Datevec to Datetime conversion results in 1 millisecond error?

When converting a datevector to a datetime, it appears that I'm losing a millisecond and I can't figure out why. It doesn't appear to be a rounding or formatting issue. Any help on what's going on here would be much appreciated.

Thanks,

Brad

Answer 1

[Peter Perkins]

Screen shots are probably not the most convenient way for others to diagnose this problem.

You haven't lost a millisecond, you never had it to begin with. Here's the deal:

>> dv = [2015 8 13 21 6 19.172; 2015 8 13 21 6 19.180]
dv =
         2015            8           13           21            6       19.172
         2015            8           13           21            6        19.18

You only think that last value is 19.18. It isn't. It's the floating point value that's closest to 19.18, which happens to be just less than the real number 19.18. The display simply shields you from that. With "numbers", that's probably what you'd want display to do, but in "time", you'd probably be unhappy if the display for something just before midnight rounded up to the next day. So:

>> dt = datetime(2015,8,13,21,6,[19.172 19.180],'Format','dd-MM-yyyy HH:mm:ss.SSSSSS')
dt = 
   13-08-2015 21:06:19.172000   13-08-2015 21:06:19.179999

Your problem is that by the time you've typed "19.18", you're done for. datetime has the precision to give you what you want, though:

>> dt = datetime('13-08-2015 21:06:19.180','Format','dd-MM-yyyy HH:mm:ss.SSSSSS')
dt = 
   13-08-2015 21:06:19.180000

>> dt = datetime(2015,8,13,21,6,19,[172 180],'Format','dd-MM-yyyy HH:mm:ss.SSSSSS')
dt = 
   13-08-2015 21:06:19.172000   13-08-2015 21:06:19.180000

[Brad]

Excellent! So it WAS a floating point / rounding / display issue. Great example process. Very clear and easy to follow.

You're right, the screenshot is difficult to see and work with. I will post inline code for future questions and answers.

Many thanks,

Brad

Answer 2

[Ingrid]

I do not seem to be able to reproduce this deviation, so how did you obtain the DV and DT variables?

>> timeVec = [2015 8 13 21 6 19.1800]

timeVec =

     1.0e+03 *

    Columns 1 through 5

     2.015000000000000   0.008000000000000   0.013000000000000   0.021000000000000   0.006000000000000

    Column 6

     0.019180000000000

>> timeNum = datenum(timeVec)

timeNum =

       7.361898793886574e+05

>> timeVecBack = datevec(timeNum)

timeVecBack =

     1.0e+03 *

    Columns 1 through 4

     2.015000000000000   0.008000000000000   0.013000000000000   0.021000000000000

    Columns 5 through 6

     0.006000000000000   0.019180000000008

[Brad]

Hi Ingrid,

I'm using the 'datetime' data type, not serial date numbers as you have done with 'datenum()'.

The DV (as in "datevector") and DT (as in "datetime) variables were obtained from timestamp metadata from TIFF images. I converted this serial string data to a datevector, and then to a datetime to be able to easily correct and fix timestamp issues.