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.
36 * The BNF-like specification of the time syntax parsed is below:
38 * As usual, [ X ] means that X is optional, { X } means that X may
39 * be either omitted or specified as many times as needed,
40 * alternatives are separated by |, brackets are used for grouping.
41 * (# marks the beginning of comment that extends to the end of line)
43 * TIME-SPECIFICATION ::= TIME-REFERENCE [ OFFSET-SPEC ] |
45 * ( START | END ) OFFSET-SPEC
47 * TIME-REFERENCE ::= NOW | TIME-OF-DAY-SPEC [ DAY-SPEC-1 ] |
48 * [ TIME-OF-DAY-SPEC ] DAY-SPEC-2
50 * TIME-OF-DAY-SPEC ::= NUMBER (':') NUMBER [am|pm] | # HH:MM
51 * 'noon' | 'midnight' | 'teatime'
53 * DAY-SPEC-1 ::= NUMBER '/' NUMBER '/' NUMBER | # MM/DD/[YY]YY
54 * NUMBER '.' NUMBER '.' NUMBER | # DD.MM.[YY]YY
55 * NUMBER # Seconds since 1970
58 * DAY-SPEC-2 ::= MONTH-NAME NUMBER [NUMBER] | # Month DD [YY]YY
59 * 'yesterday' | 'today' | 'tomorrow' |
63 * OFFSET-SPEC ::= '+'|'-' NUMBER TIME-UNIT { ['+'|'-'] NUMBER TIME-UNIT }
65 * TIME-UNIT ::= SECONDS | MINUTES | HOURS |
66 * DAYS | WEEKS | MONTHS | YEARS
70 * START ::= 'start' | 's'
73 * SECONDS ::= 'seconds' | 'second' | 'sec' | 's'
74 * MINUTES ::= 'minutes' | 'minute' | 'min' | 'm'
75 * HOURS ::= 'hours' | 'hour' | 'hr' | 'h'
76 * DAYS ::= 'days' | 'day' | 'd'
77 * WEEKS ::= 'weeks' | 'week' | 'wk' | 'w'
78 * MONTHS ::= 'months' | 'month' | 'mon' | 'm'
79 * YEARS ::= 'years' | 'year' | 'yr' | 'y'
81 * MONTH-NAME ::= 'jan' | 'january' | 'feb' | 'february' | 'mar' | 'march' |
82 * 'apr' | 'april' | 'may' | 'jun' | 'june' | 'jul' | 'july' |
83 * 'aug' | 'august' | 'sep' | 'september' | 'oct' | 'october' |
84 * 'nov' | 'november' | 'dec' | 'december'
86 * DAY-OF-WEEK ::= 'sunday' | 'sun' | 'monday' | 'mon' | 'tuesday' | 'tue' |
87 * 'wednesday' | 'wed' | 'thursday' | 'thu' | 'friday' | 'fri' |
91 * As you may note, there is an ambiguity with respect to
92 * the 'm' time unit (which can mean either minutes or months).
93 * To cope with this, code tries to read users mind :) by applying
94 * certain heuristics. There are two of them:
96 * 1. If 'm' is used in context of (i.e. right after the) years,
97 * months, weeks, or days it is assumed to mean months, while
98 * in the context of hours, minutes, and seconds it means minutes.
99 * (e.g., in -1y6m or +3w1m 'm' means 'months', while in
100 * -3h20m or +5s2m 'm' means 'minutes')
102 * 2. Out of context (i.e. right after the '+' or '-' sign) the
103 * meaning of 'm' is guessed from the number it directly follows.
104 * Currently, if the number absolute value is below 25 it is assumed
105 * that 'm' means months, otherwise it is treated as minutes.
106 * (e.g., -25m == -25 minutes, while +24m == +24 months)
114 #include "rrd_tool.h"
117 /* Structures and unions */
120 MIDNIGHT, NOON, TEATIME,
121 PM, AM, YESTERDAY, TODAY, TOMORROW, NOW, START, END,
122 SECONDS, MINUTES, HOURS, DAYS, WEEKS, MONTHS, YEARS,
124 NUMBER, PLUS, MINUS, DOT, COLON, SLASH, ID, JUNK,
125 JAN, FEB, MAR, APR, MAY, JUN,
126 JUL, AUG, SEP, OCT, NOV, DEC,
127 SUN, MON, TUE, WED, THU, FRI, SAT
130 /* the below is for plus_minus() */
131 #define PREVIOUS_OP (-1)
133 /* parse translation table - table driven parsers can be your FRIEND!
135 struct SpecialToken {
136 char *name; /* token name */
137 int value; /* token id */
139 static struct SpecialToken VariousWords[] = {
140 { "midnight", MIDNIGHT }, /* 00:00:00 of today or tomorrow */
141 { "noon", NOON }, /* 12:00:00 of today or tomorrow */
142 { "teatime", TEATIME }, /* 16:00:00 of today or tomorrow */
143 { "am", AM }, /* morning times for 0-12 clock */
144 { "pm", PM }, /* evening times for 0-12 clock */
145 { "tomorrow", TOMORROW },
146 { "yesterday", YESTERDAY },
175 { "september", SEP },
185 { "wednesday", WED },
193 { NULL, 0 } /*** SENTINEL ***/
196 static struct SpecialToken TimeMultipliers[] = {
197 { "second", SECONDS }, /* seconds multiplier */
198 { "seconds", SECONDS }, /* (pluralized) */
199 { "sec", SECONDS }, /* (generic) */
200 { "s", SECONDS }, /* (short generic) */
201 { "minute", MINUTES }, /* minutes multiplier */
202 { "minutes", MINUTES }, /* (pluralized) */
203 { "min", MINUTES }, /* (generic) */
204 { "m", MONTHS_MINUTES }, /* (short generic) */
205 { "hour", HOURS }, /* hours ... */
206 { "hours", HOURS }, /* (pluralized) */
207 { "hr", HOURS }, /* (generic) */
208 { "h", HOURS }, /* (short generic) */
209 { "day", DAYS }, /* days ... */
210 { "days", DAYS }, /* (pluralized) */
211 { "d", DAYS }, /* (short generic) */
212 { "week", WEEKS }, /* week ... */
213 { "weeks", WEEKS }, /* (pluralized) */
214 { "wk", WEEKS }, /* (generic) */
215 { "w", WEEKS }, /* (short generic) */
216 { "month", MONTHS }, /* week ... */
217 { "months", MONTHS }, /* (pluralized) */
218 { "mon", MONTHS }, /* (generic) */
219 { "year", YEARS }, /* year ... */
220 { "years", YEARS }, /* (pluralized) */
221 { "yr", YEARS }, /* (generic) */
222 { "y", YEARS }, /* (short generic) */
223 { NULL, 0 } /*** SENTINEL ***/
226 /* File scope variables */
228 /* context dependant list of specials for parser to recognize,
229 * required for us to be able distinguish between 'mon' as 'month'
230 * and 'mon' as 'monday'
232 static struct SpecialToken *Specials;
234 static char **scp; /* scanner - pointer at arglist */
235 static char scc; /* scanner - count of remaining arguments */
236 static char *sct; /* scanner - next char pointer in current argument */
237 static int need; /* scanner - need to advance to next argument */
239 static char *sc_token=NULL; /* scanner - token buffer */
240 static size_t sc_len; /* scanner - lenght of token buffer */
241 static int sc_tokid; /* scanner - token id */
243 static int need_to_free = 0; /* means that we need deallocating memory */
245 /* Local functions */
247 void EnsureMemFree ()
257 * A hack to compensate for the lack of the C++ exceptions
259 * Every function func that might generate parsing "exception"
260 * should return TIME_OK (aka NULL) or pointer to the error message,
261 * and should be called like this: try(func(args));
263 * if the try is not successfull it will reset the token pointer ...
265 * [NOTE: when try(...) is used as the only statement in the "if-true"
266 * part of the if statement that also has an "else" part it should be
267 * either enclosed in the curly braces (despite the fact that it looks
268 * like a single statement) or NOT follwed by the ";"]
280 * The panic() function was used in the original code to die, we redefine
281 * it as macro to start the chain of ascending returns that in conjunction
282 * with the try(b) above will simulate a sort of "exception handling"
290 * ve() and e() are used to set the return error,
291 * the most aprropriate use for these is inside panic(...)
293 #define MAX_ERR_MSG_LEN 1024
294 static char errmsg[ MAX_ERR_MSG_LEN ];
297 ve ( char *fmt, va_list ap )
299 #ifdef HAVE_VSNPRINTF
300 vsnprintf( errmsg, MAX_ERR_MSG_LEN, fmt, ap );
302 vsprintf( errmsg, fmt, ap );
319 /* Compare S1 and S2, ignoring case, returning less than, equal to or
320 greater than zero if S1 is lexiographically less than,
321 equal to or greater than S2. -- copied from GNU libc*/
323 mystrcasecmp (s1, s2)
327 const unsigned char *p1 = (const unsigned char *) s1;
328 const unsigned char *p2 = (const unsigned char *) s2;
329 unsigned char c1, c2;
336 c1 = tolower (*p1++);
337 c2 = tolower (*p2++);
347 * parse a token, checking if it's something special to us
350 parse_token(char *arg)
354 for (i=0; Specials[i].name != NULL; i++)
355 if (mystrcasecmp(Specials[i].name, arg) == 0)
356 return sc_tokid = Specials[i].value;
358 /* not special - must be some random id */
359 return sc_tokid = ID;
365 * init_scanner() sets up the scanner to eat arguments
368 init_scanner(int argc, char **argv)
375 sc_len += strlen(*argv++);
377 sc_token = (char *) malloc(sc_len*sizeof(char));
378 if( sc_token == NULL )
379 return "Failed to allocate memory";
385 * token() fetches a token from the input stream
393 memset(sc_token, '\0', sc_len);
397 /* if we need to read another argument, walk along the argument list;
398 * when we fall off the arglist, we'll just return EOF forever
408 /* eat whitespace now - if we walk off the end of the argument,
409 * we'll continue, which puts us up at the top of the while loop
410 * to fetch the next argument in
412 while (isspace((unsigned char)*sct) || *sct == '_' || *sct == ',' )
419 /* preserve the first character of the new token
421 sc_token[0] = *sct++;
423 /* then see what it is
425 if (isdigit((unsigned char)(sc_token[0]))) {
426 while (isdigit((unsigned char)(*sct)))
427 sc_token[++idx] = *sct++;
428 sc_token[++idx] = '\0';
429 return sc_tokid = NUMBER;
431 else if (isalpha((unsigned char)(sc_token[0]))) {
432 while (isalpha((unsigned char)(*sct)))
433 sc_token[++idx] = *sct++;
434 sc_token[++idx] = '\0';
435 return parse_token(sc_token);
437 else switch(sc_token[0]) {
438 case ':': return sc_tokid = COLON;
439 case '.': return sc_tokid = DOT;
440 case '+': return sc_tokid = PLUS;
441 case '-': return sc_tokid = MINUS;
442 case '/': return sc_tokid = SLASH;
444 /*OK, we did not make it ... */
446 return sc_tokid = EOF;
453 * expect() gets a token and complins if it's not the token we want
456 expect(int desired, char *complain_fmt, ...)
459 va_start( ap, complain_fmt );
460 if (token() != desired) {
461 panic(ve( complain_fmt, ap ));
470 * plus_minus() is used to parse a single NUMBER TIME-UNIT pair
471 * for the OFFSET-SPEC.
472 * It allso applies those m-guessing euristics.
475 plus_minus(struct time_value *ptv, int doop)
477 static int op = PLUS;
478 static int prev_multiplier = -1;
484 try(expect(NUMBER,"There should be number after '%c'", op == PLUS ? '+' : '-'));
485 prev_multiplier = -1; /* reset months-minutes guessing mechanics */
487 /* if doop is < 0 then we repeat the previous op
488 * with the prefetched number */
490 delta = atoi(sc_token);
492 if( token() == MONTHS_MINUTES )
494 /* hard job to guess what does that -5m means: -5mon or -5min? */
495 switch(prev_multiplier)
511 if( delta < 6 ) /* it may be some other value but in the context
512 * of RRD who needs less than 6 min deltas? */
518 prev_multiplier = sc_tokid;
521 ptv->tm.tm_year += (op == PLUS) ? delta : -delta;
524 ptv->tm.tm_mon += (op == PLUS) ? delta : -delta;
530 ptv->tm.tm_mday += (op == PLUS) ? delta : -delta;
533 ptv->offset += (op == PLUS) ? delta*60*60 : -delta*60*60;
536 ptv->offset += (op == PLUS) ? delta*60 : -delta*60;
539 ptv->offset += (op == PLUS) ? delta : -delta;
541 default: /*default unit is seconds */
542 ptv->offset += (op == PLUS) ? delta : -delta;
545 panic(e("well-known time unit expected after %d", delta));
547 return TIME_OK; /* to make compiler happy :) */
552 * tod() computes the time of day (TIME-OF-DAY-SPEC)
555 tod(struct time_value *ptv)
557 int hour, minute = 0;
559 /* save token status in case we must abort */
562 int sc_tokid_sv = sc_tokid;
564 tlen = strlen(sc_token);
566 /* first pick out the time of day - we assume a HH (COLON|DOT) MM time
572 hour = atoi(sc_token);
575 if (sc_tokid == SLASH || sc_tokid == DOT) {
576 /* guess we are looking at a date */
579 sc_tokid = sc_tokid_sv;
580 sprintf (sc_token,"%d", hour);
583 if (sc_tokid == COLON ) {
585 "Parsing HH:MM syntax, expecting MM as number, got none"));
586 minute = atoi(sc_token);
588 panic(e("parsing HH:MM syntax, got MM = %d (>59!)", minute ));
593 /* check if an AM or PM specifier was given
595 if (sc_tokid == AM || sc_tokid == PM) {
597 panic(e("there cannot be more than 12 AM or PM hours"));
599 if (sc_tokid == PM) {
600 if (hour != 12) /* 12:xx PM is 12:xx, not 24:xx */
603 if (hour == 12) /* 12:xx AM is 00:xx, not 12:xx */
608 else if (hour > 23) {
609 /* guess it was not a time then ... */
612 sc_tokid = sc_tokid_sv;
613 sprintf (sc_token,"%d", hour);
616 ptv->tm.tm_hour = hour;
617 ptv->tm.tm_min = minute;
619 if (ptv->tm.tm_hour == 24) {
628 * assign_date() assigns a date, adjusting year as appropriate
631 assign_date(struct time_value *ptv, long mday, long mon, long year)
637 panic(e("invalid year %d (should be either 00-99 or >1900)",
640 } else if( year >= 0 && year < 38 ) {
641 year += 100; /* Allow year 2000-2037 to be specified as */
642 } /* 00-37 until the problem of 2038 year will */
643 /* arise for unices with 32-bit time_t :) */
645 panic(e("won't handle dates before epoch (01/01/1970), sorry"));
648 ptv->tm.tm_mday = mday;
649 ptv->tm.tm_mon = mon;
650 ptv->tm.tm_year = year;
656 * day() picks apart DAY-SPEC-[12]
659 day(struct time_value *ptv)
661 long mday=0, wday, mon, year = ptv->tm.tm_year;
668 case TODAY: /* force ourselves to stay in today - no further processing */
676 case JAN: case FEB: case MAR: case APR: case MAY: case JUN:
677 case JUL: case AUG: case SEP: case OCT: case NOV: case DEC:
678 /* do month mday [year]
680 mon = (sc_tokid-JAN);
682 "the day of the month should follow month name"));
683 mday = atol(sc_token);
684 if (token() == NUMBER) {
685 year = atol(sc_token);
689 year = ptv->tm.tm_year;
690 try(assign_date(ptv, mday, mon, year));
693 case SUN: case MON: case TUE:
694 case WED: case THU: case FRI:
696 /* do a particular day of the week
698 wday = (sc_tokid-SUN);
699 ptv->tm.tm_mday += (wday - ptv->tm.tm_wday);
702 mday = ptv->tm.tm_mday;
703 mday += (wday - ptv->tm.tm_wday);
704 ptv->tm.tm_wday = wday;
706 try(assign_date(ptv, mday, ptv->tm.tm_mon, ptv->tm.tm_year));
711 /* get numeric <sec since 1970>, MM/DD/[YY]YY, or DD.MM.[YY]YY
713 tlen = strlen(sc_token);
714 mon = atol(sc_token);
715 if (mon > 10*356*24*60*60) {
716 ptv->tm=*localtime(&mon);
721 if (mon > 19700101 && mon < 24000101){ /*works between 1900 and 2400 */
722 char cmon[3],cmday[3],cyear[5];
723 strncpy(cyear,sc_token,4);cyear[4]='\0';
725 strncpy(cmon,&(sc_token[4]),2);cmon[2]='\0';
727 strncpy(cmday,&(sc_token[6]),2);cmday[2]='\0';
733 if (mon <= 31 && (sc_tokid == SLASH || sc_tokid == DOT)) {
736 try(expect(NUMBER,"there should be %s number after '%c'",
737 sep == DOT ? "month" : "day", sep == DOT ? '.' : '/'));
738 mday = atol(sc_token);
739 if (token() == sep) {
740 try(expect(NUMBER,"there should be year number after '%c'",
741 sep == DOT ? '.' : '/'));
742 year = atol(sc_token);
746 /* flip months and days for european timing
757 if(mon < 0 || mon > 11 ) {
758 panic(e("did you really mean month %d?", mon+1));
760 if(mday < 1 || mday > 31) {
761 panic(e("I'm afraid that %d is not a valid day of the month",
764 try(assign_date(ptv, mday, mon, year));
771 /* Global functions */
775 * parsetime() is the external interface that takes tspec, parses
776 * it and puts the result in the time_value structure *ptv.
777 * It can return either absolute times (these are ensured to be
778 * correct) or relative time references that are expected to be
779 * added to some absolute time value and then normalized by
780 * mktime() The return value is either TIME_OK (aka NULL) or
781 * the pointer to the error message in the case of problems
784 parsetime(char *tspec, struct time_value *ptv)
786 time_t now = time(NULL);
788 /* this MUST be initialized to zero for midnight/noon/teatime */
790 Specials = VariousWords; /* initialize special words context */
792 try(init_scanner( 1, &tspec ));
794 /* establish the default time reference */
795 ptv->type = ABSOLUTE_TIME;
797 ptv->tm = *localtime(&now);
798 ptv->tm.tm_isdst = -1; /* mk time can figure this out for us ... */
804 break; /* jump to OFFSET-SPEC part */
807 ptv->type = RELATIVE_TO_START_TIME;
810 ptv->type = RELATIVE_TO_END_TIME;
821 int time_reference = sc_tokid;
823 if( sc_tokid == PLUS || sc_tokid == MINUS )
825 if( time_reference != NOW ) {
826 panic(e("'start' or 'end' MUST be followed by +|- offset"));
829 if( sc_tokid != EOF ) {
830 panic(e("if 'now' is followed by a token it must be +|- offset"));
835 /* Only absolute time specifications below */
838 if (sc_tokid != NUMBER) break;
839 /* fix month parsing */
840 case JAN: case FEB: case MAR: case APR: case MAY: case JUN:
841 case JUL: case AUG: case SEP: case OCT: case NOV: case DEC:
843 if (sc_tokid != NUMBER) break;
847 /* evil coding for TEATIME|NOON|MIDNIGHT - we've initialised
848 * hr to zero up above, then fall into this case in such a
849 * way so we add +12 +4 hours to it for teatime, +12 hours
850 * to it for noon, and nothing at all for midnight, then
851 * set our rettime to that hour before leaping into the
861 /* if (ptv->tm.tm_hour >= hr) {
864 } */ /* shifting does not makes sense here ... noon is noon */
865 ptv->tm.tm_hour = hr;
872 panic(e("unparsable time: %s%s",sc_token,sct));
874 } /* ugly case statement */
877 * the OFFSET-SPEC part
879 * (NOTE, the sc_tokid was prefetched for us by the previous code)
881 if( sc_tokid == PLUS || sc_tokid == MINUS ) {
882 Specials = TimeMultipliers; /* switch special words context */
883 while( sc_tokid == PLUS || sc_tokid == MINUS ||
884 sc_tokid == NUMBER ) {
885 if( sc_tokid == NUMBER ) {
886 try(plus_minus(ptv, PREVIOUS_OP ));
888 try(plus_minus(ptv, sc_tokid));
889 token(); /* We will get EOF eventually but that's OK, since
890 token() will return us as many EOFs as needed */
894 /* now we should be at EOF */
895 if( sc_tokid != EOF ) {
896 panic(e("unparsable trailing text: '...%s%s'", sc_token, sct));
899 ptv->tm.tm_isdst = -1; /* for mktime to guess DST status */
900 if( ptv->type == ABSOLUTE_TIME )
901 if( mktime( &ptv->tm ) == -1 ) { /* normalize & check */
902 /* can happen for "nonexistent" times, e.g. around 3am */
903 /* when winter -> summer time correction eats a hour */
904 panic(e("the specified time is incorrect (out of range?)"));
911 int proc_start_end (struct time_value *start_tv,
912 struct time_value *end_tv,
915 if (start_tv->type == RELATIVE_TO_END_TIME && /* same as the line above */
916 end_tv->type == RELATIVE_TO_START_TIME) {
917 rrd_set_error("the start and end times cannot be specified "
918 "relative to each other");
922 if (start_tv->type == RELATIVE_TO_START_TIME) {
923 rrd_set_error("the start time cannot be specified relative to itself");
927 if (end_tv->type == RELATIVE_TO_END_TIME) {
928 rrd_set_error("the end time cannot be specified relative to itself");
932 if( start_tv->type == RELATIVE_TO_END_TIME) {
934 *end = mktime(&(end_tv->tm)) + end_tv->offset;
935 tmtmp = *localtime(end); /* reinit end including offset */
936 tmtmp.tm_mday += start_tv->tm.tm_mday;
937 tmtmp.tm_mon += start_tv->tm.tm_mon;
938 tmtmp.tm_year += start_tv->tm.tm_year;
939 *start = mktime(&tmtmp) + start_tv->offset;
941 *start = mktime(&(start_tv->tm)) + start_tv->offset;
943 if (end_tv->type == RELATIVE_TO_START_TIME) {
945 *start = mktime(&(start_tv->tm)) + start_tv->offset;
946 tmtmp = *localtime(start);
947 tmtmp.tm_mday += end_tv->tm.tm_mday;
948 tmtmp.tm_mon += end_tv->tm.tm_mon;
949 tmtmp.tm_year += end_tv->tm.tm_year;
950 *end = mktime(&tmtmp) + end_tv->offset;
952 *end = mktime(&(end_tv->tm)) + end_tv->offset;
955 } /* proc_start_end */