921 lines
23 KiB
C
921 lines
23 KiB
C
/* $NetBSD: cal.c,v 1.23 2008/01/01 17:45:01 dholland Exp $ */
|
|
|
|
/*
|
|
* Copyright (c) 1989, 1993, 1994
|
|
* The Regents of the University of California. All rights reserved.
|
|
*
|
|
* This code is derived from software contributed to Berkeley by
|
|
* Kim Letkeman.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
* 3. Neither the name of the University nor the names of its contributors
|
|
* may be used to endorse or promote products derived from this software
|
|
* without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
|
|
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
* SUCH DAMAGE.
|
|
*/
|
|
|
|
#include <sys/cdefs.h>
|
|
#ifndef lint
|
|
__COPYRIGHT("@(#) Copyright (c) 1989, 1993, 1994\n\
|
|
The Regents of the University of California. All rights reserved.\n");
|
|
#endif /* not lint */
|
|
|
|
#ifndef lint
|
|
#if 0
|
|
static char sccsid[] = "@(#)cal.c 8.4 (Berkeley) 4/2/94";
|
|
#else
|
|
__RCSID("$NetBSD: cal.c,v 1.23 2008/01/01 17:45:01 dholland Exp $");
|
|
#endif
|
|
#endif /* not lint */
|
|
|
|
#include <sys/types.h>
|
|
|
|
#include <ctype.h>
|
|
#include <err.h>
|
|
#include <errno.h>
|
|
#include <limits.h>
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <termcap.h>
|
|
#include <time.h>
|
|
#include <tzfile.h>
|
|
#include <unistd.h>
|
|
|
|
#define SATURDAY 6 /* 1 Jan 1 was a Saturday */
|
|
|
|
#define FIRST_MISSING_DAY reform->first_missing_day
|
|
#define NUMBER_MISSING_DAYS reform->missing_days
|
|
|
|
#define MAXDAYS 42 /* max slots in a month array */
|
|
#define SPACE -1 /* used in day array */
|
|
|
|
static int days_in_month[2][13] = {
|
|
{0, 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31},
|
|
{0, 31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31},
|
|
};
|
|
|
|
int empty[MAXDAYS] = {
|
|
SPACE, SPACE, SPACE, SPACE, SPACE, SPACE, SPACE,
|
|
SPACE, SPACE, SPACE, SPACE, SPACE, SPACE, SPACE,
|
|
SPACE, SPACE, SPACE, SPACE, SPACE, SPACE, SPACE,
|
|
SPACE, SPACE, SPACE, SPACE, SPACE, SPACE, SPACE,
|
|
SPACE, SPACE, SPACE, SPACE, SPACE, SPACE, SPACE,
|
|
SPACE, SPACE, SPACE, SPACE, SPACE, SPACE, SPACE,
|
|
};
|
|
int shift_days[2][4][MAXDAYS + 1];
|
|
|
|
char *month_names[12] = {
|
|
"January", "February", "March", "April", "May", "June",
|
|
"July", "August", "September", "October", "November", "December",
|
|
};
|
|
|
|
char *day_headings = " S M Tu W Th F S";
|
|
char *j_day_headings = " S M Tu W Th F S";
|
|
|
|
/* leap years according to the julian calendar */
|
|
#define j_leap_year(y, m, d) \
|
|
(((m) > 2) && \
|
|
!((y) % 4))
|
|
|
|
/* leap years according to the gregorian calendar */
|
|
#define g_leap_year(y, m, d) \
|
|
(((m) > 2) && \
|
|
((!((y) % 4) && ((y) % 100)) || \
|
|
!((y) % 400)))
|
|
|
|
/* leap year -- account for gregorian reformation at some point */
|
|
#define leap_year(yr) \
|
|
((yr) <= reform->year ? j_leap_year((yr), 3, 1) : \
|
|
g_leap_year((yr), 3, 1))
|
|
|
|
/* number of julian leap days that have passed by a given date */
|
|
#define j_leap_days(y, m, d) \
|
|
((((y) - 1) / 4) + j_leap_year(y, m, d))
|
|
|
|
/* number of gregorian leap days that have passed by a given date */
|
|
#define g_leap_days(y, m, d) \
|
|
((((y) - 1) / 4) - (((y) - 1) / 100) + (((y) - 1) / 400) + \
|
|
g_leap_year(y, m, d))
|
|
|
|
/*
|
|
* Subtracting the gregorian leap day count (for a given date) from
|
|
* the julian leap day count (for the same date) describes the number
|
|
* of days from the date before the shift to the next date that
|
|
* appears in the calendar. Since we want to know the number of
|
|
* *missing* days, not the number of days that the shift spans, we
|
|
* subtract 2.
|
|
*
|
|
* Alternately...
|
|
*
|
|
* There's a reason they call the Dark ages the Dark Ages. Part of it
|
|
* is that we don't have that many records of that period of time.
|
|
* One of the reasons for this is that a lot of the Dark Ages never
|
|
* actually took place. At some point in the first millenium A.D., a
|
|
* ruler of some power decided that he wanted the number of the year
|
|
* to be different than what it was, so he changed it to coincide
|
|
* nicely with some event (a birthday or anniversary, perhaps a
|
|
* wedding, or maybe a centennial for a largish city). One of the
|
|
* side effects of this upon the Gregorian reform is that two Julian
|
|
* leap years (leap days celebrated during centennial years that are
|
|
* not quatro-centennial years) were skipped.
|
|
*/
|
|
#define GREGORIAN_MAGIC 2
|
|
|
|
/* number of centuries since the reform, not inclusive */
|
|
#define centuries_since_reform(yr) \
|
|
((yr) > reform->year ? ((yr) / 100) - (reform->year / 100) : 0)
|
|
|
|
/* number of centuries since the reform whose modulo of 400 is 0 */
|
|
#define quad_centuries_since_reform(yr) \
|
|
((yr) > reform->year ? ((yr) / 400) - (reform->year / 400) : 0)
|
|
|
|
/* number of leap years between year 1 and this year, not inclusive */
|
|
#define leap_years_since_year_1(yr) \
|
|
((yr) / 4 - centuries_since_reform(yr) + quad_centuries_since_reform(yr))
|
|
|
|
struct reform {
|
|
const char *country;
|
|
int ambiguity, year, month, date;
|
|
long first_missing_day;
|
|
int missing_days;
|
|
/*
|
|
* That's 2 for standard/julian display, 4 for months possibly
|
|
* affected by the Gregorian shift, and MAXDAYS + 1 for the
|
|
* days that get displayed, plus a crib slot.
|
|
*/
|
|
} *reform, reforms[] = {
|
|
{ "DEFAULT", 0, 1752, 9, 3 },
|
|
{ "Italy", 1, 1582, 10, 5 },
|
|
{ "Spain", 1, 1582, 10, 5 },
|
|
{ "Portugal", 1, 1582, 10, 5 },
|
|
{ "Poland", 1, 1582, 10, 5 },
|
|
{ "France", 2, 1582, 12, 10 },
|
|
{ "Luxembourg", 2, 1582, 12, 22 },
|
|
{ "Netherlands", 2, 1582, 12, 22 },
|
|
{ "Bavaria", 0, 1583, 10, 6 },
|
|
{ "Austria", 2, 1584, 1, 7 },
|
|
{ "Switzerland", 2, 1584, 1, 12 },
|
|
{ "Hungary", 0, 1587, 10, 22 },
|
|
{ "Germany", 0, 1700, 2, 19 },
|
|
{ "Norway", 0, 1700, 2, 19 },
|
|
{ "Denmark", 0, 1700, 2, 19 },
|
|
{ "Great Britain", 0, 1752, 9, 3 },
|
|
{ "England", 0, 1752, 9, 3 },
|
|
{ "America", 0, 1752, 9, 3 },
|
|
{ "Sweden", 0, 1753, 2, 18 },
|
|
{ "Finland", 0, 1753, 2, 18 },
|
|
{ "Japan", 0, 1872, 12, 20 },
|
|
{ "China", 0, 1911, 11, 7 },
|
|
{ "Bulgaria", 0, 1916, 4, 1 },
|
|
{ "U.S.S.R.", 0, 1918, 2, 1 },
|
|
{ "Serbia", 0, 1919, 1, 19 },
|
|
{ "Romania", 0, 1919, 1, 19 },
|
|
{ "Greece", 0, 1924, 3, 10 },
|
|
{ "Turkey", 0, 1925, 12, 19 },
|
|
{ "Egypt", 0, 1928, 9, 18 },
|
|
{ NULL, 0, 0, 0, 0 },
|
|
};
|
|
|
|
int julian;
|
|
int dow;
|
|
int hilite;
|
|
char *md, *me;
|
|
|
|
void init_hilite(void);
|
|
int getnum(const char *);
|
|
void gregorian_reform(const char *);
|
|
void reform_day_array(int, int, int *, int *, int *,int *,int *,int *);
|
|
int ascii_day(char *, int);
|
|
void center(char *, int, int);
|
|
void day_array(int, int, int *);
|
|
int day_in_week(int, int, int);
|
|
int day_in_year(int, int, int);
|
|
void monthrange(int, int, int, int, int);
|
|
int main(int, char **);
|
|
void trim_trailing_spaces(char *);
|
|
void usage(void);
|
|
|
|
int
|
|
main(int argc, char **argv)
|
|
{
|
|
struct tm *local_time;
|
|
time_t now;
|
|
int ch, yflag;
|
|
long month, year;
|
|
int before, after, use_reform;
|
|
int yearly = 0;
|
|
char *when, *eoi;
|
|
|
|
before = after = 0;
|
|
use_reform = yflag = year = 0;
|
|
when = NULL;
|
|
while ((ch = getopt(argc, argv, "A:B:d:hjR:ry3")) != -1) {
|
|
switch (ch) {
|
|
case 'A':
|
|
after = getnum(optarg);
|
|
if (after < 0)
|
|
errx(1, "Argument to -A must be positive");
|
|
break;
|
|
case 'B':
|
|
before = getnum(optarg);
|
|
if (before < 0)
|
|
errx(1, "Argument to -B must be positive");
|
|
break;
|
|
case 'd':
|
|
dow = getnum(optarg);
|
|
if (dow < 0 || dow > 6)
|
|
errx(1, "illegal day of week value: use 0-6");
|
|
break;
|
|
case 'h':
|
|
init_hilite();
|
|
break;
|
|
case 'j':
|
|
julian = 1;
|
|
break;
|
|
case 'R':
|
|
when = optarg;
|
|
break;
|
|
case 'r':
|
|
use_reform = 1;
|
|
break;
|
|
case 'y':
|
|
yflag = 1;
|
|
break;
|
|
case '3':
|
|
before = after = 1;
|
|
break;
|
|
case '?':
|
|
default:
|
|
usage();
|
|
/* NOTREACHED */
|
|
}
|
|
}
|
|
|
|
argc -= optind;
|
|
argv += optind;
|
|
|
|
if (when != NULL)
|
|
gregorian_reform(when);
|
|
if (reform == NULL)
|
|
gregorian_reform("DEFAULT");
|
|
|
|
month = 0;
|
|
switch (argc) {
|
|
case 2:
|
|
month = strtol(*argv++, &eoi, 10);
|
|
if (month < 1 || month > 12 || *eoi != '\0')
|
|
errx(1, "illegal month value: use 1-12");
|
|
year = strtol(*argv, &eoi, 10);
|
|
if (year < 1 || year > 9999 || *eoi != '\0')
|
|
errx(1, "illegal year value: use 1-9999");
|
|
break;
|
|
case 1:
|
|
year = strtol(*argv, &eoi, 10);
|
|
if (year < 1 || year > 9999 || (*eoi != '\0' && *eoi != '/' && *eoi != '-'))
|
|
errx(1, "illegal year value: use 1-9999");
|
|
if (*eoi != '\0') {
|
|
month = strtol(eoi + 1, &eoi, 10);
|
|
if (month < 1 || month > 12 || *eoi != '\0')
|
|
errx(1, "illegal month value: use 1-12");
|
|
}
|
|
break;
|
|
case 0:
|
|
(void)time(&now);
|
|
local_time = localtime(&now);
|
|
if (use_reform)
|
|
year = reform->year;
|
|
else
|
|
year = local_time->tm_year + TM_YEAR_BASE;
|
|
if (!yflag) {
|
|
if (use_reform)
|
|
month = reform->month;
|
|
else
|
|
month = local_time->tm_mon + 1;
|
|
}
|
|
break;
|
|
default:
|
|
usage();
|
|
}
|
|
|
|
if (!month) {
|
|
/* yearly */
|
|
month = 1;
|
|
before = 0;
|
|
after = 11;
|
|
yearly = 1;
|
|
}
|
|
|
|
monthrange(month, year, before, after, yearly);
|
|
|
|
exit(0);
|
|
}
|
|
|
|
#define DAY_LEN 3 /* 3 spaces per day */
|
|
#define J_DAY_LEN 4 /* 4 spaces per day */
|
|
#define WEEK_LEN 20 /* 7 * 3 - one space at the end */
|
|
#define J_WEEK_LEN 27 /* 7 * 4 - one space at the end */
|
|
#define HEAD_SEP 2 /* spaces between day headings */
|
|
#define J_HEAD_SEP 2
|
|
#define MONTH_PER_ROW 3 /* how many monthes in a row */
|
|
#define J_MONTH_PER_ROW 2
|
|
|
|
void
|
|
monthrange(int month, int year, int before, int after, int yearly)
|
|
{
|
|
int startmonth, startyear;
|
|
int endmonth, endyear;
|
|
int i, row;
|
|
int days[3][MAXDAYS];
|
|
char lineout[256];
|
|
int inayear;
|
|
int newyear;
|
|
int day_len, week_len, head_sep;
|
|
int month_per_row;
|
|
int skip, r_off, w_off;
|
|
|
|
if (julian) {
|
|
day_len = J_DAY_LEN;
|
|
week_len = J_WEEK_LEN;
|
|
head_sep = J_HEAD_SEP;
|
|
month_per_row = J_MONTH_PER_ROW;
|
|
}
|
|
else {
|
|
day_len = DAY_LEN;
|
|
week_len = WEEK_LEN;
|
|
head_sep = HEAD_SEP;
|
|
month_per_row = MONTH_PER_ROW;
|
|
}
|
|
|
|
month--;
|
|
|
|
startyear = year - (before + 12 - 1 - month) / 12;
|
|
startmonth = 12 - 1 - ((before + 12 - 1 - month) % 12);
|
|
endyear = year + (month + after) / 12;
|
|
endmonth = (month + after) % 12;
|
|
|
|
if (startyear < 0 || endyear > 9999) {
|
|
errx(1, "year should be in 1-9999\n");
|
|
}
|
|
|
|
year = startyear;
|
|
month = startmonth;
|
|
inayear = newyear = (year != endyear || yearly);
|
|
if (inayear) {
|
|
skip = month % month_per_row;
|
|
month -= skip;
|
|
}
|
|
else {
|
|
skip = 0;
|
|
}
|
|
|
|
do {
|
|
if (newyear) {
|
|
(void)snprintf(lineout, sizeof(lineout), "%d", year);
|
|
center(lineout, week_len * month_per_row +
|
|
head_sep * (month_per_row - 1), 0);
|
|
(void)printf("\n\n");
|
|
newyear = 0;
|
|
}
|
|
|
|
for (i = 0; i < skip; i++)
|
|
center("", week_len, head_sep);
|
|
|
|
for (; i < month_per_row; i++) {
|
|
int sep;
|
|
|
|
if (year == endyear && month + i > endmonth)
|
|
break;
|
|
|
|
sep = (i == month_per_row - 1) ? 0 : head_sep;
|
|
day_array(month + i + 1, year, days[i]);
|
|
if (inayear) {
|
|
center(month_names[month + i], week_len, sep);
|
|
}
|
|
else {
|
|
snprintf(lineout, sizeof(lineout), "%s %d",
|
|
month_names[month + i], year);
|
|
center(lineout, week_len, sep);
|
|
}
|
|
}
|
|
printf("\n");
|
|
|
|
for (i = 0; i < skip; i++)
|
|
center("", week_len, head_sep);
|
|
|
|
for (; i < month_per_row; i++) {
|
|
int sep;
|
|
|
|
if (year == endyear && month + i > endmonth)
|
|
break;
|
|
|
|
sep = (i == month_per_row - 1) ? 0 : head_sep;
|
|
if (dow) {
|
|
printf("%s ", (julian) ?
|
|
j_day_headings + 4 * dow :
|
|
day_headings + 3 * dow);
|
|
printf("%.*s", dow * (julian ? 4 : 3) - 1,
|
|
(julian) ? j_day_headings : day_headings);
|
|
} else
|
|
printf("%s", (julian) ? j_day_headings : day_headings);
|
|
printf("%*s", sep, "");
|
|
}
|
|
printf("\n");
|
|
|
|
for (row = 0; row < 6; row++) {
|
|
char *p = NULL;
|
|
|
|
memset(lineout, ' ', sizeof(lineout));
|
|
for (i = 0; i < skip; i++) {
|
|
/* nothing */
|
|
}
|
|
w_off = 0;
|
|
for (; i < month_per_row; i++) {
|
|
int col, *dp;
|
|
|
|
if (year == endyear && month + i > endmonth)
|
|
break;
|
|
|
|
p = lineout + i * (week_len + 2) + w_off;
|
|
dp = &days[i][row * 7];
|
|
for (col = 0; col < 7;
|
|
col++, p += day_len + r_off) {
|
|
r_off = ascii_day(p, *dp++);
|
|
w_off += r_off;
|
|
}
|
|
}
|
|
*p = '\0';
|
|
trim_trailing_spaces(lineout);
|
|
(void)printf("%s\n", lineout);
|
|
}
|
|
|
|
skip = 0;
|
|
month += month_per_row;
|
|
if (month >= 12) {
|
|
month -= 12;
|
|
year++;
|
|
newyear = 1;
|
|
}
|
|
} while (year < endyear || (year == endyear && month <= endmonth));
|
|
}
|
|
|
|
/*
|
|
* day_array --
|
|
* Fill in an array of 42 integers with a calendar. Assume for a moment
|
|
* that you took the (maximum) 6 rows in a calendar and stretched them
|
|
* out end to end. You would have 42 numbers or spaces. This routine
|
|
* builds that array for any month from Jan. 1 through Dec. 9999.
|
|
*/
|
|
void
|
|
day_array(int month, int year, int *days)
|
|
{
|
|
int day, dw, dm;
|
|
time_t t;
|
|
struct tm *tm;
|
|
|
|
t = time(NULL);
|
|
tm = localtime(&t);
|
|
tm->tm_year += TM_YEAR_BASE;
|
|
tm->tm_mon++;
|
|
tm->tm_yday++; /* jan 1 is 1 for us, not 0 */
|
|
|
|
for (dm = month + year * 12, dw = 0; dw < 4; dw++) {
|
|
if (dm == shift_days[julian][dw][MAXDAYS]) {
|
|
memmove(days, shift_days[julian][dw],
|
|
MAXDAYS * sizeof(int));
|
|
return;
|
|
}
|
|
}
|
|
|
|
memmove(days, empty, MAXDAYS * sizeof(int));
|
|
dm = days_in_month[leap_year(year)][month];
|
|
dw = day_in_week(1, month, year);
|
|
day = julian ? day_in_year(1, month, year) : 1;
|
|
while (dm--) {
|
|
if (hilite && year == tm->tm_year &&
|
|
(julian ? (day == tm->tm_yday) :
|
|
(month == tm->tm_mon && day == tm->tm_mday)))
|
|
days[dw++] = SPACE - day++;
|
|
else
|
|
days[dw++] = day++;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* day_in_year --
|
|
* return the 1 based day number within the year
|
|
*/
|
|
int
|
|
day_in_year(int day, int month, int year)
|
|
{
|
|
int i, leap;
|
|
|
|
leap = leap_year(year);
|
|
for (i = 1; i < month; i++)
|
|
day += days_in_month[leap][i];
|
|
return (day);
|
|
}
|
|
|
|
/*
|
|
* day_in_week
|
|
* return the 0 based day number for any date from 1 Jan. 1 to
|
|
* 31 Dec. 9999. Returns the day of the week of the first
|
|
* missing day for any given Gregorian shift.
|
|
*/
|
|
int
|
|
day_in_week(int day, int month, int year)
|
|
{
|
|
long temp;
|
|
|
|
temp = (long)(year - 1) * 365 + leap_years_since_year_1(year - 1)
|
|
+ day_in_year(day, month, year);
|
|
if (temp < FIRST_MISSING_DAY)
|
|
return ((temp - dow + 6 + SATURDAY) % 7);
|
|
if (temp >= (FIRST_MISSING_DAY + NUMBER_MISSING_DAYS))
|
|
return (((temp - dow + 6 + SATURDAY) - NUMBER_MISSING_DAYS) % 7);
|
|
return ((FIRST_MISSING_DAY - dow + 6 + SATURDAY) % 7);
|
|
}
|
|
|
|
int
|
|
ascii_day(char *p, int day)
|
|
{
|
|
int display, val, rc;
|
|
char *b;
|
|
static char *aday[] = {
|
|
"",
|
|
" 1", " 2", " 3", " 4", " 5", " 6", " 7",
|
|
" 8", " 9", "10", "11", "12", "13", "14",
|
|
"15", "16", "17", "18", "19", "20", "21",
|
|
"22", "23", "24", "25", "26", "27", "28",
|
|
"29", "30", "31",
|
|
};
|
|
|
|
if (day == SPACE) {
|
|
memset(p, ' ', julian ? J_DAY_LEN : DAY_LEN);
|
|
return (0);
|
|
}
|
|
if (day < SPACE) {
|
|
b = p;
|
|
day = SPACE - day;
|
|
} else
|
|
b = NULL;
|
|
if (julian) {
|
|
if ((val = day / 100) != 0) {
|
|
day %= 100;
|
|
*p++ = val + '0';
|
|
display = 1;
|
|
} else {
|
|
*p++ = ' ';
|
|
display = 0;
|
|
}
|
|
val = day / 10;
|
|
if (val || display)
|
|
*p++ = val + '0';
|
|
else
|
|
*p++ = ' ';
|
|
*p++ = day % 10 + '0';
|
|
} else {
|
|
*p++ = aday[day][0];
|
|
*p++ = aday[day][1];
|
|
}
|
|
|
|
rc = 0;
|
|
if (b != NULL) {
|
|
char *t, h[64];
|
|
int l;
|
|
|
|
l = p - b;
|
|
memcpy(h, b, l);
|
|
p = b;
|
|
|
|
if (md != NULL) {
|
|
for (t = md; *t; rc++)
|
|
*p++ = *t++;
|
|
memcpy(p, h, l);
|
|
p += l;
|
|
for (t = me; *t; rc++)
|
|
*p++ = *t++;
|
|
} else {
|
|
for (t = &h[0]; l--; t++) {
|
|
*p++ = *t;
|
|
rc++;
|
|
*p++ = '\b';
|
|
rc++;
|
|
*p++ = *t;
|
|
}
|
|
}
|
|
}
|
|
|
|
*p = ' ';
|
|
return (rc);
|
|
}
|
|
|
|
void
|
|
trim_trailing_spaces(char *s)
|
|
{
|
|
char *p;
|
|
|
|
for (p = s; *p; ++p)
|
|
continue;
|
|
while (p > s && isspace((unsigned char)*--p))
|
|
continue;
|
|
if (p > s)
|
|
++p;
|
|
*p = '\0';
|
|
}
|
|
|
|
void
|
|
center(char *str, int len, int separate)
|
|
{
|
|
|
|
len -= strlen(str);
|
|
(void)printf("%*s%s%*s", len / 2, "", str, len / 2 + len % 2, "");
|
|
if (separate)
|
|
(void)printf("%*s", separate, "");
|
|
}
|
|
|
|
/*
|
|
* gregorian_reform --
|
|
* Given a description of date on which the Gregorian Reform was
|
|
* applied. The argument can be any of the "country" names
|
|
* listed in the reforms array (case insensitive) or a date of
|
|
* the form YYYY/MM/DD. The date and month can be omitted if
|
|
* doing so would not select more than one different built-in
|
|
* reform point.
|
|
*/
|
|
void
|
|
gregorian_reform(const char *p)
|
|
{
|
|
int year, month, date;
|
|
int i, days, diw, diy;
|
|
char c;
|
|
|
|
i = sscanf(p, "%d%*[/,-]%d%*[/,-]%d%c", &year, &month, &date, &c);
|
|
switch (i) {
|
|
case 4:
|
|
/*
|
|
* If the character was sscanf()ed, then there's more
|
|
* stuff than we need.
|
|
*/
|
|
errx(1, "date specifier %s invalid", p);
|
|
case 0:
|
|
/*
|
|
* Not a form we can sscanf(), so void these, and we
|
|
* can try matching "country" names later.
|
|
*/
|
|
year = month = date = -1;
|
|
break;
|
|
case 1:
|
|
month = 0;
|
|
/*FALLTHROUGH*/
|
|
case 2:
|
|
date = 0;
|
|
/*FALLTHROUGH*/
|
|
case 3:
|
|
/*
|
|
* At last, some sanity checking on the values we were
|
|
* given.
|
|
*/
|
|
if (year < 1 || year > 9999)
|
|
errx(1, "%d: illegal year value: use 1-9999", year);
|
|
if (i > 1 && (month < 1 || month > 12))
|
|
errx(1, "%d: illegal month value: use 1-12", month);
|
|
if ((i == 3 && date < 1) || date < 0 ||
|
|
date > days_in_month[1][month])
|
|
/*
|
|
* What about someone specifying a leap day in
|
|
* a non-leap year? Well...that's a tricky
|
|
* one. We can't yet *say* whether the year
|
|
* in question is a leap year. What if the
|
|
* date given was, for example, 1700/2/29? is
|
|
* that a valid leap day?
|
|
*
|
|
* So...we punt, and hope that saying 29 in
|
|
* the case of February isn't too bad an idea.
|
|
*/
|
|
errx(1, "%d: illegal date value: use 1-%d", date,
|
|
days_in_month[1][month]);
|
|
break;
|
|
}
|
|
|
|
/*
|
|
* A complete date was specified, so use the other pope.
|
|
*/
|
|
if (date > 0) {
|
|
static struct reform Goestheveezl;
|
|
|
|
reform = &Goestheveezl;
|
|
reform->country = "Bompzidaize";
|
|
reform->year = year;
|
|
reform->month = month;
|
|
reform->date = date;
|
|
}
|
|
|
|
/*
|
|
* No date information was specified, so let's try to match on
|
|
* country name.
|
|
*/
|
|
else if (year == -1) {
|
|
for (reform = &reforms[0]; reform->year; reform++) {
|
|
if (strcasecmp(p, reform->country) == 0)
|
|
break;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* We have *some* date information, but not a complete date.
|
|
* Let's see if we have enough to pick a single entry from the
|
|
* list that's not ambiguous.
|
|
*/
|
|
else {
|
|
for (reform = &reforms[0]; reform->year; reform++) {
|
|
if ((year == 0 || year == reform->year) &&
|
|
(month == 0 || month == reform->month) &&
|
|
(date == 0 || month == reform->date))
|
|
break;
|
|
}
|
|
|
|
if (i <= reform->ambiguity)
|
|
errx(1, "%s: ambiguous short reform date specification", p);
|
|
}
|
|
|
|
/*
|
|
* Oops...we reached the end of the list.
|
|
*/
|
|
if (reform->year == 0)
|
|
errx(1, "reform name %s invalid", p);
|
|
|
|
/*
|
|
*
|
|
*/
|
|
reform->missing_days =
|
|
j_leap_days(reform->year, reform->month, reform->date) -
|
|
g_leap_days(reform->year, reform->month, reform->date) -
|
|
GREGORIAN_MAGIC;
|
|
|
|
reform->first_missing_day =
|
|
(reform->year - 1) * 365 +
|
|
day_in_year(reform->date, reform->month, reform->year) +
|
|
date +
|
|
j_leap_days(reform->year, reform->month, reform->date);
|
|
|
|
/*
|
|
* Once we know the day of the week of the first missing day,
|
|
* skip back to the first of the month's day of the week.
|
|
*/
|
|
diw = day_in_week(reform->date, reform->month, reform->year);
|
|
diw = (diw + 8 - (reform->date % 7)) % 7;
|
|
diy = day_in_year(1, reform->month, reform->year);
|
|
|
|
/*
|
|
* We might need all four of these (if you switch from Julian
|
|
* to Gregorian at some point after 9900, you get a gap of 73
|
|
* days, and that can affect four months), and it doesn't hurt
|
|
* all that much to precompute them, so there.
|
|
*/
|
|
date = 1;
|
|
days = 0;
|
|
for (i = 0; i < 4; i++)
|
|
reform_day_array(reform->month + i, reform->year,
|
|
&days, &date, &diw, &diy,
|
|
shift_days[0][i],
|
|
shift_days[1][i]);
|
|
}
|
|
|
|
/*
|
|
* reform_day_array --
|
|
* Pre-calculates the given month's calendar (in both "standard"
|
|
* and "julian day" representations) with respect for days
|
|
* skipped during a reform period.
|
|
*/
|
|
void
|
|
reform_day_array(int month, int year, int *done, int *date, int *diw, int *diy,
|
|
int *scal, int *jcal)
|
|
{
|
|
int mdays;
|
|
|
|
/*
|
|
* If the reform was in the month of october or later, then
|
|
* the month number from the caller could "overflow".
|
|
*/
|
|
if (month > 12) {
|
|
month -= 12;
|
|
year++;
|
|
}
|
|
|
|
/*
|
|
* Erase months, and set crib number. The crib number is used
|
|
* later to determine if the month to be displayed is here or
|
|
* should be built on the fly with the generic routine
|
|
*/
|
|
memmove(scal, empty, MAXDAYS * sizeof(int));
|
|
scal[MAXDAYS] = month + year * 12;
|
|
memmove(jcal, empty, MAXDAYS * sizeof(int));
|
|
jcal[MAXDAYS] = month + year * 12;
|
|
|
|
/*
|
|
* It doesn't matter what the actual month is when figuring
|
|
* out if this is a leap year or not, just so long as February
|
|
* gets the right number of days in it.
|
|
*/
|
|
mdays = days_in_month[g_leap_year(year, 3, 1)][month];
|
|
|
|
/*
|
|
* Bounce back to the first "row" in the day array, and fill
|
|
* in any days that actually occur.
|
|
*/
|
|
for (*diw %= 7; (*date - *done) <= mdays; (*date)++, (*diy)++) {
|
|
/*
|
|
* "date" doesn't get reset by the caller across calls
|
|
* to this routine, so we can actually tell that we're
|
|
* looking at April the 41st. Much easier than trying
|
|
* to calculate the absolute julian day for a given
|
|
* date and then checking that.
|
|
*/
|
|
if (*date < reform->date ||
|
|
*date >= reform->date + reform->missing_days) {
|
|
scal[*diw] = *date - *done;
|
|
jcal[*diw] = *diy;
|
|
(*diw)++;
|
|
}
|
|
}
|
|
*done += mdays;
|
|
}
|
|
|
|
int
|
|
getnum(const char *p)
|
|
{
|
|
long result;
|
|
char *ep;
|
|
|
|
errno = 0;
|
|
result = strtoul(p, &ep, 10);
|
|
if (p[0] == '\0' || *ep != '\0')
|
|
goto error;
|
|
if (errno == ERANGE && result == ULONG_MAX)
|
|
goto error;
|
|
if (result > INT_MAX)
|
|
goto error;
|
|
|
|
return (int)result;
|
|
|
|
error:
|
|
errx(1, "bad number: %s", p);
|
|
/*NOTREACHED*/
|
|
}
|
|
|
|
void
|
|
init_hilite(void)
|
|
{
|
|
static char control[128];
|
|
char cap[1024];
|
|
char *tc;
|
|
|
|
hilite++;
|
|
|
|
if (!isatty(fileno(stdout)))
|
|
return;
|
|
|
|
tc = getenv("TERM");
|
|
if (tc == NULL)
|
|
tc = "dumb";
|
|
if (tgetent(&cap[0], tc) != 1)
|
|
return;
|
|
|
|
tc = &control[0];
|
|
if ((md = tgetstr(hilite > 1 ? "mr" : "md", &tc)))
|
|
*tc++ = '\0';
|
|
if ((me = tgetstr("me", &tc)))
|
|
*tc++ = '\0';
|
|
if (me == NULL || md == NULL)
|
|
md = me = NULL;
|
|
}
|
|
|
|
void
|
|
usage(void)
|
|
{
|
|
|
|
(void)fprintf(stderr,
|
|
"usage: cal [-3hjry] [-A after] [-B before] [-d day-of-week] "
|
|
"[-R reform-spec]\n [[month] year]\n");
|
|
exit(1);
|
|
}
|