Trailing-Edge
-
PDP-10 Archives
-
SRI_NIC_PERM_FS_1_19910112
-
kcc-5/lib/timepa.c
There are 4 other files named timepa.c in the archive. Click here to see a list.
/*
** TIME_PARSE parse date/time string into a TMX structure
**
** Copyright (c) 1980, 1987 by Ken Harrenstien, SRI International.
**
** This code is quasi-public; it may be used freely in like software.
** It is not to be sold, nor used in licensed software without
** permission of the author.
** For everyone's benefit, please report bugs and improvements!
** (Internet: <KLH@SRI-NIC.ARPA>)
**
** Usage:
** #include <timex.h>
** int time_parse(const char *str, struct tmx *tp, char **endptr);
**
** Time values are stored in specified TMX structure (unspecified values
** are all set to TMX_NULL). The tmx_err member will be set to NULL
** or will point to a constant error message string.
** If endptr is not NULL, it will be used to return an updated string ptr.
**
** Return value:
** < 0 Clash error or ambiguous keyword.
** Err msg, *endptr points to start of bad token.
**
** >= 0 OK up to a point. Value is # of valid tokens read.
** Note that 0 means nothing was set. There are several
** possibilities:
** OK up to EOF = no err msg, *endptr points to terminating null char.
** OK up to break char = err msg, *endptr points to (non-alpha) char.
** OK up to unknown keyword = err msg, *endptr points to start of token;
** note that isalpha(**endptr) will be true!
*/
#include <stdio.h>
#include <ctype.h>
#include <timex.h>
/* Internal functions */
static char *ptfixup();
static int ptstash();
static int pttoken(), ptpeektok(),
ptitoken(), ptmatchstr(), ptscntok(), ptbrkc();
#define MAXTOKEN 20 /* Max # chars in a token */
struct token {
char *tcp; /* pointer to string */
int tcnt; /* # chars in token */
char tbrk; /* "break" char */
char tbrkl; /* last break char */
char tflg; /* 0 = alpha, 1 = numeric */
long tlong; /* value if numeric */
int tnum; /* value cast to (int), for efficiency */
struct tmwent *ttmw; /* ptr to a tmwent, if alpha */
char tbuf[MAXTOKEN+1]; /* Lowercase copy of alpha token */
};
struct tmwent { /* Time-Word Entry */
char *went; /* word string */
int wval; /* word value */
int wtype; /* word type (and flags) */
};
/* Word types */
#define TW_MASK 07 /* Mask for type */
#define TW_NOISE 0 /* Noise word, ignore */
#define TW_MON 1
#define TW_WDAY 2
#define TW_ZON 3
#define TW_AMPM 4
#define TW_1200 5 /* Word is NOON or MIDNIGHT (sigh) */
/* Flags */
#define TWF_TIM 020 /* Word is a time value (absence implies date) */
#define TWF_DST 040 /* Word is a DST-type timezone */
static struct tmwent tmwords [] = {
{"january", 0, TW_MON},
{"february", 1, TW_MON},
{"march", 2, TW_MON},
{"april", 3, TW_MON},
{"may", 4, TW_MON},
{"june", 5, TW_MON},
{"july", 6, TW_MON},
{"august", 7, TW_MON},
{"september", 8, TW_MON},
{"october", 9, TW_MON},
{"november", 10, TW_MON},
{"december", 11, TW_MON},
{"sunday", 0, TW_WDAY},
{"monday", 1, TW_WDAY},
{"tuesday", 2, TW_WDAY},
{"wednesday", 3, TW_WDAY},
{"thursday", 4, TW_WDAY},
{"friday", 5, TW_WDAY},
{"saturday", 6, TW_WDAY},
{"ut", 0*60, TWF_TIM|TW_ZON}, /* Universal Time */
{"gmt", 0*60, TWF_TIM|TW_ZON}, /* Greenwich Mean Time */
{"gst", 0*60, TWF_TIM|TW_ZON},
{"gdt", 0*60, TWF_TIM+TWF_DST|TW_ZON}, /* ?? */
{"ast", 4*60, TWF_TIM|TW_ZON}, /* Atlantic */
{"est", 5*60, TWF_TIM|TW_ZON}, /* Eastern */
{"cst", 6*60, TWF_TIM|TW_ZON}, /* Central */
{"mst", 7*60, TWF_TIM|TW_ZON}, /* Mountain */
{"pst", 8*60, TWF_TIM|TW_ZON}, /* Pacific */
{"yst", 9*60, TWF_TIM|TW_ZON}, /* Yukon */
{"hst", 10*60, TWF_TIM|TW_ZON}, /* Hawaii */
{"bst", 11*60, TWF_TIM|TW_ZON}, /* Bering */
{"adt", 4*60, TWF_TIM+TWF_DST|TW_ZON}, /* Atlantic */
{"edt", 5*60, TWF_TIM+TWF_DST|TW_ZON}, /* Eastern */
{"cdt", 6*60, TWF_TIM+TWF_DST|TW_ZON}, /* Central */
{"mdt", 7*60, TWF_TIM+TWF_DST|TW_ZON}, /* Mountain */
{"pdt", 8*60, TWF_TIM+TWF_DST|TW_ZON}, /* Pacific */
{"ydt", 9*60, TWF_TIM+TWF_DST|TW_ZON}, /* Yukon */
{"hdt", 10*60, TWF_TIM+TWF_DST|TW_ZON}, /* Hawaii */
{"bdt", 11*60, TWF_TIM+TWF_DST|TW_ZON}, /* Bering */
{"standard", 1, TWF_TIM|TW_ZON}, /* Local Standard */
{"std", 1, TWF_TIM|TW_ZON}, /* " " */
{"daylight", 1, TWF_TIM+TWF_DST|TW_ZON}, /* " Daylight */
{"am", 1, TWF_TIM|TW_AMPM}, /* Special frobs */
{"pm", 2, TWF_TIM|TW_AMPM},
{"noon", 12,TWF_TIM|TW_1200},
{"midnight", 0, TWF_TIM|TW_1200},
{"at", 0, TW_NOISE}, /* Noise word */
#if 0 /* Military timezone designations; single letters */
{"z", 0*60, TWF_TIM|TW_ZON}, /* Z = UT */
{"a", 1*60, TWF_TIM|TW_ZON}, /* A = UT-0100 */
{"b", 2*60, TWF_TIM|TW_ZON}, /* etc */
{"c", 3*60, TWF_TIM|TW_ZON},
{"d", 4*60, TWF_TIM|TW_ZON},
{"e", 5*60, TWF_TIM|TW_ZON},
{"f", 6*60, TWF_TIM|TW_ZON},
{"g", 7*60, TWF_TIM|TW_ZON},
{"h", 8*60, TWF_TIM|TW_ZON},
{"i", 9*60, TWF_TIM|TW_ZON}, /* note: 'J' not used */
{"k", 10*60, TWF_TIM|TW_ZON},
{"l", 11*60, TWF_TIM|TW_ZON},
{"m", 12*60, TWF_TIM|TW_ZON},
{"n", -1*60, TWF_TIM|TW_ZON},
{"o", -2*60, TWF_TIM|TW_ZON},
{"p", -3*60, TWF_TIM|TW_ZON},
{"q", -4*60, TWF_TIM|TW_ZON},
{"r", -5*60, TWF_TIM|TW_ZON},
{"s", -6*60, TWF_TIM|TW_ZON},
{"t", -7*60, TWF_TIM|TW_ZON},
{"u", -8*60, TWF_TIM|TW_ZON},
{"v", -9*60, TWF_TIM|TW_ZON},
{"w", -10*60, TWF_TIM|TW_ZON},
{"x", -11*60, TWF_TIM|TW_ZON},
{"y", -12*60, TWF_TIM|TW_ZON},
#endif
{0, 0, 0}, /* Zero entry to terminate searches */
};
static struct tmz inittmz = { TMX_NULL, TMX_NULL, TMX_NULL };
#define TERR(str) { err = str; goto edone; }
int
time_parse(astr, tp, endptr)
char *astr, **endptr;
register struct tmx *tp;
{
register struct tmwent *twp;
register int i;
struct token tok, xtok;
char *cp;
int ch, ord, *ip;
int ntoks = 0;
char *err = 0;
int midnoon = TMX_NULL, ampm = TMX_NULL;
/* Initialize the TM structure. Can't do it cleverly because
** we don't know what the implementation's TM struct looks like. Barf.
*/
tp->tm.tm_sec = TMX_NULL;
tp->tm.tm_min = TMX_NULL;
tp->tm.tm_hour = TMX_NULL;
tp->tm.tm_mday = TMX_NULL;
tp->tm.tm_mon = TMX_NULL;
tp->tm.tm_year = TMX_NULL;
tp->tm.tm_wday = TMX_NULL;
tp->tm.tm_yday = TMX_NULL;
tp->tm.tm_isdst = TMX_NULL;
tp->tmz = inittmz; /* Can do TMZ part semi-cleverly anyway */
if (astr == 0) { /* Check for initialize-only invocation */
tp->tmx_err = 0;
if (endptr) *endptr = 0;
return 0;
}
tok.tcp = astr; /* Initialize token input */
tok.tcnt = tok.tbrkl = 0;
/* MAIN TOKEN PROCESSING LOOP */
while (pttoken(&tok)) { /* Get next token */
ntoks++;
if (tok.tflg == 0) { /* Alpha? */
/* Handle general alpha token */
twp = tok.ttmw; /* Get ptr to entry */
i = twp->wval; /* And value */
switch (twp->wtype&TW_MASK) {
case TW_NOISE: /* If noise word, */
break; /* just ignore it. */
case TW_MON:
if (ptstash(&tp->tm.tm_mon, i))
TERR("mon clash")
break;
case TW_WDAY:
if (ptstash(&tp->tm.tm_wday, i))
TERR("wday clash")
break;
case TW_ZON:
if (ptstash(&tp->tmz.tmz_minwest, i))
TERR("zone clash")
if (ptstash(&tp->tm.tm_isdst, (twp->wtype&TWF_DST)?1:0))
TERR("DST clash")
#if 0 /* Special check for "GMT+hh:mm" */
if (i == 0 &&
(tok.tbrk == '+' || tok.tbrk == '-')) {
}
#endif
break;
case TW_AMPM:
if (ptstash(&m, i))
TERR("am/pm clash")
break;
case TW_1200:
if (ptstash(&midnoon, i))
TERR("noon/midnight clash")
break;
default:
TERR("bad keyword")
}
continue; /* Process next token */
}
/* Token is number. LOTS of hairy heuristics. */
if (tok.tcnt > 6) /* 7 or more digits in string? (barf!) */
TERR("number too big")
if (tok.tcnt == 6) { /* 6 digits = HHMMSS. Needs special crock */
/* since 6 are too big for 16-bit integer! */
if (ptstash(&tp->tm.tm_hour, (int)(tok.tlong/10000)))
TERR("hour clash")
tok.tnum = tok.tlong % 10000;
goto coltm4;
}
i = tok.tnum; /* Value now known to be valid; get it. */
if (tok.tcnt == 5 /* 5 digits = HMMSS */
|| tok.tcnt == 3) { /* 3 digits = HMM */
if (tok.tcnt != 3) {
if (ptstash(&tp->tm.tm_sec, i%100))
TERR("sec clash")
i /= 100;
}
hhmm4: if (ptstash(&tp->tm.tm_min, i%100))
TERR("min clash")
i /= 100;
hh2: if (ptstash(&tp->tm.tm_hour, i))
TERR("hour clash")
continue; /* Continue normal loop */
}
if (tok.tcnt == 4) { /* 4 digits = YEAR or HHMM */
if (tp->tm.tm_year != TMX_NULL) goto hhmm4; /* Already got yr? */
if (tp->tm.tm_hour != TMX_NULL) goto year4; /* Already got hr? */
if ((i%100) > 59) goto year4; /* MM >= 60? */
if (tok.tbrk == ':') /* HHMM:SS ? */
if (ptstash(&tp->tm.tm_hour, i/100)
|| ptstash(&tp->tm.tm_min, i%100))
TERR("hr/min clash")
else goto coltm2; /* Go handle SS */
if ( tok.tbrk != ','
&& tok.tbrk != '/'
&& ptpeektok(&tok, &xtok) /* Peek into xtok */
&& xtok.tflg == 0 /* alpha */
&& (xtok.ttmw->wtype&TWF_TIM)) /* HHMM-ZON */
goto hhmm4;
if ( tok.tbrkl == '-' /* DD-Mon-YYYY */
|| tok.tbrkl == ',' /* Mon DD, YYYY */
|| tok.tbrkl == '/' /* MM/DD/YYYY */
|| tok.tbrkl == '.' /* DD.MM.YYYY */
|| tok.tbrk == '-' /* YYYY-MM-DD */
) goto year4;
goto hhmm4; /* Give up, assume HHMM. */
}
/* From this point on, assume tcnt == 1 or 2 */
/* 2 digits = YY, MM, DD, or HH (MM and SS caught at coltime) */
if (tok.tbrk == ':') /* HH:MM[:SS] */
goto coltime; /* must be part of time. */
if (i > 31) goto yy2; /* If >= 32, only YY poss. */
/* Check for numerical-format date. Very messy. */
for (cp = "/-."; ch = *cp++;) {
ord = (ch == '.' ? 0 : 1); /* n/m = D/M or M/D */
if (tok.tbrk == ch) { /* "NN-" */
if (tok.tbrkl != ch) {
if (ptpeektok(&tok, &xtok)
&& xtok.tflg == 0
&& xtok.ttmw->wtype == TW_MON)
goto dd2;
if(ord) goto mm2; else goto dd2; /* "NN-" */
} /* "-NN-" */
if ( tp->tm.tm_mday == TMX_NULL
&& tp->tm.tm_year != TMX_NULL) /* If "YY-NN-" */
goto mm2; /* then always MM */
if (ord) goto dd2;
goto mm2;
}
if (tok.tbrkl == ch /* "-NN" */
&& (ord ? tp->tm.tm_mon : tp->tm.tm_mday) != TMX_NULL) {
if ((ord ? tp->tm.tm_mday
: tp->tm.tm_mon) == TMX_NULL) /* MM/DD */
if (ord) goto dd2; else goto mm2;
goto yy2; /* "-YY" */
}
} /* End of messy for() loop */
/* At this point only YY, DD, and HH are left.
* YY is very unlikely since value is <= 32 and there was
* no numerical format date. Make one last try at YY
* before dropping through to DD vs HH code.
*/
if (tok.tcnt == 2 /* If 2 digits */
&& tp->tm.tm_hour != TMX_NULL /* and already have hour */
&& tp->tm.tm_mday != TMX_NULL /* and day, but */
&& tp->tm.tm_year == TMX_NULL) /* no year, then assume */
goto yy2; /* that's what we have. */
/* Now reduced to choice between HH and DD */
if (tp->tm.tm_hour != TMX_NULL) goto dd2; /* Have hour? Assume day. */
if (tp->tm.tm_mday != TMX_NULL) goto hh2; /* Have day? Assume hour. */
if (i > 24) goto dd2; /* Impossible HH means DD */
if (ptpeektok(&tok, &xtok) /* If next token exists */
&& xtok.tflg == 0 /* and is an alpha */
&& xtok.ttmw->wtype&TWF_TIM) /* time-spec, assume hour! */
goto hh2; /* e.g. "3 PM", "11-EDT" */
/* Otherwise assume day and drop thru */
dd2: if (ptstash(&tp->tm.tm_mday, i)) /* Store day (1 based) */
TERR("day clash")
continue; /* Continue normal loop */
mm2: if (ptstash(&tp->tm.tm_mon, i-1)) /* Store month (0-based) */
TERR("month clash")
continue; /* Continue normal loop */
yy2: i += 1900;
year4: if (ptstash(&tp->tm.tm_year, i)) /* Store year (full number) */
TERR("year clash")
continue; /* Continue normal loop */
/* Hack HH:MM[[:]SS] */
coltime:
if (ptstash(&tp->tm.tm_hour, i))
TERR("hour clash")
if (!pttoken(&tok)) /* Get next token */
break; /* EOF? Done, break out of loop! */
ntoks++;
if (!tok.tflg)
TERR("bad syntax - hh:?") /* HH:<alpha> ?? */
if (tok.tcnt == 4) { /* MMSS */
coltm4: if (ptstash(&tp->tm.tm_min, tok.tnum/100)
|| ptstash(&tp->tm.tm_sec, tok.tnum%100))
TERR("min/sec clash")
continue; /* Continue normal loop */
}
if (tok.tcnt != 2
|| ptstash(&tp->tm.tm_min, tok.tnum))
TERR("min clash or bad")
if (tok.tbrk != ':') /* Seconds follow? */
continue; /* Nope, back to normal loop */
coltm2:
if (!pttoken(&tok)) /* Get next token */
break; /* EOF, break out of loop! */
ntoks++;
if (!tok.tflg || tok.tcnt != 2) /* Verify SS */
TERR("secs expected")
if (ptstash(&tp->tm.tm_sec, tok.tnum))
TERR("sec clash")
/* Back to start of loop */
}
/* End of gigantic token-processing while() loop */
if (midnoon != TMX_NULL || ampm != TMX_NULL)
err = ptfixup(tp, midnoon, ampm); /* Do final fixups */
edone: /* TERR macro jumps here */
i = err ? -1 : ntoks; /* Determine return value */
if (i >= 0) { /* If no errs, check last token attempt */
if (tok.tcnt) { /* Was a token read? */
if (tok.tnum) /* Yes, was a match found? */
TERR("ambiguous keyword") /* Yes, ambiguity is error! */
else err = "unknown keyword"; /* No, just unknown, semi-OK */
} else if (tok.tbrk) /* No token read, was break EOF? */
err = "non-null break char"; /* No, semi-OK */
}
tp->tmx_err = err; /* Set error string (NULL if none) */
if (endptr)
*endptr = tok.tcp;
return i;
}
/* Auxiliary routines, all static */
/* PTSTASH - Store date/time value, checking for possible clash.
* Returns 0 if successful (for easier coding).
* Fails if entry already set to a different value.
*/
static int
ptstash(ip, val)
register int *ip;
{
if(*ip != TMX_NULL) /* If entry is already set, */
return *ip != val; /* fail if new value isn't same. */
*ip = val; /* Not set, so new val always OK. */
return 0;
}
/* PT12HACK - This subroutine is invoked for NOON or MIDNIGHT when wrapping up
* just prior to returning from time_parse.
* Returns NULL if everything's OK, else an error string pointer.
*/
static char *
ptfixup(tp, midnoon, ampm)
register struct tmx *tp;
int midnoon, ampm;
{
register int i, h;
if (midnoon != TMX_NULL) {
if (((i=tp->tm.tm_min) && i != TMX_NULL) /* Ensure mins, secs */
|| ((i=tp->tm.tm_sec) && i != TMX_NULL)) /* are 0 or unspec'd */
return "mm:ss must be 00:00";
i = midnoon; /* Get 0 or 12 (midnite or noon) */
if ((h = tp->tm.tm_hour) == TMX_NULL /* If hour unspec'd, win */
|| h == 12) /* or if 12:00 (matches either) */
tp->tm.tm_hour = i; /* Then set time */
else if (!(i == 0 /* Nope, but if midnight and */
&& (h == 0 || h == 24))) /* time matches, can pass. */
return "midnight hr clash";
/* midnoon spec won. Cannot co-exist with AM/PM... */
if (ampm != TMX_NULL)
return "am/pm clash";
} else if (ampm != TMX_NULL) {
/* Adjust hour properly for AM/PM spec */
switch (ampm) {
default: return "bad am/pm val";
case TMX_NULL: /* Ignore these values */
case 0:
break;
case 1: /* AM */
case 2: /* PM */
if ((i = tp->tm.tm_hour) == TMX_NULL)
i = 0;
if (i > 12) return "am/pm out of range";
if (i == 12) i = 0; /* Modulo 12 */
if (ampm == 2) /* If PM, then */
i += 12; /* get 24-hour time */
tp->tm.tm_hour = i;
break;
}
}
return 0; /* Won, no problems */
}
/* PTTOKEN - Given pointer to a token struct, reads the next token into the
* struct and identifies it to some degree.
* Returns 0 if failure or no more tokens, in which case:
* token.tcnt != 0 if bad keyword. token.tnum has # matches
* (either 0 for none, or 2 for ambiguous partial match)
* token.tcnt == 0 if EOF. token.tbrk will be 0 for normal EOF,
* else will hold the illegal break char.
*/
/* PTPEEKTOK - Same but takes input from 1st token struct and leaves
** peeked-at token in 2nd struct.
*/
static int
pttoken(tkp)
register struct token *tkp;
{ return ptitoken(tkp->tcp + tkp->tcnt, tkp);
}
static int
ptpeektok(otkp, ntkp)
struct token *otkp, *ntkp;
{ return ptitoken(otkp->tcp + otkp->tcnt, ntkp);
}
static int
ptitoken(str, tkp)
char *str;
register struct token *tkp;
{
/* Read token from string */
if (ptscntok(str, tkp) == 0) {
#ifdef DEBUG
printf("EOF\n");
#endif
return 0; /* No token found, tcnt and tnum both 0. */
}
/* Got a token, now try to identify it as number or keyword */
#ifdef DEBUG
{ char *cp = tkp->tcp;
i = cp[tkp->tcnt];
cp[tkp->tcnt] = 0;
printf("Token: \"%s\" ", cp);
cp[tkp->tcnt] = i;
}
#endif
if (tkp->tflg == 0) { /* Check num or alpha */
/* Alpha, look up in keyword table */
if ((tkp->tnum = ptmatchstr(tkp->tbuf, &tkp->ttmw)) != 1) {
#ifdef DEBUG
printf("Not found!\n");
#endif
return 0;
}
}
#ifdef DEBUG
if (tkp->tflg)
printf("Val: %d.\n",tkp->tnum);
else printf("Found: \"%s\", val: %d., type %o\n",
tkp->ttmw->went, tkp->ttmw->wval, tkp->ttmw->wtype);
#endif
return 1;
}
static int
ptmatchstr(str, atmw)
char *str;
struct tmwent **atmw; /* Holds last match found */
{
register char *cp, *mp;
register int c;
struct tmwent *tmw = &tmwords[0];
*atmw = 0; /* Initialize to null */
for ( ; mp = tmw->went; tmw++) {
cp = str;
while ((c = *cp++) == *mp++) /* Compare loop */
if (c == 0) {
*atmw = tmw; /* Exact match */
return 1;
}
if (c == 0) /* Partial match? */
if (*atmw) return 2; /* Ambiguous */
else *atmw = tmw; /* 1st ambig */
}
return (*atmw ? 1 : 0); /* Win if any non-ambiguous partial match */
}
/* PTSCNTOK - Read token from input string into token structure.
* Skips over all initial whitespace and break chars.
* Returns the # of chars in the token; this will be 0 if
* hit a EOF-type delimiter or end of string before any token read.
*/
static int
ptscntok(cp, tkp)
register char *cp;
register struct token *tkp;
{
register int c;
tkp->tbrkl = tkp->tbrk; /* Set "last break" */
tkp->tcnt = 0; /* Clear # chars in token */
tkp->tcp = cp; /* Set pointer to start of token */
/* First flush all whitespace and legal breaks */
while (c = *cp++) {
if (!isspace(c))
if (ptbrkc(c))
tkp->tbrkl = c; /* Set new last-break-char */
else break; /* Not space or break, stop loop */
tkp->tcp = cp; /* Update new start of token */
}
/* c has first non-space, non-break char. */
if (isdigit(c)) { /* Numerical token? */
register long i = 0; /* Initialize accumulated value */
do { tkp->tcnt++;
i = i*10 + (c-'0');
} while (isdigit(c = *cp++));
tkp->tnum = tkp->tlong = i; /* Set token value */
tkp->tflg = 1; /* say numeric */
} else if (isalpha(c)) { /* Alpha word token? */
register char *bp = &tkp->tbuf[0]; /* Initialize ptr into buff */
do {
if (++tkp->tcnt <= MAXTOKEN) *bp++ = tolower(c);
} while (isalpha(c = *cp++));
*bp = '\0';
tkp->tflg = 0; /* say alpha */
} else {
/* Char is illegal (most likely the terminating null).
** Nothing special needs to be done.
*/
}
if (tkp->tcnt)
while (isspace(c)) c = *cp++; /* If token read, flush wsp */
tkp->tbrk = c; /* Use 1st non-wsp as break char */
return tkp->tcnt; /* Then return # token chars seen! */
}
/* PTBRKC(ch) - returns true if char is a valid date/time break char.
*/
static int
ptbrkc(ch)
{
switch (ch) {
case '(': case ')':
case '-': case ',':
case '/': case ':':
case '.':
return 1;
default:
return 0;
}
}