2 * parsetime.c - parse time for at(1)
3 * Copyright (C) 1993, 1994 Thomas Koenig
5 * modifications for English-language times
6 * Copyright (C) 1993 David Parsons
8 * A lot of modifications and extensions
9 * (including the new syntax being useful for RRDB)
10 * Copyright (C) 1999 Oleg Cherevko (aka Olwi Deer)
12 * severe structural damage inflicted by Tobi Oetiker in 1999
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. The name of the author(s) may not be used to endorse or promote
20 * products derived from this software without specific prior written
23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
24 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
25 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
26 * IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
28 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
32 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 /* NOTE: nothing in here is thread-safe!!!! Not even the localtime
39 * The BNF-like specification of the time syntax parsed is below:
41 * As usual, [ X ] means that X is optional, { X } means that X may
42 * be either omitted or specified as many times as needed,
43 * alternatives are separated by |, brackets are used for grouping.
44 * (# marks the beginning of comment that extends to the end of line)
46 * TIME-SPECIFICATION ::= TIME-REFERENCE [ OFFSET-SPEC ] |
48 * ( START | END ) OFFSET-SPEC
50 * TIME-REFERENCE ::= NOW | TIME-OF-DAY-SPEC [ DAY-SPEC-1 ] |
51 * [ TIME-OF-DAY-SPEC ] DAY-SPEC-2
53 * TIME-OF-DAY-SPEC ::= NUMBER (':') NUMBER [am|pm] | # HH:MM
54 * 'noon' | 'midnight' | 'teatime'
56 * DAY-SPEC-1 ::= NUMBER '/' NUMBER '/' NUMBER | # MM/DD/[YY]YY
57 * NUMBER '.' NUMBER '.' NUMBER | # DD.MM.[YY]YY
58 * NUMBER # Seconds since 1970
61 * DAY-SPEC-2 ::= MONTH-NAME NUMBER [NUMBER] | # Month DD [YY]YY
62 * 'yesterday' | 'today' | 'tomorrow' |
66 * OFFSET-SPEC ::= '+'|'-' NUMBER TIME-UNIT { ['+'|'-'] NUMBER TIME-UNIT }
68 * TIME-UNIT ::= SECONDS | MINUTES | HOURS |
69 * DAYS | WEEKS | MONTHS | YEARS
73 * START ::= 'start' | 's'
76 * SECONDS ::= 'seconds' | 'second' | 'sec' | 's'
77 * MINUTES ::= 'minutes' | 'minute' | 'min' | 'm'
78 * HOURS ::= 'hours' | 'hour' | 'hr' | 'h'
79 * DAYS ::= 'days' | 'day' | 'd'
80 * WEEKS ::= 'weeks' | 'week' | 'wk' | 'w'
81 * MONTHS ::= 'months' | 'month' | 'mon' | 'm'
82 * YEARS ::= 'years' | 'year' | 'yr' | 'y'
84 * MONTH-NAME ::= 'jan' | 'january' | 'feb' | 'february' | 'mar' | 'march' |
85 * 'apr' | 'april' | 'may' | 'jun' | 'june' | 'jul' | 'july' |
86 * 'aug' | 'august' | 'sep' | 'september' | 'oct' | 'october' |
87 * 'nov' | 'november' | 'dec' | 'december'
89 * DAY-OF-WEEK ::= 'sunday' | 'sun' | 'monday' | 'mon' | 'tuesday' | 'tue' |
90 * 'wednesday' | 'wed' | 'thursday' | 'thu' | 'friday' | 'fri' |
94 * As you may note, there is an ambiguity with respect to
95 * the 'm' time unit (which can mean either minutes or months).
96 * To cope with this, code tries to read users mind :) by applying
97 * certain heuristics. There are two of them:
99 * 1. If 'm' is used in context of (i.e. right after the) years,
100 * months, weeks, or days it is assumed to mean months, while
101 * in the context of hours, minutes, and seconds it means minutes.
102 * (e.g., in -1y6m or +3w1m 'm' means 'months', while in
103 * -3h20m or +5s2m 'm' means 'minutes')
105 * 2. Out of context (i.e. right after the '+' or '-' sign) the
106 * meaning of 'm' is guessed from the number it directly follows.
107 * Currently, if the number absolute value is below 25 it is assumed
108 * that 'm' means months, otherwise it is treated as minutes.
109 * (e.g., -25m == -25 minutes, while +24m == +24 months)
117 #include "rrd_tool.h"
120 /* Structures and unions */
123 MIDNIGHT, NOON, TEATIME,
124 PM, AM, YESTERDAY, TODAY, TOMORROW, NOW, START, END,
125 SECONDS, MINUTES, HOURS, DAYS, WEEKS, MONTHS, YEARS,
127 NUMBER, PLUS, MINUS, DOT, COLON, SLASH, ID, JUNK,
128 JAN, FEB, MAR, APR, MAY, JUN,
129 JUL, AUG, SEP, OCT, NOV, DEC,
130 SUN, MON, TUE, WED, THU, FRI, SAT
133 /* the below is for plus_minus() */
134 #define PREVIOUS_OP (-1)
136 /* parse translation table - table driven parsers can be your FRIEND!
138 struct SpecialToken {
139 char *name; /* token name */
140 int value; /* token id */
142 static struct SpecialToken VariousWords[] = {
143 {"midnight", MIDNIGHT}, /* 00:00:00 of today or tomorrow */
144 {"noon", NOON}, /* 12:00:00 of today or tomorrow */
145 {"teatime", TEATIME}, /* 16:00:00 of today or tomorrow */
146 {"am", AM}, /* morning times for 0-12 clock */
147 {"pm", PM}, /* evening times for 0-12 clock */
148 {"tomorrow", TOMORROW},
149 {"yesterday", YESTERDAY},
196 {NULL, 0} /*** SENTINEL ***/
199 static struct SpecialToken TimeMultipliers[] = {
200 {"second", SECONDS}, /* seconds multiplier */
201 {"seconds", SECONDS}, /* (pluralized) */
202 {"sec", SECONDS}, /* (generic) */
203 {"s", SECONDS}, /* (short generic) */
204 {"minute", MINUTES}, /* minutes multiplier */
205 {"minutes", MINUTES}, /* (pluralized) */
206 {"min", MINUTES}, /* (generic) */
207 {"m", MONTHS_MINUTES}, /* (short generic) */
208 {"hour", HOURS}, /* hours ... */
209 {"hours", HOURS}, /* (pluralized) */
210 {"hr", HOURS}, /* (generic) */
211 {"h", HOURS}, /* (short generic) */
212 {"day", DAYS}, /* days ... */
213 {"days", DAYS}, /* (pluralized) */
214 {"d", DAYS}, /* (short generic) */
215 {"week", WEEKS}, /* week ... */
216 {"weeks", WEEKS}, /* (pluralized) */
217 {"wk", WEEKS}, /* (generic) */
218 {"w", WEEKS}, /* (short generic) */
219 {"month", MONTHS}, /* week ... */
220 {"months", MONTHS}, /* (pluralized) */
221 {"mon", MONTHS}, /* (generic) */
222 {"year", YEARS}, /* year ... */
223 {"years", YEARS}, /* (pluralized) */
224 {"yr", YEARS}, /* (generic) */
225 {"y", YEARS}, /* (short generic) */
226 {NULL, 0} /*** SENTINEL ***/
229 /* File scope variables */
231 /* context dependent list of specials for parser to recognize,
232 * required for us to be able distinguish between 'mon' as 'month'
233 * and 'mon' as 'monday'
235 static struct SpecialToken *Specials;
237 static const char **scp; /* scanner - pointer at arglist */
238 static char scc; /* scanner - count of remaining arguments */
239 static const char *sct; /* scanner - next char pointer in current argument */
240 static int need; /* scanner - need to advance to next argument */
242 static char *sc_token = NULL; /* scanner - token buffer */
243 static size_t sc_len; /* scanner - length of token buffer */
244 static int sc_tokid; /* scanner - token id */
246 /* Local functions */
247 static void EnsureMemFree(
250 static void EnsureMemFree(
260 * A hack to compensate for the lack of the C++ exceptions
262 * Every function func that might generate parsing "exception"
263 * should return TIME_OK (aka NULL) or pointer to the error message,
264 * and should be called like this: try(func(args));
266 * if the try is not successful it will reset the token pointer ...
268 * [NOTE: when try(...) is used as the only statement in the "if-true"
269 * part of the if statement that also has an "else" part it should be
270 * either enclosed in the curly braces (despite the fact that it looks
271 * like a single statement) or NOT followed by the ";"]
283 * The panic() function was used in the original code to die, we redefine
284 * it as macro to start the chain of ascending returns that in conjunction
285 * with the try(b) above will simulate a sort of "exception handling"
293 * ve() and e() are used to set the return error,
294 * the most appropriate use for these is inside panic(...)
296 #define MAX_ERR_MSG_LEN 1024
297 static char errmsg[MAX_ERR_MSG_LEN];
303 #ifdef HAVE_VSNPRINTF
304 vsnprintf(errmsg, MAX_ERR_MSG_LEN, fmt, ap);
306 vsprintf(errmsg, fmt, ap);
325 /* Compare S1 and S2, ignoring case, returning less than, equal to or
326 greater than zero if S1 is lexicographically less than,
327 equal to or greater than S2. -- copied from GNU libc*/
328 static int mystrcasecmp(
334 const unsigned char *p1 = (const unsigned char *) s1;
335 const unsigned char *p2 = (const unsigned char *) s2;
336 unsigned char c1, c2;
353 * parse a token, checking if it's something special to us
355 static int parse_token(
360 for (i = 0; Specials[i].name != NULL; i++)
361 if (mystrcasecmp(Specials[i].name, arg) == 0)
362 return sc_tokid = Specials[i].value;
364 /* not special - must be some random id */
365 return sc_tokid = ID;
371 * init_scanner() sets up the scanner to eat arguments
373 static char *init_scanner(
382 sc_len += strlen(*argv++);
384 sc_token = (char *) malloc(sc_len * sizeof(char));
385 if (sc_token == NULL)
386 return "Failed to allocate memory";
391 * token() fetches a token from the input stream
399 memset(sc_token, '\0', sc_len);
403 /* if we need to read another argument, walk along the argument list;
404 * when we fall off the arglist, we'll just return EOF forever
414 /* eat whitespace now - if we walk off the end of the argument,
415 * we'll continue, which puts us up at the top of the while loop
416 * to fetch the next argument in
418 while (isspace((unsigned char) *sct) || *sct == '_' || *sct == ',')
425 /* preserve the first character of the new token
427 sc_token[0] = *sct++;
429 /* then see what it is
431 if (isdigit((unsigned char) (sc_token[0]))) {
432 while (isdigit((unsigned char) (*sct)))
433 sc_token[++idx] = *sct++;
434 sc_token[++idx] = '\0';
435 return sc_tokid = NUMBER;
436 } else if (isalpha((unsigned char) (sc_token[0]))) {
437 while (isalpha((unsigned char) (*sct)))
438 sc_token[++idx] = *sct++;
439 sc_token[++idx] = '\0';
440 return parse_token(sc_token);
442 switch (sc_token[0]) {
444 return sc_tokid = COLON;
446 return sc_tokid = DOT;
448 return sc_tokid = PLUS;
450 return sc_tokid = MINUS;
452 return sc_tokid = SLASH;
454 /*OK, we did not make it ... */
456 return sc_tokid = EOF;
463 * expect2() gets a token and complains if it's not the token we want
465 static char *expect2(
472 va_start(ap, complain_fmt);
473 if (token() != desired) {
474 panic(ve(complain_fmt, ap));
483 * plus_minus() is used to parse a single NUMBER TIME-UNIT pair
484 * for the OFFSET-SPEC.
485 * It also applies those m-guessing heuristics.
487 static char *plus_minus(
488 struct rrd_time_value *ptv,
491 static int op = PLUS;
492 static int prev_multiplier = -1;
498 (NUMBER, "There should be number after '%c'",
499 op == PLUS ? '+' : '-'));
500 prev_multiplier = -1; /* reset months-minutes guessing mechanics */
502 /* if doop is < 0 then we repeat the previous op
503 * with the prefetched number */
505 delta = atoi(sc_token);
507 if (token() == MONTHS_MINUTES) {
508 /* hard job to guess what does that -5m means: -5mon or -5min? */
509 switch (prev_multiplier) {
524 if (delta < 6) /* it may be some other value but in the context
525 * of RRD who needs less than 6 min deltas? */
531 prev_multiplier = sc_tokid;
534 ptv->tm. tm_year += (
535 op == PLUS) ? delta : -delta;
540 op == PLUS) ? delta : -delta;
547 ptv->tm. tm_mday += (
548 op == PLUS) ? delta : -delta;
552 ptv->offset += (op == PLUS) ? delta * 60 * 60 : -delta * 60 * 60;
555 ptv->offset += (op == PLUS) ? delta * 60 : -delta * 60;
558 ptv->offset += (op == PLUS) ? delta : -delta;
560 default: /*default unit is seconds */
561 ptv->offset += (op == PLUS) ? delta : -delta;
564 panic(e("well-known time unit expected after %d", delta));
566 return TIME_OK; /* to make compiler happy :) */
571 * tod() computes the time of day (TIME-OF-DAY-SPEC)
574 struct rrd_time_value *ptv)
576 int hour, minute = 0;
579 /* save token status in case we must abort */
581 const char *sct_sv = sct;
582 int sc_tokid_sv = sc_tokid;
584 tlen = strlen(sc_token);
586 /* first pick out the time of day - we assume a HH (COLON|DOT) MM time
592 hour = atoi(sc_token);
595 if (sc_tokid == SLASH || sc_tokid == DOT) {
596 /* guess we are looking at a date */
599 sc_tokid = sc_tokid_sv;
600 sprintf(sc_token, "%d", hour);
603 if (sc_tokid == COLON) {
605 "Parsing HH:MM syntax, expecting MM as number, got none"));
606 minute = atoi(sc_token);
608 panic(e("parsing HH:MM syntax, got MM = %d (>59!)", minute));
613 /* check if an AM or PM specifier was given
615 if (sc_tokid == AM || sc_tokid == PM) {
617 panic(e("there cannot be more than 12 AM or PM hours"));
619 if (sc_tokid == PM) {
620 if (hour != 12) /* 12:xx PM is 12:xx, not 24:xx */
623 if (hour == 12) /* 12:xx AM is 00:xx, not 12:xx */
627 } else if (hour > 23) {
628 /* guess it was not a time then ... */
631 sc_tokid = sc_tokid_sv;
632 sprintf(sc_token, "%d", hour);
635 ptv->tm. tm_hour = hour;
636 ptv->tm. tm_min = minute;
639 if (ptv->tm.tm_hour == 24) {
640 ptv->tm. tm_hour = 0;
648 * assign_date() assigns a date, adjusting year as appropriate
650 static char *assign_date(
651 struct rrd_time_value *ptv,
660 panic(e("invalid year %d (should be either 00-99 or >1900)",
663 } else if (year >= 0 && year < 38) {
664 year += 100; /* Allow year 2000-2037 to be specified as */
666 /* 00-37 until the problem of 2038 year will */
667 /* arise for unices with 32-bit time_t :) */
669 panic(e("won't handle dates before epoch (01/01/1970), sorry"));
672 ptv->tm. tm_mday = mday;
673 ptv->tm. tm_mon = mon;
674 ptv->tm. tm_year = year;
681 * day() picks apart DAY-SPEC-[12]
684 struct rrd_time_value *ptv)
686 /* using time_t seems to help portability with 64bit oses */
687 time_t mday = 0, wday, mon, year = ptv->tm.tm_year;
695 case TODAY: /* force ourselves to stay in today - no further processing */
716 /* do month mday [year]
718 mon = (sc_tokid - JAN);
719 try(expect2(NUMBER, "the day of the month should follow month name"));
720 mday = atol(sc_token);
721 if (token() == NUMBER) {
722 year = atol(sc_token);
725 year = ptv->tm.tm_year;
727 try(assign_date(ptv, mday, mon, year));
737 /* do a particular day of the week
739 wday = (sc_tokid - SUN);
740 ptv->tm. tm_mday += (
741 wday - ptv->tm.tm_wday);
746 mday = ptv->tm.tm_mday;
747 mday += (wday - ptv->tm.tm_wday);
748 ptv->tm.tm_wday = wday;
750 try(assign_date(ptv, mday, ptv->tm.tm_mon, ptv->tm.tm_year));
755 /* get numeric <sec since 1970>, MM/DD/[YY]YY, or DD.MM.[YY]YY
757 tlen = strlen(sc_token);
758 mon = atol(sc_token);
759 if (mon > 10 * 365 * 24 * 60 * 60) {
760 ptv->tm = *localtime(&mon);
766 if (mon > 19700101 && mon < 24000101) { /*works between 1900 and 2400 */
767 char cmon[3], cmday[3], cyear[5];
769 strncpy(cyear, sc_token, 4);
772 strncpy(cmon, &(sc_token[4]), 2);
775 strncpy(cmday, &(sc_token[6]), 2);
782 if (mon <= 31 && (sc_tokid == SLASH || sc_tokid == DOT)) {
786 try(expect2(NUMBER, "there should be %s number after '%c'",
787 sep == DOT ? "month" : "day",
788 sep == DOT ? '.' : '/'));
789 mday = atol(sc_token);
790 if (token() == sep) {
792 (NUMBER, "there should be year number after '%c'",
793 sep == DOT ? '.' : '/'));
794 year = atol(sc_token);
798 /* flip months and days for European timing
810 if (mon < 0 || mon > 11) {
811 panic(e("did you really mean month %d?", mon + 1));
813 if (mday < 1 || mday > 31) {
814 panic(e("I'm afraid that %d is not a valid day of the month",
817 try(assign_date(ptv, mday, mon, year));
824 /* Global functions */
828 * parsetime() is the external interface that takes tspec, parses
829 * it and puts the result in the rrd_time_value structure *ptv.
830 * It can return either absolute times (these are ensured to be
831 * correct) or relative time references that are expected to be
832 * added to some absolute time value and then normalized by
833 * mktime() The return value is either TIME_OK (aka NULL) or
834 * the pointer to the error message in the case of problems
838 struct rrd_time_value *ptv)
840 time_t now = time(NULL);
843 /* this MUST be initialized to zero for midnight/noon/teatime */
845 Specials = VariousWords; /* initialize special words context */
847 try(init_scanner(1, &tspec));
849 /* establish the default time reference */
850 ptv->type = ABSOLUTE_TIME;
852 ptv->tm = *localtime(&now);
853 ptv->tm. tm_isdst = -1; /* mk time can figure this out for us ... */
859 break; /* jump to OFFSET-SPEC part */
862 ptv->type = RELATIVE_TO_START_TIME;
865 ptv->type = RELATIVE_TO_END_TIME;
869 ptv->tm. tm_hour = 0;
870 ptv->tm. tm_mday = 0;
872 ptv->tm. tm_year = 0;
877 int time_reference = sc_tokid;
880 if (sc_tokid == PLUS || sc_tokid == MINUS)
882 if (time_reference != NOW) {
883 panic(e("'start' or 'end' MUST be followed by +|- offset"));
884 } else if (sc_tokid != EOF) {
885 panic(e("if 'now' is followed by a token it must be +|- offset"));
890 /* Only absolute time specifications below */
893 long hour_sv = ptv->tm.tm_hour;
894 long year_sv = ptv->tm.tm_year;
896 ptv->tm. tm_hour = 30;
897 ptv->tm. tm_year = 30000;
901 if (ptv->tm.tm_hour == 30 && ptv->tm.tm_year != 30000) {
904 if (ptv->tm.tm_hour == 30) {
905 ptv->tm. tm_hour = hour_sv;
907 if (ptv->tm.tm_year == 30000) {
908 ptv->tm. tm_year = year_sv;
912 /* fix month parsing */
926 if (sc_tokid != NUMBER)
931 /* evil coding for TEATIME|NOON|MIDNIGHT - we've initialized
932 * hr to zero up above, then fall into this case in such a
933 * way so we add +12 +4 hours to it for teatime, +12 hours
934 * to it for noon, and nothing at all for midnight, then
935 * set our rettime to that hour before leaping into the
945 /* if (ptv->tm.tm_hour >= hr) {
948 } *//* shifting does not makes sense here ... noon is noon */
949 ptv->tm. tm_hour = hr;
957 panic(e("unparsable time: %s%s", sc_token, sct));
959 } /* ugly case statement */
962 * the OFFSET-SPEC part
964 * (NOTE, the sc_tokid was prefetched for us by the previous code)
966 if (sc_tokid == PLUS || sc_tokid == MINUS) {
967 Specials = TimeMultipliers; /* switch special words context */
968 while (sc_tokid == PLUS || sc_tokid == MINUS || sc_tokid == NUMBER) {
969 if (sc_tokid == NUMBER) {
970 try(plus_minus(ptv, PREVIOUS_OP));
972 try(plus_minus(ptv, sc_tokid));
973 token(); /* We will get EOF eventually but that's OK, since
974 token() will return us as many EOFs as needed */
978 /* now we should be at EOF */
979 if (sc_tokid != EOF) {
980 panic(e("unparsable trailing text: '...%s%s'", sc_token, sct));
983 ptv->tm. tm_isdst = -1; /* for mktime to guess DST status */
985 if (ptv->type == ABSOLUTE_TIME)
986 if (mktime(&ptv->tm) == -1) { /* normalize & check */
987 /* can happen for "nonexistent" times, e.g. around 3am */
988 /* when winter -> summer time correction eats a hour */
989 panic(e("the specified time is incorrect (out of range?)"));
997 struct rrd_time_value *start_tv,
998 struct rrd_time_value *end_tv,
1002 if (start_tv->type == RELATIVE_TO_END_TIME && /* same as the line above */
1003 end_tv->type == RELATIVE_TO_START_TIME) {
1004 rrd_set_error("the start and end times cannot be specified "
1005 "relative to each other");
1009 if (start_tv->type == RELATIVE_TO_START_TIME) {
1011 ("the start time cannot be specified relative to itself");
1015 if (end_tv->type == RELATIVE_TO_END_TIME) {
1016 rrd_set_error("the end time cannot be specified relative to itself");
1020 if (start_tv->type == RELATIVE_TO_END_TIME) {
1023 *end = mktime(&(end_tv->tm)) + end_tv->offset;
1024 tmtmp = *localtime(end); /* reinit end including offset */
1025 tmtmp.tm_mday += start_tv->tm.tm_mday;
1026 tmtmp.tm_mon += start_tv->tm.tm_mon;
1027 tmtmp.tm_year += start_tv->tm.tm_year;
1029 *start = mktime(&tmtmp) + start_tv->offset;
1031 *start = mktime(&(start_tv->tm)) + start_tv->offset;
1033 if (end_tv->type == RELATIVE_TO_START_TIME) {
1036 *start = mktime(&(start_tv->tm)) + start_tv->offset;
1037 tmtmp = *localtime(start);
1038 tmtmp.tm_mday += end_tv->tm.tm_mday;
1039 tmtmp.tm_mon += end_tv->tm.tm_mon;
1040 tmtmp.tm_year += end_tv->tm.tm_year;
1042 *end = mktime(&tmtmp) + end_tv->offset;
1044 *end = mktime(&(end_tv->tm)) + end_tv->offset;
1047 } /* proc_start_end */