converte UNIX time to Human readable format.

Hello,

Work is almost done, but there came one more problem: how can I converte the UNIX time to human readable format, like ISO086..

My UNIX time is in int64 format, for example: 1352434077. So when I tried to converte it with this code:

unix_epoch = datenum(1970,1,1,0,0,0);
for i=1:1:size(data_mat)
matlab_time(i,1) = data_mat(i,1)./86400 + unix_epoch; 
end

It gave 735087..not something like YYYY-MM-DD: hh:mm:ss ...can anyone gave me a hint? Thanks a lot!

Answer 1

[Bjorn Gustavsson]

Take a look at datestr (or some of the other functions in the date-family):

datestr(matlab_time,'yyyymmddTHH:MM:SS')

or whatever format you might choose.

HTH

[buer]

Thanks for answering...In fact, I tried this function many times..

datestr(1352434077/86400+datenum('1/1/1970'))
09-Nov-2012 04:07:57

So when I input only one number, it works fine. But my time series are stored in 13892780 * 2 int64 format..so when I tried this function in for loop, it always gave errors:

    Error using dateformverify (line 28)
    DATESTR failed converting date number to date vector.

    Error in datestr (line 195)
    S = dateformverify(dtnumber, dateformstr, islocal);

    Caused by:
        Error using datevecmx
        The datevecmx function only accepts double arrays.

That is why I am thinking is it related to int64 format..

[buer]

very interesting..

datestr(1352435088/86400 + datenum('1/1/1970'))

ans =

09-Nov-2012 04:24:48

if only one number. works fine. but with a mat file does not work

[Bjorn Gustavsson]

Well this seems to work just fine:

datestr([1;1;1]*matlab_time,'yyyymmddTHH:MM:SS')

...so you'll have to cast the time-array to double as the error message indicates...

HTH

[buer]

Yes, Bjorn...I also tried this way..This is my data:

Then I do

xx=typecast(Data(:,1),'double');

Then the result is like this:

 How can it be like this...I need to converte them to date time...please check it for me...I have spend quite few days already....thanks

[Bjorn Gustavsson]

Hi,

I think you miss the most obvious way:

>> time64 = int64(now);  % Not the best way to handle a time...
>> datestr([1;1;1]*double(time64),'yyyymmddTHH:MM:SS')
ans =
20141222T00:00:00
20141222T00:00:00
20141222T00:00:00

Should be fine if you convert your dates into doubles - if not then you might have something faulty either in the date format or in the actual date-numbers?

If the above doesn't work check your implementation with some dates that you set yourself, both the conversion to doubles and then to date-strings, and saveing to the date-format your trying to read...

HTH

[buer]

Thanks Bjorn,

I tried with

x=datestr([1]*double(Data(1:10,1)./86400+datenum(1970,1,1,0,0,0)),'yyyymmddTHH:MM:SS');

The result is

20121109T00:00:00
20121109T00:00:00
20121109T00:00:00
20121109T00:00:00
20121109T00:00:00

20121109T00:00:00
20121109T00:00:00
20121109T00:00:00
20121109T00:00:00
20121109T00:00:00

Why the hour, minute and second are all 0? they should be different...

[Bjorn Gustavsson]

I guess you should do the typecasting to double ASAP, that is on the Data(1:10,1) before dividing by 86400 - as it is now that division will give you an int - and that corresponds to midnight since the time of day is in fractions of a day...

HTH

[buer]

Thanks Bjorn....it is really about the typecasting..thanks a lot