2 * rrd_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)
121 #include "rrd_tool.h"
123 /* Structures and unions */
126 MIDNIGHT, NOON, TEATIME,
127 PM, AM, YESTERDAY, TODAY, TOMORROW, NOW, START, END,
128 SECONDS, MINUTES, HOURS, DAYS, WEEKS, MONTHS, YEARS,
130 NUMBER, PLUS, MINUS, DOT, COLON, SLASH, ID, JUNK,
131 JAN, FEB, MAR, APR, MAY, JUN,
132 JUL, AUG, SEP, OCT, NOV, DEC,
133 SUN, MON, TUE, WED, THU, FRI, SAT
136 /* the below is for plus_minus() */
137 #define PREVIOUS_OP (-1)
139 /* parse translation table - table driven parsers can be your FRIEND!
141 struct SpecialToken {
142 char *name; /* token name */
143 int value; /* token id */
145 static const struct SpecialToken VariousWords[] = {
146 {"midnight", MIDNIGHT}, /* 00:00:00 of today or tomorrow */
147 {"noon", NOON}, /* 12:00:00 of today or tomorrow */
148 {"teatime", TEATIME}, /* 16:00:00 of today or tomorrow */
149 {"am", AM}, /* morning times for 0-12 clock */
150 {"pm", PM}, /* evening times for 0-12 clock */
151 {"tomorrow", TOMORROW},
152 {"yesterday", YESTERDAY},
199 {NULL, 0} /*** SENTINEL ***/
202 static const struct SpecialToken TimeMultipliers[] = {
203 {"second", SECONDS}, /* seconds multiplier */
204 {"seconds", SECONDS}, /* (pluralized) */
205 {"sec", SECONDS}, /* (generic) */
206 {"s", SECONDS}, /* (short generic) */
207 {"minute", MINUTES}, /* minutes multiplier */
208 {"minutes", MINUTES}, /* (pluralized) */
209 {"min", MINUTES}, /* (generic) */
210 {"m", MONTHS_MINUTES}, /* (short generic) */
211 {"hour", HOURS}, /* hours ... */
212 {"hours", HOURS}, /* (pluralized) */
213 {"hr", HOURS}, /* (generic) */
214 {"h", HOURS}, /* (short generic) */
215 {"day", DAYS}, /* days ... */
216 {"days", DAYS}, /* (pluralized) */
217 {"d", DAYS}, /* (short generic) */
218 {"week", WEEKS}, /* week ... */
219 {"weeks", WEEKS}, /* (pluralized) */
220 {"wk", WEEKS}, /* (generic) */
221 {"w", WEEKS}, /* (short generic) */
222 {"month", MONTHS}, /* week ... */
223 {"months", MONTHS}, /* (pluralized) */
224 {"mon", MONTHS}, /* (generic) */
225 {"year", YEARS}, /* year ... */
226 {"years", YEARS}, /* (pluralized) */
227 {"yr", YEARS}, /* (generic) */
228 {"y", YEARS}, /* (short generic) */
229 {NULL, 0} /*** SENTINEL ***/
232 /* File scope variables */
234 /* context dependent list of specials for parser to recognize,
235 * required for us to be able distinguish between 'mon' as 'month'
236 * and 'mon' as 'monday'
238 static const struct SpecialToken *Specials;
240 static const char **scp; /* scanner - pointer at arglist */
241 static char scc; /* scanner - count of remaining arguments */
242 static const char *sct; /* scanner - next char pointer in current argument */
243 static int need; /* scanner - need to advance to next argument */
245 static char *sc_token = NULL; /* scanner - token buffer */
246 static size_t sc_len; /* scanner - length of token buffer */
247 static int sc_tokid; /* scanner - token id */
249 /* Local functions */
250 static void EnsureMemFree(
253 static void EnsureMemFree(
263 * A hack to compensate for the lack of the C++ exceptions
265 * Every function func that might generate parsing "exception"
266 * should return TIME_OK (aka NULL) or pointer to the error message,
267 * and should be called like this: try(func(args));
269 * if the try is not successful it will reset the token pointer ...
271 * [NOTE: when try(...) is used as the only statement in the "if-true"
272 * part of the if statement that also has an "else" part it should be
273 * either enclosed in the curly braces (despite the fact that it looks
274 * like a single statement) or NOT followed by the ";"]
286 * The panic() function was used in the original code to die, we redefine
287 * it as macro to start the chain of ascending returns that in conjunction
288 * with the try(b) above will simulate a sort of "exception handling"
296 * ve() and e() are used to set the return error,
297 * the most appropriate use for these is inside panic(...)
299 #define MAX_ERR_MSG_LEN 1024
300 static char errmsg[MAX_ERR_MSG_LEN];
306 #ifdef HAVE_VSNPRINTF
307 vsnprintf(errmsg, MAX_ERR_MSG_LEN, fmt, ap);
309 vsprintf(errmsg, fmt, ap);
328 /* Compare S1 and S2, ignoring case, returning less than, equal to or
329 greater than zero if S1 is lexicographically less than,
330 equal to or greater than S2. -- copied from GNU libc*/
331 static int mystrcasecmp(
335 const unsigned char *p1 = (const unsigned char *) s1;
336 const unsigned char *p2 = (const unsigned char *) s2;
337 unsigned char c1, c2;
354 * parse a token, checking if it's something special to us
356 static int parse_token(
361 for (i = 0; Specials[i].name != NULL; i++)
362 if (mystrcasecmp(Specials[i].name, arg) == 0)
363 return sc_tokid = Specials[i].value;
365 /* not special - must be some random id */
366 return sc_tokid = ID;
372 * init_scanner() sets up the scanner to eat arguments
374 static char *init_scanner(
383 sc_len += strlen(*argv++);
385 sc_token = (char *) malloc(sc_len * sizeof(char));
386 if (sc_token == NULL)
387 return "Failed to allocate memory";
392 * token() fetches a token from the input stream
400 memset(sc_token, '\0', sc_len);
404 /* if we need to read another argument, walk along the argument list;
405 * when we fall off the arglist, we'll just return EOF forever
415 /* eat whitespace now - if we walk off the end of the argument,
416 * we'll continue, which puts us up at the top of the while loop
417 * to fetch the next argument in
419 while (isspace((unsigned char) *sct) || *sct == '_' || *sct == ',')
426 /* preserve the first character of the new token
428 sc_token[0] = *sct++;
430 /* then see what it is
432 if (isdigit((unsigned char) (sc_token[0]))) {
433 while (isdigit((unsigned char) (*sct)))
434 sc_token[++idx] = *sct++;
435 sc_token[++idx] = '\0';
436 return sc_tokid = NUMBER;
437 } else if (isalpha((unsigned char) (sc_token[0]))) {
438 while (isalpha((unsigned char) (*sct)))
439 sc_token[++idx] = *sct++;
440 sc_token[++idx] = '\0';
441 return parse_token(sc_token);
443 switch (sc_token[0]) {
445 return sc_tokid = COLON;
447 return sc_tokid = DOT;
449 return sc_tokid = PLUS;
451 return sc_tokid = MINUS;
453 return sc_tokid = SLASH;
455 /*OK, we did not make it ... */
457 return sc_tokid = EOF;
464 * expect2() gets a token and complains if it's not the token we want
466 static char *expect2(
473 va_start(ap, complain_fmt);
474 if (token() != desired) {
475 panic(ve(complain_fmt, ap));
484 * plus_minus() is used to parse a single NUMBER TIME-UNIT pair
485 * for the OFFSET-SPEC.
486 * It also applies those m-guessing heuristics.
488 static char *plus_minus(
489 rrd_time_value_t * ptv,
492 static int op = PLUS;
493 static int prev_multiplier = -1;
499 (NUMBER, "There should be number after '%c'",
500 op == PLUS ? '+' : '-'));
501 prev_multiplier = -1; /* reset months-minutes guessing mechanics */
503 /* if doop is < 0 then we repeat the previous op
504 * with the prefetched number */
506 delta = atoi(sc_token);
508 if (token() == MONTHS_MINUTES) {
509 /* hard job to guess what does that -5m means: -5mon or -5min? */
510 switch (prev_multiplier) {
525 if (delta < 6) /* it may be some other value but in the context
526 * of RRD who needs less than 6 min deltas? */
532 prev_multiplier = sc_tokid;
535 ptv->tm. tm_year += (
536 op == PLUS) ? delta : -delta;
541 op == PLUS) ? delta : -delta;
548 ptv->tm. tm_mday += (
549 op == PLUS) ? delta : -delta;
553 ptv->offset += (op == PLUS) ? delta * 60 * 60 : -delta * 60 * 60;
556 ptv->offset += (op == PLUS) ? delta * 60 : -delta * 60;
559 ptv->offset += (op == PLUS) ? delta : -delta;
561 default: /*default unit is seconds */
562 ptv->offset += (op == PLUS) ? delta : -delta;
565 panic(e("well-known time unit expected after %d", delta));
567 return TIME_OK; /* to make compiler happy :) */
572 * tod() computes the time of day (TIME-OF-DAY-SPEC)
575 rrd_time_value_t * ptv)
577 int hour, minute = 0;
580 /* save token status in case we must abort */
582 const char *sct_sv = sct;
583 int sc_tokid_sv = sc_tokid;
585 tlen = strlen(sc_token);
587 /* first pick out the time of day - we assume a HH (COLON|DOT) MM time
593 hour = atoi(sc_token);
596 if (sc_tokid == SLASH || sc_tokid == DOT) {
597 /* guess we are looking at a date */
600 sc_tokid = sc_tokid_sv;
601 sprintf(sc_token, "%d", hour);
604 if (sc_tokid == COLON) {
606 "Parsing HH:MM syntax, expecting MM as number, got none"));
607 minute = atoi(sc_token);
609 panic(e("parsing HH:MM syntax, got MM = %d (>59!)", minute));
614 /* check if an AM or PM specifier was given
616 if (sc_tokid == AM || sc_tokid == PM) {
618 panic(e("there cannot be more than 12 AM or PM hours"));
620 if (sc_tokid == PM) {
621 if (hour != 12) /* 12:xx PM is 12:xx, not 24:xx */
624 if (hour == 12) /* 12:xx AM is 00:xx, not 12:xx */
628 } else if (hour > 23) {
629 /* guess it was not a time then ... */
632 sc_tokid = sc_tokid_sv;
633 sprintf(sc_token, "%d", hour);
636 ptv->tm. tm_hour = hour;
637 ptv->tm. tm_min = minute;
640 if (ptv->tm.tm_hour == 24) {
641 ptv->tm. tm_hour = 0;
649 * assign_date() assigns a date, adjusting year as appropriate
651 static char *assign_date(
652 rrd_time_value_t * ptv,
661 panic(e("invalid year %d (should be either 00-99 or >1900)",
664 } else if (year >= 0 && year < 38) {
665 year += 100; /* Allow year 2000-2037 to be specified as */
667 /* 00-37 until the problem of 2038 year will */
668 /* arise for unices with 32-bit time_t :) */
670 panic(e("won't handle dates before epoch (01/01/1970), sorry"));
673 ptv->tm. tm_mday = mday;
674 ptv->tm. tm_mon = mon;
675 ptv->tm. tm_year = year;
682 * day() picks apart DAY-SPEC-[12]
685 rrd_time_value_t * ptv)
687 /* using time_t seems to help portability with 64bit oses */
688 time_t mday = 0, wday, mon, year = ptv->tm.tm_year;
696 case TODAY: /* force ourselves to stay in today - no further processing */
717 /* do month mday [year]
719 mon = (sc_tokid - JAN);
720 try(expect2(NUMBER, "the day of the month should follow month name"));
721 mday = atol(sc_token);
722 if (token() == NUMBER) {
723 year = atol(sc_token);
726 year = ptv->tm.tm_year;
728 try(assign_date(ptv, mday, mon, year));
738 /* do a particular day of the week
740 wday = (sc_tokid - SUN);
741 ptv->tm. tm_mday += (
742 wday - ptv->tm.tm_wday);
747 mday = ptv->tm.tm_mday;
748 mday += (wday - ptv->tm.tm_wday);
749 ptv->tm.tm_wday = wday;
751 try(assign_date(ptv, mday, ptv->tm.tm_mon, ptv->tm.tm_year));
756 /* get numeric <sec since 1970>, MM/DD/[YY]YY, or DD.MM.[YY]YY
758 tlen = strlen(sc_token);
759 mon = atol(sc_token);
760 if (mon > 10 * 365 * 24 * 60 * 60) {
761 ptv->tm = *localtime(&mon);
767 if (mon > 19700101 && mon < 24000101) { /*works between 1900 and 2400 */
768 char cmon[3], cmday[3], cyear[5];
770 strncpy(cyear, sc_token, 4);
773 strncpy(cmon, &(sc_token[4]), 2);
776 strncpy(cmday, &(sc_token[6]), 2);
783 if (mon <= 31 && (sc_tokid == SLASH || sc_tokid == DOT)) {
787 try(expect2(NUMBER, "there should be %s number after '%c'",
788 sep == DOT ? "month" : "day",
789 sep == DOT ? '.' : '/'));
790 mday = atol(sc_token);
791 if (token() == sep) {
793 (NUMBER, "there should be year number after '%c'",
794 sep == DOT ? '.' : '/'));
795 year = atol(sc_token);
799 /* flip months and days for European timing
811 if (mon < 0 || mon > 11) {
812 panic(e("did you really mean month %d?", mon + 1));
814 if (mday < 1 || mday > 31) {
815 panic(e("I'm afraid that %d is not a valid day of the month",
818 try(assign_date(ptv, mday, mon, year));
825 /* Global functions */
829 * rrd_parsetime() is the external interface that takes tspec, parses
830 * it and puts the result in the rrd_time_value structure *ptv.
831 * It can return either absolute times (these are ensured to be
832 * correct) or relative time references that are expected to be
833 * added to some absolute time value and then normalized by
834 * mktime() The return value is either TIME_OK (aka NULL) or
835 * the pointer to the error message in the case of problems
839 rrd_time_value_t * ptv)
841 time_t now = time(NULL);
844 /* this MUST be initialized to zero for midnight/noon/teatime */
846 Specials = VariousWords; /* initialize special words context */
848 try(init_scanner(1, &tspec));
850 /* establish the default time reference */
851 ptv->type = ABSOLUTE_TIME;
853 ptv->tm = *localtime(&now);
854 ptv->tm. tm_isdst = -1; /* mk time can figure dst by default ... */
860 break; /* jump to OFFSET-SPEC part */
863 ptv->type = RELATIVE_TO_START_TIME;
866 ptv->type = RELATIVE_TO_END_TIME;
870 ptv->tm. tm_hour = 0;
871 ptv->tm. tm_mday = 0;
873 ptv->tm. tm_year = 0;
878 int time_reference = sc_tokid;
881 if (sc_tokid == PLUS || sc_tokid == MINUS)
883 if (time_reference != NOW) {
884 panic(e("'start' or 'end' MUST be followed by +|- offset"));
885 } else if (sc_tokid != EOF) {
886 panic(e("if 'now' is followed by a token it must be +|- offset"));
891 /* Only absolute time specifications below */
894 long hour_sv = ptv->tm.tm_hour;
895 long year_sv = ptv->tm.tm_year;
897 ptv->tm. tm_hour = 30;
898 ptv->tm. tm_year = 30000;
902 if (ptv->tm.tm_hour == 30 && ptv->tm.tm_year != 30000) {
905 if (ptv->tm.tm_hour == 30) {
906 ptv->tm. tm_hour = hour_sv;
908 if (ptv->tm.tm_year == 30000) {
909 ptv->tm. tm_year = year_sv;
913 /* fix month parsing */
927 if (sc_tokid != NUMBER)
932 /* evil coding for TEATIME|NOON|MIDNIGHT - we've initialized
933 * hr to zero up above, then fall into this case in such a
934 * way so we add +12 +4 hours to it for teatime, +12 hours
935 * to it for noon, and nothing at all for midnight, then
936 * set our rettime to that hour before leaping into the
946 /* if (ptv->tm.tm_hour >= hr) {
949 } *//* shifting does not makes sense here ... noon is noon */
950 ptv->tm. tm_hour = hr;
958 panic(e("unparsable time: %s%s", sc_token, sct));
960 } /* ugly case statement */
963 * the OFFSET-SPEC part
965 * (NOTE, the sc_tokid was prefetched for us by the previous code)
967 if (sc_tokid == PLUS || sc_tokid == MINUS) {
968 Specials = TimeMultipliers; /* switch special words context */
969 while (sc_tokid == PLUS || sc_tokid == MINUS || sc_tokid == NUMBER) {
970 if (sc_tokid == NUMBER) {
971 try(plus_minus(ptv, PREVIOUS_OP));
973 try(plus_minus(ptv, sc_tokid));
974 token(); /* We will get EOF eventually but that's OK, since
975 token() will return us as many EOFs as needed */
979 /* now we should be at EOF */
980 if (sc_tokid != EOF) {
981 panic(e("unparsable trailing text: '...%s%s'", sc_token, sct));
984 if (ptv->type == ABSOLUTE_TIME)
985 if (mktime(&ptv->tm) == -1) { /* normalize & check */
986 /* can happen for "nonexistent" times, e.g. around 3am */
987 /* when winter -> summer time correction eats a hour */
988 panic(e("the specified time is incorrect (out of range?)"));
992 } /* rrd_parsetime */
995 int rrd_proc_start_end(
996 rrd_time_value_t * start_tv,
997 rrd_time_value_t * end_tv,
1001 if (start_tv->type == RELATIVE_TO_END_TIME && /* same as the line above */
1002 end_tv->type == RELATIVE_TO_START_TIME) {
1003 rrd_set_error("the start and end times cannot be specified "
1004 "relative to each other");
1008 if (start_tv->type == RELATIVE_TO_START_TIME) {
1010 ("the start time cannot be specified relative to itself");
1014 if (end_tv->type == RELATIVE_TO_END_TIME) {
1015 rrd_set_error("the end time cannot be specified relative to itself");
1019 if (start_tv->type == RELATIVE_TO_END_TIME) {
1022 *end = mktime(&(end_tv->tm)) + end_tv->offset;
1023 tmtmp = *localtime(end); /* reinit end including offset */
1024 tmtmp.tm_mday += start_tv->tm.tm_mday;
1025 tmtmp.tm_mon += start_tv->tm.tm_mon;
1026 tmtmp.tm_year += start_tv->tm.tm_year;
1028 *start = mktime(&tmtmp) + start_tv->offset;
1030 *start = mktime(&(start_tv->tm)) + start_tv->offset;
1032 if (end_tv->type == RELATIVE_TO_START_TIME) {
1035 *start = mktime(&(start_tv->tm)) + start_tv->offset;
1036 tmtmp = *localtime(start);
1037 tmtmp.tm_mday += end_tv->tm.tm_mday;
1038 tmtmp.tm_mon += end_tv->tm.tm_mon;
1039 tmtmp.tm_year += end_tv->tm.tm_year;
1041 *end = mktime(&tmtmp) + end_tv->offset;
1043 *end = mktime(&(end_tv->tm)) + end_tv->offset;
1046 } /* rrd_proc_start_end */