How to convert 64 bit integer timestamp to yyyymmddHHMMSSffffff format?

Hi

I have a timeseries data whose each record has a timestamp in microseconds and are recorded in 64 bit integer format. Data starts from May 1,2011, Sunday, 19:00 EDT. However, I want create a histogram of the interarrival times. Therefore, I want to first convert the timestamps into yyyymmddHHMMSSffffff format and then store it as an integer/double.

A few sample values are: 3679705205, 4998039591, 5638258864, 10464755368, and so on. Therefore, I shall be really thankful if someone can help write a small piece of code to solve this issue.

Thanks,

Sourav

[Cris LaPierre]

Help us out by sharing what the datetime equivalent of 3679705205, 4998039591, 5638258864, and 10464755368 are.

[Sourav Mondal]

I don't know exactly. All I know is that the data starts from May 1, 2011, Sunday, 19:00 EDT.

Answer 1

[Cris LaPierre]

My best guess would be to use datetime and set the 'Epoch' and 'TicksPerSecond' name-value pairs. Perhaps something like this?

t = [3679705205; 4998039591; 5638258864; 10464755368];
d = datetime(t,'ConvertFrom','epochtime','Epoch',datetime(2011,5,1,19,0,0),'TicksPerSecond',1e6)

d = 

  4×1 datetime array

   01-May-2011 20:01:19
   01-May-2011 20:23:18
   01-May-2011 20:33:58
   01-May-2011 21:54:24

[Cris LaPierre]

To modify the output format, use the 'Format' name-value pair as well.

d = datetime(t,'ConvertFrom','epochtime','Epoch',datetime(2011,5,1,19,0,0),'TicksPerSecond',1e6,'Format','yyyyMMddHHmmssSSSSSS')

d = 

  4×1 datetime array

   20110501200119705205
   20110501202318039591
   20110501203358258863
   20110501215424755367

[Sourav Mondal]

Thank you very much for the answer. This works perfectly. Can I convert the datetime array into an integer/double array that stores exactly the same numbers shown above?

[Steven Lord]

No. There's no integer type in MATLAB large enough to store those exactly and it is greater than flintmax. If you try to make it a uint64 it saturates at intmax.

>> uint64(20110501200119705205)

ans =

  uint64

   18446744073709551615

But if you want to create a histogram you can do that with the original datetime array. There's no need to convert to a numeric array.

t = [3679705205; 4998039591; 5638258864; 10464755368];
d = datetime(t,'ConvertFrom','epochtime',...
               'Epoch',datetime(2011,5,1,19,0,0),...
               'TicksPerSecond',1e6,...
               'Format','yyyyMMddHHmmssSSSSSS');
histogram(d)

Or if you want the histogram of the difference between those times:

histogram(diff(d), 'BinWidth', minutes(5)) % 5 minute wide bins

[Sourav Mondal]

Thank you!