DateTime compare

[kmp84]

Hello All,

Can you point me some easy way to compare two datetime value?
I have date/time from RTC(DS1307) and data/time from RS232 device.
I want to know if date/time from device are equal or greater 30 min from DS1307 RTC.


Best Regards!

[temtronic]

One way....
Kinda rough but you get to work out the code... !

1st create 8 bit( byte) variables....
rtc_yr
rtc_mt
rtc_dy
rtc_hr
rtc_mn

ser_yr
ser_mt
ser_dy
ser_hr
ser_mn

now 2, 16 bit vaiables
rtc_time_sum
ser_time_sum


rtc_time_sum = 60*rtc_hr+ rtc_mn // total # of minutes
ser_time_sum = 60*rtc_hr+ rtc_mn // total # of minutes

now simply compare
yr to yr
mt to mt
dy to dy
then rtc_time_sum to ser_time_sum+30

The last compare decides if the 2 times are =>30 minutes apart.

Probably a better way but this should be fast and easy to do...

Jay

[kmp84]

Thanks mr. temtronic,

But if date/time ser= 2020/12/31 23:55 and RTC=2021/01/01 01:05 it will not trigger to true value?

[Ttelmah]

The other way is to use the time.h/time.c functions. These provide Unix
compatible time functions. So you can convert 'time' in ASCII format into a
int32 value, and convert the other way. You can then do comparisons etc.,
on the value. difftime, gives you the difference (in seconds) between two
time values.

[temtronic]

re:

[kmp84]

Thank you mr. temtronic, mr. Ttelmah!

I think "time.c" and "time.h" will do the my job.

Best Regards!

[kmp84]

Hi again,
I want to ask you why in <time.h>

[Ttelmah]

That is standard in Unix. It's done to allow the code to handle going
back in time before the origin point. Otherwise it couldn't cope with times
before this.
Also, remember that when doing arithmetic, the type used is the type of
the variables. If the variables were 'unsigned', difftime couldn't have a
-ve result.

[kmp84]

I ask you because range of sec. in years variables are reduce to half.

Thanks,

[Ttelmah]

I understand totally.
I had the same thought myself. Using an instrument that can only be
recording 'now' times, realised one could double the available range
of years, by using unsigned not signed. However in fact it worked much
better to leave it as signed, and just set the option to move the origin to
2010. (#define TIME_T_USES_2010, before you load time.c/time.h).

[kmp84]

Yes,
I'm using this option "#define TIME_T_USES_2010" and for current project it will be useful.

Thanks for help mr. Ttelmah!

Best Regards!

[jaka]

I have gone through all these same considerations.

With "#define TIME_T_USES_2010" the time overflows 2078. One of my projects has a remote possibility that it could still be in use then. By that time, I am 99 years old and too demented to fix it (not to mention that any of the required tools would be available!).

My solution was to modify time.h and time.c to add own functions which use 'unix minutes' instead of 'unix seconds'. At least mktime() and localtime() were fairly easy to modify this way. In my application I am using these functions to convert time between time zones, so accuracy of one minute is OK. The seconds don't need conversion, and can be directly copied from the original time.

[kmp84]

It's a good idea, I've also think about it but not realized yet.

Best Regards!

[kmp84]

Hello,

I have modify CCS 's "time_t mktime(rtc_time_t *timeT)" function as as follows :

[Ttelmah]

Glaring issue:

MM[1-12]

Month numbers are 0-11.

Having the month number wrong will mean that the number of days
in the month will be calculated incorrectly. Result "calculation for different
months are wrong".....