATES:
3863 /* compatible with Oracle/Ingres date formats */
3864 if (DateOrder == DATEORDER_DMY)
3865 {
3866 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3867 *str++ = '/';
3868 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
3869 }
3870 else
3871 {
3872 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
3873 *str++ = '/';
3874 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3875 }
3876 *str++ = '/';
3877 str = pg_ltostr_zeropad(str,
3878 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
3879 break;
3880
3881 case USE_GERMAN_DATES:
3882 /* German-style date format */
3883 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3884 *str++ = '.';
3885 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
3886 *str++ = '.';
3887 str = pg_ltostr_zeropad(str,
3888 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
3889 break;
3890
3891 case USE_POSTGRES_DATES:
3892 default:
3893 /* traditional date-only style for Postgres */
3894 if (DateOrder == DATEORDER_DMY)
3895 {
3896 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3897 *str++ = '-';
3898 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
3899 }
3900 else
3901 {
3902 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
3903 *str++ = '-';
3904 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3905 }
3906 *str++ = '-';
3907 str = pg_ltostr_zeropad(str,
3908 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
3909 break;
3910 }
3911
3912 if (tm->tm_year <= 0)
3913 {
3914 memcpy(str, " BC", 3); /* Don't copy NUL */
3915 str += 3;
3916 }
3917 *str = '\0';
3918 }
3919
3920
3921 /* EncodeTimeOnly()
3922 * Encode time fields only.
3923 *
3924 * tm and fsec are the value to encode, print_tz determines whether to include
3925 * a time zone (the difference between time and timetz types), tz is the
3926 * numeric time zone offset, style is the date style, str is where to write the
3927 * output.
3928 */
3929 void
EncodeTimeOnly(struct pg_tm * tm,fsec_t fsec,bool print_tz,int tz,int style,char * str)3930 EncodeTimeOnly(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, int style, char *str)
3931 {
3932 str = pg_ltostr_zeropad(str, tm->tm_hour, 2);
3933 *str++ = ':';
3934 str = pg_ltostr_zeropad(str, tm->tm_min, 2);
3935 *str++ = ':';
3936 str = AppendSeconds(str, tm->tm_sec, fsec, MAX_TIME_PRECISION, true);
3937 if (print_tz)
3938 str = EncodeTimezone(str, tz, style);
3939 *str = '\0';
3940 }
3941
3942
3943 /* EncodeDateTime()
3944 * Encode date and time interpreted as local time.
3945 *
3946 * tm and fsec are the value to encode, print_tz determines whether to include
3947 * a time zone (the difference between timestamp and timestamptz types), tz is
3948 * the numeric time zone offset, tzn is the textual time zone, which if
3949 * specified will be used instead of tz by some styles, style is the date
3950 * style, str is where to write the output.
3951 *
3952 * Supported date styles:
3953 * Postgres - day mon hh:mm:ss yyyy tz
3954 * SQL - mm/dd/yyyy hh:mm:ss.ss tz
3955 * ISO - yyyy-mm-dd hh:mm:ss+/-tz
3956 * German - dd.mm.yyyy hh:mm:ss tz
3957 * XSD - yyyy-mm-ddThh:mm:ss.ss+/-tz
3958 */
3959 void
EncodeDateTime(struct pg_tm * tm,fsec_t fsec,bool print_tz,int tz,const char * tzn,int style,char * str)3960 EncodeDateTime(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, const char *tzn, int style, char *str)
3961 {
3962 int day;
3963
3964 Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR);
3965
3966 /*
3967 * Negative tm_isdst means we have no valid time zone translation.
3968 */
3969 if (tm->tm_isdst < 0)
3970 print_tz = false;
3971
3972 switch (style)
3973 {
3974 case USE_ISO_DATES:
3975 case USE_XSD_DATES:
3976 /* Compatible with ISO-8601 date formats */
3977 str = pg_ltostr_zeropad(str,
3978 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
3979 *str++ = '-';
3980 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
3981 *str++ = '-';
3982 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3983 *str++ = (style == USE_ISO_DATES) ? ' ' : 'T';
3984 str = pg_ltostr_zeropad(str, tm->tm_hour, 2);
3985 *str++ = ':';
3986 str = pg_ltostr_zeropad(str, tm->tm_min, 2);
3987 *str++ = ':';
3988 str = AppendTimestampSeconds(str, tm, fsec);
3989 if (print_tz)
3990 str = EncodeTimezone(str, tz, style);
3991 break;
3992
3993 case USE_SQL_DATES:
3994 /* Compatible with Oracle/Ingres date formats */
3995 if (DateOrder == DATEORDER_DMY)
3996 {
3997 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
3998 *str++ = '/';
3999 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
4000 }
4001 else
4002 {
4003 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
4004 *str++ = '/';
4005 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
4006 }
4007 *str++ = '/';
4008 str = pg_ltostr_zeropad(str,
4009 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4010 *str++ = ' ';
4011 str = pg_ltostr_zeropad(str, tm->tm_hour, 2);
4012 *str++ = ':';
4013 str = pg_ltostr_zeropad(str, tm->tm_min, 2);
4014 *str++ = ':';
4015 str = AppendTimestampSeconds(str, tm, fsec);
4016
4017 /*
4018 * Note: the uses of %.*s in this function would be risky if the
4019 * timezone names ever contain non-ASCII characters. However, all
4020 * TZ abbreviations in the IANA database are plain ASCII.
4021 */
4022 if (print_tz)
4023 {
4024 if (tzn)
4025 {
4026 sprintf(str, " %.*s", MAXTZLEN, tzn);
4027 str += strlen(str);
4028 }
4029 else
4030 str = EncodeTimezone(str, tz, style);
4031 }
4032 break;
4033
4034 case USE_GERMAN_DATES:
4035 /* German variant on European style */
4036 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
4037 *str++ = '.';
4038 str = pg_ltostr_zeropad(str, tm->tm_mon, 2);
4039 *str++ = '.';
4040 str = pg_ltostr_zeropad(str,
4041 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4042 *str++ = ' ';
4043 str = pg_ltostr_zeropad(str, tm->tm_hour, 2);
4044 *str++ = ':';
4045 str = pg_ltostr_zeropad(str, tm->tm_min, 2);
4046 *str++ = ':';
4047 str = AppendTimestampSeconds(str, tm, fsec);
4048
4049 if (print_tz)
4050 {
4051 if (tzn)
4052 {
4053 sprintf(str, " %.*s", MAXTZLEN, tzn);
4054 str += strlen(str);
4055 }
4056 else
4057 str = EncodeTimezone(str, tz, style);
4058 }
4059 break;
4060
4061 case USE_POSTGRES_DATES:
4062 default:
4063 /* Backward-compatible with traditional Postgres abstime dates */
4064 day = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday);
4065 tm->tm_wday = j2day(day);
4066 memcpy(str, days[tm->tm_wday], 3);
4067 str += 3;
4068 *str++ = ' ';
4069 if (DateOrder == DATEORDER_DMY)
4070 {
4071 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
4072 *str++ = ' ';
4073 memcpy(str, months[tm->tm_mon - 1], 3);
4074 str += 3;
4075 }
4076 else
4077 {
4078 memcpy(str, months[tm->tm_mon - 1], 3);
4079 str += 3;
4080 *str++ = ' ';
4081 str = pg_ltostr_zeropad(str, tm->tm_mday, 2);
4082 }
4083 *str++ = ' ';
4084 str = pg_ltostr_zeropad(str, tm->tm_hour, 2);
4085 *str++ = ':';
4086 str = pg_ltostr_zeropad(str, tm->tm_min, 2);
4087 *str++ = ':';
4088 str = AppendTimestampSeconds(str, tm, fsec);
4089 *str++ = ' ';
4090 str = pg_ltostr_zeropad(str,
4091 (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4);
4092
4093 if (print_tz)
4094 {
4095 if (tzn)
4096 {
4097 sprintf(str, " %.*s", MAXTZLEN, tzn);
4098 str += strlen(str);
4099 }
4100 else
4101 {
4102 /*
4103 * We have a time zone, but no string version. Use the
4104 * numeric form, but be sure to include a leading space to
4105 * avoid formatting something which would be rejected by
4106 * the date/time parser later. - thomas 2001-10-19
4107 */
4108 *str++ = ' ';
4109 str = EncodeTimezone(str, tz, style);
4110 }
4111 }
4112 break;
4113 }
4114
4115 if (tm->tm_year <= 0)
4116 {
4117 memcpy(str, " BC", 3); /* Don't copy NUL */
4118 str += 3;
4119 }
4120 *str = '\0';
4121 }
4122
4123
4124 /*
4125 * Helper functions to avoid duplicated code in EncodeInterval.
4126 */
4127
4128 /* Append an ISO-8601-style interval field, but only if value isn't zero */
4129 static char *
AddISO8601IntPart(char * cp,int value,char units)4130 AddISO8601IntPart(char *cp, int value, char units)
4131 {
4132 if (value == 0)
4133 return cp;
4134 sprintf(cp, "%d%c", value, units);
4135 return cp + strlen(cp);
4136 }
4137
4138 /* Append a postgres-style interval field, but only if value isn't zero */
4139 static char *
AddPostgresIntPart(char * cp,int value,const char * units,bool * is_zero,bool * is_before)4140 AddPostgresIntPart(char *cp, int value, const char *units,
4141 bool *is_zero, bool *is_before)
4142 {
4143 if (value == 0)
4144 return cp;
4145 sprintf(cp, "%s%s%d %s%s",
4146 (!*is_zero) ? " " : "",
4147 (*is_before && value > 0) ? "+" : "",
4148 value,
4149 units,
4150 (value != 1) ? "s" : "");
4151
4152 /*
4153 * Each nonzero field sets is_before for (only) the next one. This is a
4154 * tad bizarre but it's how it worked before...
4155 */
4156 *is_before = (value < 0);
4157 *is_zero = false;
4158 return cp + strlen(cp);
4159 }
4160
4161 /* Append a verbose-style interval field, but only if value isn't zero */
4162 static char *
AddVerboseIntPart(char * cp,int value,const char * units,bool * is_zero,bool * is_before)4163 AddVerboseIntPart(char *cp, int value, const char *units,
4164 bool *is_zero, bool *is_before)
4165 {
4166 if (value == 0)
4167 return cp;
4168 /* first nonzero value sets is_before */
4169 if (*is_zero)
4170 {
4171 *is_before = (value < 0);
4172 value = abs(value);
4173 }
4174 else if (*is_before)
4175 value = -value;
4176 sprintf(cp, " %d %s%s", value, units, (value == 1) ? "" : "s");
4177 *is_zero = false;
4178 return cp + strlen(cp);
4179 }
4180
4181
4182 /* EncodeInterval()
4183 * Interpret time structure as a delta time and convert to string.
4184 *
4185 * Support "traditional Postgres" and ISO-8601 styles.
4186 * Actually, afaik ISO does not address time interval formatting,
4187 * but this looks similar to the spec for absolute date/time.
4188 * - thomas 1998-04-30
4189 *
4190 * Actually, afaik, ISO 8601 does specify formats for "time
4191 * intervals...[of the]...format with time-unit designators", which
4192 * are pretty ugly. The format looks something like
4193 * P1Y1M1DT1H1M1.12345S
4194 * but useful for exchanging data with computers instead of humans.
4195 * - ron 2003-07-14
4196 *
4197 * And ISO's SQL 2008 standard specifies standards for
4198 * "year-month literal"s (that look like '2-3') and
4199 * "day-time literal"s (that look like ('4 5:6:7')
4200 */
4201 void
EncodeInterval(struct pg_tm * tm,fsec_t fsec,int style,char * str)4202 EncodeInterval(struct pg_tm *tm, fsec_t fsec, int style, char *str)
4203 {
4204 char *cp = str;
4205 int year = tm->tm_year;
4206 int mon = tm->tm_mon;
4207 int mday = tm->tm_mday;
4208 int hour = tm->tm_hour;
4209 int min = tm->tm_min;
4210 int sec = tm->tm_sec;
4211 bool is_before = false;
4212 bool is_zero = true;
4213
4214 /*
4215 * The sign of year and month are guaranteed to match, since they are
4216 * stored internally as "month". But we'll need to check for is_before and
4217 * is_zero when determining the signs of day and hour/minute/seconds
4218 * fields.
4219 */
4220 switch (style)
4221 {
4222 /* SQL Standard interval format */
4223 case INTSTYLE_SQL_STANDARD:
4224 {
4225 bool has_negative = year < 0 || mon < 0 ||
4226 mday < 0 || hour < 0 ||
4227 min < 0 || sec < 0 || fsec < 0;
4228 bool has_positive = year > 0 || mon > 0 ||
4229 mday > 0 || hour > 0 ||
4230 min > 0 || sec > 0 || fsec > 0;
4231 bool has_year_month = year != 0 || mon != 0;
4232 bool has_day_time = mday != 0 || hour != 0 ||
4233 min != 0 || sec != 0 || fsec != 0;
4234 bool has_day = mday != 0;
4235 bool sql_standard_value = !(has_negative && has_positive) &&
4236 !(has_year_month && has_day_time);
4237
4238 /*
4239 * SQL Standard wants only 1 "<sign>" preceding the whole
4240 * interval ... but can't do that if mixed signs.
4241 */
4242 if (has_negative && sql_standard_value)
4243 {
4244 *cp++ = '-';
4245 year = -year;
4246 mon = -mon;
4247 mday = -mday;
4248 hour = -hour;
4249 min = -min;
4250 sec = -sec;
4251 fsec = -fsec;
4252 }
4253
4254 if (!has_negative && !has_positive)
4255 {
4256 sprintf(cp, "0");
4257 }
4258 else if (!sql_standard_value)
4259 {
4260 /*
4261 * For non sql-standard interval values, force outputting
4262 * the signs to avoid ambiguities with intervals with
4263 * mixed sign components.
4264 */
4265 char year_sign = (year < 0 || mon < 0) ? '-' : '+';
4266 char day_sign = (mday < 0) ? '-' : '+';
4267 char sec_sign = (hour < 0 || min < 0 ||
4268 sec < 0 || fsec < 0) ? '-' : '+';
4269
4270 sprintf(cp, "%c%d-%d %c%d %c%d:%02d:",
4271 year_sign, abs(year), abs(mon),
4272 day_sign, abs(mday),
4273 sec_sign, abs(hour), abs(min));
4274 cp += strlen(cp);
4275 cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4276 *cp = '\0';
4277 }
4278 else if (has_year_month)
4279 {
4280 sprintf(cp, "%d-%d", year, mon);
4281 }
4282 else if (has_day)
4283 {
4284 sprintf(cp, "%d %d:%02d:", mday, hour, min);
4285 cp += strlen(cp);
4286 cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4287 *cp = '\0';
4288 }
4289 else
4290 {
4291 sprintf(cp, "%d:%02d:", hour, min);
4292 cp += strlen(cp);
4293 cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4294 *cp = '\0';
4295 }
4296 }
4297 break;
4298
4299 /* ISO 8601 "time-intervals by duration only" */
4300 case INTSTYLE_ISO_8601:
4301 /* special-case zero to avoid printing nothing */
4302 if (year == 0 && mon == 0 && mday == 0 &&
4303 hour == 0 && min == 0 && sec == 0 && fsec == 0)
4304 {
4305 sprintf(cp, "PT0S");
4306 break;
4307 }
4308 *cp++ = 'P';
4309 cp = AddISO8601IntPart(cp, year, 'Y');
4310 cp = AddISO8601IntPart(cp, mon, 'M');
4311 cp = AddISO8601IntPart(cp, mday, 'D');
4312 if (hour != 0 || min != 0 || sec != 0 || fsec != 0)
4313 *cp++ = 'T';
4314 cp = AddISO8601IntPart(cp, hour, 'H');
4315 cp = AddISO8601IntPart(cp, min, 'M');
4316 if (sec != 0 || fsec != 0)
4317 {
4318 if (sec < 0 || fsec < 0)
4319 *cp++ = '-';
4320 cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4321 *cp++ = 'S';
4322 *cp++ = '\0';
4323 }
4324 break;
4325
4326 /* Compatible with postgresql < 8.4 when DateStyle = 'iso' */
4327 case INTSTYLE_POSTGRES:
4328 cp = AddPostgresIntPart(cp, year, "year", &is_zero, &is_before);
4329
4330 /*
4331 * Ideally we should spell out "month" like we do for "year" and
4332 * "day". However, for backward compatibility, we can't easily
4333 * fix this. bjm 2011-05-24
4334 */
4335 cp = AddPostgresIntPart(cp, mon, "mon", &is_zero, &is_before);
4336 cp = AddPostgresIntPart(cp, mday, "day", &is_zero, &is_before);
4337 if (is_zero || hour != 0 || min != 0 || sec != 0 || fsec != 0)
4338 {
4339 bool minus = (hour < 0 || min < 0 || sec < 0 || fsec < 0);
4340
4341 sprintf(cp, "%s%s%02d:%02d:",
4342 is_zero ? "" : " ",
4343 (minus ? "-" : (is_before ? "+" : "")),
4344 abs(hour), abs(min));
4345 cp += strlen(cp);
4346 cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true);
4347 *cp = '\0';
4348 }
4349 break;
4350
4351 /* Compatible with postgresql < 8.4 when DateStyle != 'iso' */
4352 case INTSTYLE_POSTGRES_VERBOSE:
4353 default:
4354 strcpy(cp, "@");
4355 cp++;
4356 cp = AddVerboseIntPart(cp, year, "year", &is_zero, &is_before);
4357 cp = AddVerboseIntPart(cp, mon, "mon", &is_zero, &is_before);
4358 cp = AddVerboseIntPart(cp, mday, "day", &is_zero, &is_before);
4359 cp = AddVerboseIntPart(cp, hour, "hour", &is_zero, &is_before);
4360 cp = AddVerboseIntPart(cp, min, "min", &is_zero, &is_before);
4361 if (sec != 0 || fsec != 0)
4362 {
4363 *cp++ = ' ';
4364 if (sec < 0 || (sec == 0 && fsec < 0))
4365 {
4366 if (is_zero)
4367 is_before = true;
4368 else if (!is_before)
4369 *cp++ = '-';
4370 }
4371 else if (is_before)
4372 *cp++ = '-';
4373 cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false);
4374 sprintf(cp, " sec%s",
4375 (abs(sec) != 1 || fsec != 0) ? "s" : "");
4376 is_zero = false;
4377 }
4378 /* identically zero? then put in a unitless zero... */
4379 if (is_zero)
4380 strcat(cp, " 0");
4381 if (is_before)
4382 strcat(cp, " ago");
4383 break;
4384 }
4385 }
4386
4387
4388 /*
4389 * We've been burnt by stupid errors in the ordering of the datetkn tables
4390 * once too often. Arrange to check them during postmaster start.
4391 */
4392 static bool
CheckDateTokenTable(const char * tablename,const datetkn * base,int nel)4393 CheckDateTokenTable(const char *tablename, const datetkn *base, int nel)
4394 {
4395 bool ok = true;
4396 int i;
4397
4398 for (i = 0; i < nel; i++)
4399 {
4400 /* check for token strings that don't fit */
4401 if (strlen(base[i].token) > TOKMAXLEN)
4402 {
4403 /* %.*s is safe since all our tokens are ASCII */
4404 elog(LOG, "token too long in %s table: \"%.*s\"",
4405 tablename,
4406 TOKMAXLEN + 1, base[i].token);
4407 ok = false;
4408 break; /* don't risk applying strcmp */
4409 }
4410 /* check for out of order */
4411 if (i > 0 &&
4412 strcmp(base[i - 1].token, base[i].token) >= 0)
4413 {
4414 elog(LOG, "ordering error in %s table: \"%s\" >= \"%s\"",
4415 tablename,
4416 base[i - 1].token,
4417 base[i].token);
4418 ok = false;
4419 }
4420 }
4421 return ok;
4422 }
4423
4424 bool
CheckDateTokenTables(void)4425 CheckDateTokenTables(void)
4426 {
4427 bool ok = true;
4428
4429 Assert(UNIX_EPOCH_JDATE == date2j(1970, 1, 1));
4430 Assert(POSTGRES_EPOCH_JDATE == date2j(2000, 1, 1));
4431
4432 ok &= CheckDateTokenTable("datetktbl", datetktbl, szdatetktbl);
4433 ok &= CheckDateTokenTable("deltatktbl", deltatktbl, szdeltatktbl);
4434 return ok;
4435 }
4436
4437 /*
4438 * Common code for temporal prosupport functions: simplify, if possible,
4439 * a call to a temporal type's length-coercion function.
4440 *
4441 * Types time, timetz, timestamp and timestamptz each have a range of allowed
4442 * precisions. An unspecified precision is rigorously equivalent to the
4443 * highest specifiable precision. We can replace the function call with a
4444 * no-op RelabelType if it is coercing to the same or higher precision as the
4445 * input is known to have.
4446 *
4447 * The input Node is always a FuncExpr, but to reduce the #include footprint
4448 * of datetime.h, we declare it as Node *.
4449 *
4450 * Note: timestamp_scale throws an error when the typmod is out of range, but
4451 * we can't get there from a cast: our typmodin will have caught it already.
4452 */
4453 Node *
TemporalSimplify(int32 max_precis,Node * node)4454 TemporalSimplify(int32 max_precis, Node *node)
4455 {
4456 FuncExpr *expr = castNode(FuncExpr, node);
4457 Node *ret = NULL;
4458 Node *typmod;
4459
4460 Assert(list_length(expr->args) >= 2);
4461
4462 typmod = (Node *) lsecond(expr->args);
4463
4464 if (IsA(typmod, Const) &&!((Const *) typmod)->constisnull)
4465 {
4466 Node *source = (Node *) linitial(expr->args);
4467 int32 old_precis = exprTypmod(source);
4468 int32 new_precis = DatumGetInt32(((Const *) typmod)->constvalue);
4469
4470 if (new_precis < 0 || new_precis == max_precis ||
4471 (old_precis >= 0 && new_precis >= old_precis))
4472 ret = relabel_to_typmod(source, new_precis);
4473 }
4474
4475 return ret;
4476 }
4477
4478 /*
4479 * This function gets called during timezone config file load or reload
4480 * to create the final array of timezone tokens. The argument array
4481 * is already sorted in name order.
4482 *
4483 * The result is a TimeZoneAbbrevTable (which must be a single malloc'd chunk)
4484 * or NULL on malloc failure. No other error conditions are defined.
4485 */
4486 TimeZoneAbbrevTable *
ConvertTimeZoneAbbrevs(struct tzEntry * abbrevs,int n)4487 ConvertTimeZoneAbbrevs(struct tzEntry *abbrevs, int n)
4488 {
4489 TimeZoneAbbrevTable *tbl;
4490 Size tbl_size;
4491 int i;
4492
4493 /* Space for fixed fields and datetkn array */
4494 tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
4495 n * sizeof(datetkn);
4496 tbl_size = MAXALIGN(tbl_size);
4497 /* Count up space for dynamic abbreviations */
4498 for (i = 0; i < n; i++)
4499 {
4500 struct tzEntry *abbr = abbrevs + i;
4501
4502 if (abbr->zone != NULL)
4503 {
4504 Size dsize;
4505
4506 dsize = offsetof(DynamicZoneAbbrev, zone) +
4507 strlen(abbr->zone) + 1;
4508 tbl_size += MAXALIGN(dsize);
4509 }
4510 }
4511
4512 /* Alloc the result ... */
4513 tbl = malloc(tbl_size);
4514 if (!tbl)
4515 return NULL;
4516
4517 /* ... and fill it in */
4518 tbl->tblsize = tbl_size;
4519 tbl->numabbrevs = n;
4520 /* in this loop, tbl_size reprises the space calculation above */
4521 tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) +
4522 n * sizeof(datetkn);
4523 tbl_size = MAXALIGN(tbl_size);
4524 for (i = 0; i < n; i++)
4525 {
4526 struct tzEntry *abbr = abbrevs + i;
4527 datetkn *dtoken = tbl->abbrevs + i;
4528
4529 /* use strlcpy to truncate name if necessary */
4530 strlcpy(dtoken->token, abbr->abbrev, TOKMAXLEN + 1);
4531 if (abbr->zone != NULL)
4532 {
4533 /* Allocate a DynamicZoneAbbrev for this abbreviation */
4534 DynamicZoneAbbrev *dtza;
4535 Size dsize;
4536
4537 dtza = (DynamicZoneAbbrev *) ((char *) tbl + tbl_size);
4538 dtza->tz = NULL;
4539 strcpy(dtza->zone, abbr->zone);
4540
4541 dtoken->type = DYNTZ;
4542 /* value is offset from table start to DynamicZoneAbbrev */
4543 dtoken->value = (int32) tbl_size;
4544
4545 dsize = offsetof(DynamicZoneAbbrev, zone) +
4546 strlen(abbr->zone) + 1;
4547 tbl_size += MAXALIGN(dsize);
4548 }
4549 else
4550 {
4551 dtoken->type = abbr->is_dst ? DTZ : TZ;
4552 dtoken->value = abbr->offset;
4553 }
4554 }
4555
4556 /* Assert the two loops above agreed on size calculations */
4557 Assert(tbl->tblsize == tbl_size);
4558
4559 /* Check the ordering, if testing */
4560 Assert(CheckDateTokenTable("timezone abbreviations", tbl->abbrevs, n));
4561
4562 return tbl;
4563 }
4564
4565 /*
4566 * Install a TimeZoneAbbrevTable as the active table.
4567 *
4568 * Caller is responsible that the passed table doesn't go away while in use.
4569 */
4570 void
InstallTimeZoneAbbrevs(TimeZoneAbbrevTable * tbl)4571 InstallTimeZoneAbbrevs(TimeZoneAbbrevTable *tbl)
4572 {
4573 zoneabbrevtbl = tbl;
4574 /* reset abbrevcache, which may contain pointers into old table */
4575 memset(abbrevcache, 0, sizeof(abbrevcache));
4576 }
4577
4578 /*
4579 * Helper subroutine to locate pg_tz timezone for a dynamic abbreviation.
4580 */
4581 static pg_tz *
FetchDynamicTimeZone(TimeZoneAbbrevTable * tbl,const datetkn * tp)4582 FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp)
4583 {
4584 DynamicZoneAbbrev *dtza;
4585
4586 /* Just some sanity checks to prevent indexing off into nowhere */
4587 Assert(tp->type == DYNTZ);
4588 Assert(tp->value > 0 && tp->value < tbl->tblsize);
4589
4590 dtza = (DynamicZoneAbbrev *) ((char *) tbl + tp->value);
4591
4592 /* Look up the underlying zone if we haven't already */
4593 if (dtza->tz == NULL)
4594 {
4595 dtza->tz = pg_tzset(dtza->zone);
4596
4597 /*
4598 * Ideally we'd let the caller ereport instead of doing it here, but
4599 * then there is no way to report the bad time zone name.
4600 */
4601 if (dtza->tz == NULL)
4602 ereport(ERROR,
4603 (errcode(ERRCODE_CONFIG_FILE_ERROR),
4604 errmsg("time zone \"%s\" not recognized",
4605 dtza->zone),
4606 errdetail("This time zone name appears in the configuration file for time zone abbreviation \"%s\".",
4607 tp->token)));
4608 }
4609 return dtza->tz;
4610 }
4611
4612
4613 /*
4614 * This set-returning function reads all the available time zone abbreviations
4615 * and returns a set of (abbrev, utc_offset, is_dst).
4616 */
4617 Datum
pg_timezone_abbrevs(PG_FUNCTION_ARGS)4618 pg_timezone_abbrevs(PG_FUNCTION_ARGS)
4619 {
4620 FuncCallContext *funcctx;
4621 int *pindex;
4622 Datum result;
4623 HeapTuple tuple;
4624 Datum values[3];
4625 bool nulls[3];
4626 const datetkn *tp;
4627 char buffer[TOKMAXLEN + 1];
4628 int gmtoffset;
4629 bool is_dst;
4630 unsigned char *p;
4631 struct pg_tm tm;
4632 Interval *resInterval;
4633
4634 /* stuff done only on the first call of the function */
4635 if (SRF_IS_FIRSTCALL())
4636 {
4637 TupleDesc tupdesc;
4638 MemoryContext oldcontext;
4639
4640 /* create a function context for cross-call persistence */
4641 funcctx = SRF_FIRSTCALL_INIT();
4642
4643 /*
4644 * switch to memory context appropriate for multiple function calls
4645 */
4646 oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx);
4647
4648 /* allocate memory for user context */
4649 pindex = (int *) palloc(sizeof(int));
4650 *pindex = 0;
4651 funcctx->user_fctx = (void *) pindex;
4652
4653 /*
4654 * build tupdesc for result tuples. This must match this function's
4655 * pg_proc entry!
4656 */
4657 tupdesc = CreateTemplateTupleDesc(3);
4658 TupleDescInitEntry(tupdesc, (AttrNumber) 1, "abbrev",
4659 TEXTOID, -1, 0);
4660 TupleDescInitEntry(tupdesc, (AttrNumber) 2, "utc_offset",
4661 INTERVALOID, -1, 0);
4662 TupleDescInitEntry(tupdesc, (AttrNumber) 3, "is_dst",
4663 BOOLOID, -1, 0);
4664
4665 funcctx->tuple_desc = BlessTupleDesc(tupdesc);
4666 MemoryContextSwitchTo(oldcontext);
4667 }
4668
4669 /* stuff done on every call of the function */
4670 funcctx = SRF_PERCALL_SETUP();
4671 pindex = (int *) funcctx->user_fctx;
4672
4673 if (zoneabbrevtbl == NULL ||
4674 *pindex >= zoneabbrevtbl->numabbrevs)
4675 SRF_RETURN_DONE(funcctx);
4676
4677 tp = zoneabbrevtbl->abbrevs + *pindex;
4678
4679 switch (tp->type)
4680 {
4681 case TZ:
4682 gmtoffset = tp->value;
4683 is_dst = false;
4684 break;
4685 case DTZ:
4686 gmtoffset = tp->value;
4687 is_dst = true;
4688 break;
4689 case DYNTZ:
4690 {
4691 /* Determine the current meaning of the abbrev */
4692 pg_tz *tzp;
4693 TimestampTz now;
4694 int isdst;
4695
4696 tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp);
4697 now = GetCurrentTransactionStartTimestamp();
4698 gmtoffset = -DetermineTimeZoneAbbrevOffsetTS(now,
4699 tp->token,
4700 tzp,
4701 &isdst);
4702 is_dst = (bool) isdst;
4703 break;
4704 }
4705 default:
4706 elog(ERROR, "unrecognized timezone type %d", (int) tp->type);
4707 gmtoffset = 0; /* keep compiler quiet */
4708 is_dst = false;
4709 break;
4710 }
4711
4712 MemSet(nulls, 0, sizeof(nulls));
4713
4714 /*
4715 * Convert name to text, using upcasing conversion that is the inverse of
4716 * what ParseDateTime() uses.
4717 */
4718 strlcpy(buffer, tp->token, sizeof(buffer));
4719 for (p = (unsigned char *) buffer; *p; p++)
4720 *p = pg_toupper(*p);
4721
4722 values[0] = CStringGetTextDatum(buffer);
4723
4724 /* Convert offset (in seconds) to an interval */
4725 MemSet(&tm, 0, sizeof(struct pg_tm));
4726 tm.tm_sec = gmtoffset;
4727 resInterval = (Interval *) palloc(sizeof(Interval));
4728 tm2interval(&tm, 0, resInterval);
4729 values[1] = IntervalPGetDatum(resInterval);
4730
4731 values[2] = BoolGetDatum(is_dst);
4732
4733 (*pindex)++;
4734
4735 tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls);
4736 result = HeapTupleGetDatum(tuple);
4737
4738 SRF_RETURN_NEXT(funcctx, result);
4739 }
4740
4741 /*
4742 * This set-returning function reads all the available full time zones
4743 * and returns a set of (name, abbrev, utc_offset, is_dst).
4744 */
4745 Datum
pg_timezone_names(PG_FUNCTION_ARGS)4746 pg_timezone_names(PG_FUNCTION_ARGS)
4747 {
4748 ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo;
4749 bool randomAccess;
4750 TupleDesc tupdesc;
4751 Tuplestorestate *tupstore;
4752 pg_tzenum *tzenum;
4753 pg_tz *tz;
4754 Datum values[4];
4755 bool nulls[4];
4756 int tzoff;
4757 struct pg_tm tm;
4758 fsec_t fsec;
4759 const char *tzn;
4760 Interval *resInterval;
4761 struct pg_tm itm;
4762 MemoryContext oldcontext;
4763
4764 /* check to see if caller supports us returning a tuplestore */
4765 if (rsinfo == NULL || !IsA(rsinfo, ReturnSetInfo))
4766 ereport(ERROR,
4767 (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
4768 errmsg("set-valued function called in context that cannot accept a set")));
4769 if (!(rsinfo->allowedModes & SFRM_Materialize))
4770 ereport(ERROR,
4771 (errcode(ERRCODE_SYNTAX_ERROR),
4772 errmsg("materialize mode required, but it is not allowed in this context")));
4773
4774 /* The tupdesc and tuplestore must be created in ecxt_per_query_memory */
4775 oldcontext = MemoryContextSwitchTo(rsinfo->econtext->ecxt_per_query_memory);
4776
4777 if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE)
4778 elog(ERROR, "return type must be a row type");
4779
4780 randomAccess = (rsinfo->allowedModes & SFRM_Materialize_Random) != 0;
4781 tupstore = tuplestore_begin_heap(randomAccess, false, work_mem);
4782 rsinfo->returnMode = SFRM_Materialize;
4783 rsinfo->setResult = tupstore;
4784 rsinfo->setDesc = tupdesc;
4785
4786 MemoryContextSwitchTo(oldcontext);
4787
4788 /* initialize timezone scanning code */
4789 tzenum = pg_tzenumerate_start();
4790
4791 /* search for another zone to display */
4792 for (;;)
4793 {
4794 tz = pg_tzenumerate_next(tzenum);
4795 if (!tz)
4796 break;
4797
4798 /* Convert now() to local time in this zone */
4799 if (timestamp2tm(GetCurrentTransactionStartTimestamp(),
4800 &tzoff, &tm, &fsec, &tzn, tz) != 0)
4801 continue; /* ignore if conversion fails */
4802
4803 /*
4804 * IANA's rather silly "Factory" time zone used to emit ridiculously
4805 * long "abbreviations" such as "Local time zone must be set--see zic
4806 * manual page" or "Local time zone must be set--use tzsetup". While
4807 * modern versions of tzdb emit the much saner "-00", it seems some
4808 * benighted packagers are hacking the IANA data so that it continues
4809 * to produce these strings. To prevent producing a weirdly wide
4810 * abbrev column, reject ridiculously long abbreviations.
4811 */
4812 if (tzn && strlen(tzn) > 31)
4813 continue;
4814
4815 MemSet(nulls, 0, sizeof(nulls));
4816
4817 values[0] = CStringGetTextDatum(pg_get_timezone_name(tz));
4818 values[1] = CStringGetTextDatum(tzn ? tzn : "");
4819
4820 MemSet(&itm, 0, sizeof(struct pg_tm));
4821 itm.tm_sec = -tzoff;
4822 resInterval = (Interval *) palloc(sizeof(Interval));
4823 tm2interval(&itm, 0, resInterval);
4824 values[2] = IntervalPGetDatum(resInterval);
4825
4826 values[3] = BoolGetDatum(tm.tm_isdst > 0);
4827
4828 tuplestore_putvalues(tupstore, tupdesc, values, nulls);
4829 }
4830
4831 pg_tzenumerate_end(tzenum);
4832 return (Datum) 0;
4833 }
4834