Parsing time string

[Aleksander Braula]

Hi,

I'm in a need of an advice for parsing specific time strings (duration) to get the number of seconds.

The time string can be either

X,Y - number, one or two digit

"Xh Ymn" or "Xmn Ys" or "Xs Yms" (output of mediainfo command line interface)

and I need a way of converting it into seconds or milliseconds.

I though of using regexps for matching different combination an than multiplying the result accordingly but maybe there is a simpler way.

[Ulf Dittmer]

The java.text.SimpleDateFormat class and its various parse methods may help with that.

[Aleksander Braula]

I've come up with:

import java.text.SimpleDateFormat;
import java.util.Date;
public class Testo {
public static void main(String[] args) {
Testo t = new Testo();
t.run();
}
private void run(){
String[] times = {
"1h 1mn", "10h 1mn", "10h 10mn", "1h 10mn",
"1mn 1s", "10mn 1s", "10mn 10s", "1mn 10s",
"1s 1ms", "10s 1ms", "10s 10ms", "1s 10ms",
"0h 1mn", "0mn 1s", "0s 1ms", "0s 10ms"
};
for (String time : times)
parse(time);
}
private void parse(String dateString) {
SimpleDateFormat[] sdf = new SimpleDateFormat[3];
sdf[0] = new SimpleDateFormat("kk'h' mm'mn'");
sdf[1] = new SimpleDateFormat("mm'mn' ss's'");
sdf[2] = new SimpleDateFormat("ss's' SSS'ms'");
for (SimpleDateFormat df : sdf ) {
try {
Date date = df.parse(dateString);
System.out.println("Parsed string: " + dateString + " as seconds: "
+ (date.getTime()/1000 + 3600) + " with pattern: " + df.toPattern());
break;
} catch (Exception e){
}
}
}
}

but for some reason I have to correct the result by 3600s (or 3600000ms if I dont divide it by 1000) otherwise I get results like:

Parsed string: 1h 1mn as seconds: 60 with pattern: kk'h' mm'mn'
Parsed string: 10h 1mn as seconds: 32460 with pattern: kk'h' mm'mn'
Parsed string: 10h 10mn as seconds: 33000 with pattern: kk'h' mm'mn'
Parsed string: 1h 10mn as seconds: 600 with pattern: kk'h' mm'mn'
Parsed string: 1mn 1s as seconds: -3539 with pattern: mm'mn' ss's'
Parsed string: 10mn 1s as seconds: -2999 with pattern: mm'mn' ss's'
Parsed string: 10mn 10s as seconds: -2990 with pattern: mm'mn' ss's'
Parsed string: 1mn 10s as seconds: -3530 with pattern: mm'mn' ss's'
Parsed string: 1s 1ms as seconds: -3598 with pattern: ss's' SSS'ms'
Parsed string: 10s 1ms as seconds: -3589 with pattern: ss's' SSS'ms'
Parsed string: 10s 10ms as seconds: -3589 with pattern: ss's' SSS'ms'
Parsed string: 1s 10ms as seconds: -3598 with pattern: ss's' SSS'ms'
Parsed string: 0h 1mn as seconds: -3540 with pattern: kk'h' mm'mn'
Parsed string: 0mn 1s as seconds: -3599 with pattern: mm'mn' ss's'
Parsed string: 0s 1ms as seconds: -3599 with pattern: ss's' SSS'ms'
Parsed string: 0s 10ms as seconds: -3599 with pattern: ss's' SSS'ms'

Anyone knows why it happens ?

[Ulf Dittmer]

Date.getTime returns something that is in terms of the GMT timezone, while the DateFormat classes use the timezone that's the default for where you're based. So I'd guess that you're one hour away from GMT. Try dates like "1h 1mn GMT" with a pattern like "kk'h' mm'mn' z" to accommodate that.

There are probably methods in the Calendar API to find out the current offset from GMT so that you can add/subtract it.

[Jesper de Jong]

Note that a java.util.Date object represents a specific point in time. It does not represent a duration, and you can't really use java.util.Date objects to store durations. In fact, the standard Java API does not have anything to store time durations at all. Despite what Ulf says, you can't use SimpleDateFormat really well for working with durations.

To work with durations, you can use a third-party library such as the popular open source Joda Time library. You can use the org.joda.time.format.PeriodFormatter class of that library to parse and format strings that represent durations.

[Aleksander Braula]

Yeah, I understand that might not be the best solution but it seem to be working for now.

I'll check the Joda Time later on, thanks.