Calculate no. of days

[Kitti Thanapuasuwan]

Hi,

Anyone has program to calculate no. of days between two date. I want to use it in DCL program.

Thanks

[Jan van den Ende · Accepted solution]

Kitti,

upgrade to V7.3-2, and you will have F$DELTA_TIME


If you can not, a little DCL can do (we had it for years):

$ diff = 0
$ ldat=f$cvtim("","comparison","date")
$loop:
$ if f$cvtim("first_date+''diff'-","comparison","date") .lts. ldat
$ then
$ diff=diff+1
$ goto loop
$ endif

Sorry, indentation gets lost by forum makeup :-(

Proost.

Have one on me.

jpe

[Hein van den Heuvel]

Similar idea, but starting closer to the target:

$! DELTA_DAYS.COM original by J. FINKA Nov 1990
$! assume p1 <= p2 and delta < 9999
$ p2 = f$cvtime(p2,"absolute","date") !normalize
$ years = 'f$cvtime(p2,,"year") - 'f$cvtime(p1,,"year")
$ months = 'f$cvtime(p2,,"month") - 'f$cvtime(p1,,"month")
$ days = 'f$cvtime(p2,,"day") - 'f$cvtime(p1,,"day")
$ y = (36525 * years + 3044 * months) / 100 + days
$ i = 0
$ l: x = y + i
$ i = - i
$ if i .ge. 0 then i = i + 1
$ q = f$cvtime(p1 + " +" + f$string(x) + "-","absolute","date")
$ if q .nes. p2 then goto l
$ write sys$output "Delta days : ", x


$ @delta_days 01-jan-2001 01-jan-2005
Delta days : 1461

And one of many possible perl implemenations.
Calculate each date in seconds, subtract and divide by seconds in a day:

$ type delta_days.p
use Time::Local;
$_ = shift @ARGV;
if (/(\d{4})-(\d+)-(\d+)/) {
$d1 = timelocal(0,0,0,$3,$2-1,$1-1900);
} else {die "please provide date 1 as YYYY-MM-DD"};
$_ = shift @ARGV;
if (/(\d{4})-(\d+)-(\d+)/) {
$d2 = timelocal(0,0,0,$3,$2-1,$1-1900);
} else {die "please provide date 2 as YYYY-MM-DD"};
$days = ($d1 - $d2) / (24*60*60);
print "$days days";

$ perl delta_days.p 2005-1-1 2001-1-1
1461 days


Hein.

[Jan van den Ende]

@Hein:

Yes,

our _operational_ routine also had 1000 day steps with backtrack, then 100's, then 10's. but I did I Q & D re-create from memory...

And like any example given here:
A - use at own risk
B - use as starting point, take the idea & refine it.

cu in Nashua, I guess?

Proost.

Have one on me.

jpe

[Antoniov.]

Hi Kitti,
here there is a dcl proceure I converted from a my C program.
This procedure takes any date in dcl format (dd-mmm-yyyy) and converts it in julien date. Julien date is the number of days since a specific date. Within julien date you can add or subtract two dates.
Original routine works since 01-01-0000 but this dcl procedure works only since 1924 (read header for furthermore information).

Hope this help

Antonio Vigliotti

[Robert Atkinson]

Kitti - all of these are good solutions.

If you want a program to do this, ICALC (http://nucwww.chem.sunysb.edu/htbin/software_list.cgi?package=icalc) converts between Julian dates, so could be used to compare the two.

Robert.

[Lachnitt_1]

Hi Kitti,

starting with VMS 8.2 you can use the lexical Function F$DELTA:

$ write sys$output F$DELTA("01-JAN-2001","01-JAN-2005")
1461 00:00:00.00

Bye Kuddel

[Jan van den Ende]

Kuddel,

like I already wrote in my first reply, F$DELTA is there since V7.3-2 (to be precise, in the DCL patch)

Proost.

Have one on me.

jpe

[Wim Van den Wyngaert]

Small remark for Antonio's solution :
1) the 2 dates must be in the same year to allow + - operations.
2) leap years are not supported (29 days in feb)

Wim

[Wim Van den Wyngaert]

To be clear : 2) isn't working because the leap year is added for all dates in a year, not only for days after 28-feb.

Wim

[Antoniov.]

Wim,
to be clear, dcl routine is convertion of C program. Original program is more complex because manage all dates since 01-01-0000 until today, calculating leap year and missed day in 1582 (Italy, France, Spain). C program is certified by a complex test unit. I removed complex calculation for missed days of 1582 and other minor feature so dcl routine is simple to write and easy to execute. I didn't remove leap year calculation therefore conversion works fine since 01-01-1925 for every country.

Antonio Vigliotti

[Wim Van den Wyngaert]

Antonio,

Just test it with the diff between 28-feb and 1-mar. It's always going to return 1, even for leap years.

Wim

[Joseph Huber_1]

Just for fun (and frustrated at VMS 7.3-1),
I made an executable f$delta_time for pre-7.3-2.

Look at it in f$delta_time.html

[Wim Van den Wyngaert]

Make that
http://wwwvms.mppmu.mpg.de/~huber/util/main/f$delta_time.html

Wim

[Joseph Huber_1]

Wim,

since it's different in every other forum:
to put such a link here in ITRC forum, is it enough to put the URL, not the

[Wim Van den Wyngaert]

Yes Joseph. Just paste what you have cut out of the address bar in IE.

Wim

[Hein van den Heuvel]

Antonio,

I'm sorry to also report that your DCL example is truly broken. You may want to verify that the C program was correct to begin with.
- As reported, it does not really do leapyears.
- It also 'lucks out' that 2000 actually was a leap year unlike 1900 and 2100
- It fails to accept input like 01-feb 01-apr because 'apr' is less than 'feb' in text form.
- It fails to accept 28-feb-2005 01-mar-2005 because 01 sorts befor 28 in text form.
- The days-so-far-in-this-year calculation uses a clumsy loop while you can just lookup the startdays of each month in the table instead of using the number of days in the table.

btw... I used a perl one-liner to make that new lis from your list:
$ perl -e "foreach (qw(0 31 28 31 30 31 30 31 31 30 31 30 31)) { $x +=$_; print "",$x"" }"
,0,31,59,90,120,151,181,212,243,273,304,334,365

The update code below seems to work.
Regards,
Hein.


$! Convertion date from DCL format to julien format
$! Author: Antonio Vigliotti antoniov@shs-av.it
$! Tweaks by Hein van den Heuvel
$! Julien date is the number of days since 1-1-0
$! Converted from a C routine
$! Original routine calculate change of calendar by pope Gregor XIII in
$! year 1582. Therefor this DCL routine works only since 1582 (in Italy)
$! or since 1824 (in Greek). For other countries read history!
$! Julien date can be use to compare two dates.
$begdat=p1
$enddat=p2
$if p2.nes."" then goto mloop1
$mloop:
$ inquire begdat "Beg.Date [dd-mmm-yyyy]"
$ if begdat.eqs."" then exit
$ inquire enddat "End Date [dd-mmm-yyyy]"
$ if enddat.eqs."" then exit
$mloop1:
$ if f$cvt(enddat).lts.f$cvt(begdat) then goto mloop
$ dt=begdat
$ gosub c2julien
$ begj=j
$ dt=enddat
$ gosub c2julien
$ endj=j
$ write sys$output enddat," - ",begdat," = ",endj-begj
$if p2.nes."" then exit
$ goto mloop
$c2julien:
$ y=f$cvtime(dt,,"year")
$ m=f$cvtime(dt,,"month")
$ days = "0,0,31,59,90,120,151,181,212,243,273,304,334"
$! Good untill 2100. Need 100/400 year rule after that
$ j=y*365 + (y+3)/4 + 'f$elem(m,",",days) + 'f$cvtim(dt,,"day")
$if m.gt.2 then j = j + 3 - f$cvtim("28-FEB-''y' +1-",,"MONTH") ! Leap?
$ return

--------------------


$ @tmp 28-feb-2004 01-mar-2004
01-MAR-2004 - 28-FEB-2004 = 2
$ @tmp 28-feb-2005 01-mar-2005
01-MAR-2005 - 28-FEB-2005 = 1

[Antoniov.]

Sorry for mistakes.
I worked too hurry :-(
Now, attachment is right routine. It works with leap year since 1582 in Italy, Spain and France. Since 1711 in UK, 1918 in Russia, 1925 in Greek.
I checked accurately original c routine so DCL is more complex.
I converted part of C test unit. Now you see test since 01-01-1900 until 31-12-2005.
Julien calculate yeap year every 4 years but not every century but every 400 years.
Before 15-10-1582 (in Italy, ...) yeap year was every 4 years. Pope Gregorio XIII changed calendar missed 10 days. England corrected calendar on 1710, Russia after revolution of 1917.
For furthermore information ask me.

Antonio Vigliotti

[Antoniov.]

Just for curiosity you can see original C routine. It is a complex project to calculate various date items (easter day lien date, holyday days and some other items).
Original C routine is not translated in english language so you can understand by C source code.
All routine are certificated by a very complex test unit (not attacched).
You can ask me for every information.

Thanks to Wim and Hein for their warning.

Antonio Vigliotti