How to convert UTC and local timezone in Java

Question

I am curious about timezone in Java. I want to get UTC time in milliseconds from a device and send to server. Server will convert it to local timezone when it displays time to users. Timezone in my system is Australia/Sydney( UTC + 11:00), and I have got the result below when I tested timezone:

int year = 2014;
int month = 0;
int date = 14;
int hourOfDay = 11;
int minute = 12;
int second = 0;
Calendar c1 = Calendar.getInstance();
c1.set(year, month, date, hourOfDay, minute, second);

SimpleDateFormat sdf = new SimpleDateFormat("dd/MM/yyyy HH:mm:ss z");
System.out.println(sdf.format(c1.getTime()));

Calendar c2 = Calendar.getInstance(TimeZone.getTimeZone("UTC"));
c2.set(year, month, date, hourOfDay, minute, second);
System.out.println(sdf.format(c2.getTime()));

output:

14/01/2014 11:12:00 EST
14/01/2014 22:12:00 EST

I thought I could have 13/01/2014 00:12:00 for c2 because UTC time is 11 hours later than mine. Does not Calendar work the way I expect?

Your help would be appreciated.

EDIT

Added z to display timezone. This makes me more confused because Mac says its timezone is (AEDT) Australian Eastern Daylight Time but Java is EST. Anyway still result is different because EST is UTC-5 hours.

Answer 1

tl;dr

Use modern java.time classes.

ZonedDateTime
.of( 2014 , 1 , 14 , 11 , 12 , 0 , 0 , ZoneId.of( "Australia/Sydney" ) )
.toInstant()
.toEpochMilli() 

1389658320000

Going the other direction.

Instant
.ofEpochMilli( 1_389_658_320_000L )  // .toString(): 2014-01-14T00:12:00Z
.atZone( 
    ZoneId.of( "Australia/Sydney" ) 
)                                    // .toString(): 2014-01-14T11:12+11:00[Australia/Sydney]
.format(
    DateTimeFormatter
    .ofPattern ( 
        "dd/MM/uuuu HH:mm:ss z" , 
        new Locale( "en" , "AU" ) 
    )
)

14/01/2014 11:12:00 AEDT

java.time

You are using terrible date-time classes that were made obsolete years ago by the adoption of JSR 310 defining the modern java.time classes.

I am curious about timezone in Java.

FYI, an offset-from-UTC is merely a number of hours-minutes-seconds. When we say “UTC” or put a Z at the end of a string, we mean an offset of zero hours-minutes-seconds, for UTC itself.

A time zone is much more. A time zone is a history of past, present, and future changes to the offset used by the people of a particular region. Politicians around the world have an odd penchant for changing the offset of their jurisdiction.

I want to get UTC time in milliseconds from a device and send to server.

For the current moment, use Instant. An Instant internally is the number of whole seconds seconds since the epoch reference of first moment of 1970 UTC, plus a fraction of a second in nanoseconds.

Instant now = Instant.now() ;  // Capture current moment in UTC.
long millisecondsSinceEpoch = now.toEpochMilli() ;

Going the other direction.

Instant instant = Instant.ofEpochMilli( millisecondsSinceEpoch ) ;

Server will convert it to local timezone …

Specify the time zone desired/expected by the user.

If no time zone is specified, the JVM implicitly applies its current default time zone. That default may change at any moment during runtime(!), so your results may vary. Better to specify your desired/expected time zone explicitly as an argument. If critical, confirm the zone with your user.

Specify a proper time zone name in the format of Continent/Region, such as America/Montreal, Africa/Casablanca, or Pacific/Auckland. Never use the 2-4 letter abbreviation such as EST or IST as they are not true time zones, not standardized, and not even unique(!).

ZoneId z = ZoneId.of( "America/Montreal" ) ;  
ZonedDateTime zdt = instant.atZone( z ) ;

… when it displays time to users

Automatically localize for the user's language and culture.

To localize, specify:

• FormatStyle to determine how long or abbreviated should the string be.
• Locale to determine:
  • The human language for translation of name of day, name of month, and such.
  • The cultural norms deciding issues of abbreviation, capitalization, punctuation, separators, and such.

Example:

Locale l = Locale.CANADA_FRENCH ;   // Or Locale.US, Locale.JAPAN, etc.
DateTimeFormatter f = 
    DateTimeFormatter
    .ofLocalizedDateTime( FormatStyle.FULL )
    .withLocale( l )
;
String output = zdt.format( f );

Timezone in my system is Australia/Sydney( UTC + 11:00)

The current default time zone of your server should be irrelevant to your program. Always specify the desired/expected time zone. Frankly, making optional the time zone (and Locale) argument of the various date-time methods is one of the very few design flaws in java.time framework.

Tip: Generally best to set your servers to UTC as their current default time zone.

By the way, be clear that time zone and locale have nothing to do with one another. You might want Japanese language for displaying a moment as seen in Africa/Tunis time zone.

ZoneID zAuSydney = ZoneId.of( "Australia/Sydney" ) ;
ZonedDateTime zdt = instant.atZone( zAuSydney ) ;
String output = zdt.format(
    DateTimeFormatter
    .localizedDateTime( FormatStyle.LONG )
    .withLocale( new Locale( "en" , "AU" ) ;
) ;

int year = 2014; …

Note that java.time uses sane numbering, unlike the legacy classes. Months are 1-12 for January-December, and weekdays are 1-7 for Monday-Sunday.

LocalDate ld = LocalDate.of( 2014 , 1 , 14 ) ;
LocalTime lt = LocalTime.of( 11 , 12 ) ;
ZoneId z = ZoneId.of( "Australia/Sydney" ) ;
ZonedDateTime zdt = ZonedDateTime.of( ld , lt , z ) ;

zdt.toString() = 2014-01-14T11:12+11:00[Australia/Sydney]

Generally best to automatically localize for display, as seen above. But if you insist, you can hard-code a formatting pattern.

Locale locale = new Locale ( "en" , "AU" );
ZoneId z = ZoneId.of ( "Australia/Sydney" );
ZonedDateTime zdt = ZonedDateTime.of ( 2014 , 1 , 14 , 11 , 12 , 0 , 0 , z );

zdt.toString(): 2014-01-14T11:12+11:00[Australia/Sydney]

Specify that formatting pattern of yours.

DateTimeFormatter f = DateTimeFormatter.ofPattern ( "dd/MM/uuuu HH:mm:ss z" , locale );
String output = zdt.format ( f );

output = 14/01/2014 11:12:00 AEDT

Your Question was interested in a count of milliseconds since epoch of 1970-01-01T00:00:00Z. So adjust from the Australia time zone to UTC. Same moment, same point on the timeline, different wall-clock time.

Instant instant = zdt.toInstant() ;  // Adjust from time zone to UTC.

instant.toString(): 2014-01-14T00:12:00Z

Note the difference in hour-of-day between instant and zdt.

I thought I could have 13/01/2014 00:12:00 for c2 because UTC time is 11 hours later than mine.

➥ As you asked for, twelve minutes after 11 AM in Sydney zone is the same moment as twelve minutes after midnight in UTC, because Australia/Sydney on that date is eleven hours ahead of UTC.

Calculate milliseconds since epoch.

long millisecondsSinceEpoch = instant.toEpochMilli() ;

Answer 2

UPDATE: This Answer is now out-of-date. The Joda-Time library is now supplanted by the java.time framework built into Java 8 and later. See this new Answer.

Three-Letter Codes

You should avoid using 3 or 4 letter time zone codes such as EST or IST. They are neither standard nor unique.

Use proper time zone names, mostly Continent/CityOrRegion such as America/Montreal or Asia/Kolkata.

Joda-Time

The java.util.Date/Calendar classes are notoriously bad. Avoid using them. Use either Joda-Time or, in Java 8, the new java.time.* classes defined by JSR 310 and inspired by Joda-Time.

Notice how much simpler and more obvious is the Joda-Time code shown below. Joda-Time even knows how to count – January is 1, not 0!

Time Zone

In Joda-Time, a DateTime instance knows its own time zone.

Sydney Australia has a standard time of 10 hours ahead of UTC/GMT, and a Daylight Saving Time (DST) of 11 hours ahead. DST applies to the date specified by the question.

Tip: Don't think like this…

UTC time is 11 hours later than mine

Think like this…

Sydney DST is 11 hours ahead of UTC/GMT.

Date-time work becomes easier and less error-prone if you think, work, and store in UTC/GMT. Only convert to localized date-time for presentation in the user-interface. Think globally, display locally. Your users and your servers can easily move to other time zones, so forget about your own time zone. Always specify a time zone, never assume or rely on default.

Example Code

Here is some example code using Joda-Time 2.3 and Java 8.

// Better to specify a time zone explicitly than rely on default.
// Use time zone names, not 3-letter codes. 
// This list is not quite up-to-date (read page for details): http://joda-time.sourceforge.net/timezones.html
DateTimeZone timeZone = DateTimeZone.forID("Australia/Sydney");
DateTime dateTime = new DateTime(2014, 1, 14, 11, 12, 0, timeZone);
DateTime dateTimeUtc = dateTime.toDateTime(DateTimeZone.UTC); // Built-in constant for UTC (no time zone offset).

Dump to console…

System.out.println("dateTime: " + dateTime);
System.out.println("dateTimeUtc: " + dateTimeUtc);

When run…

dateTime: 2014-01-14T11:12:00.000+11:00
dateTime in UTC: 2014-01-14T00:12:00.000Z

Answer 3

You probably meant to set the timezone on your formatter, not the Calendar (or in addition the the Calendar, it is not 100% clear what you mean to accomplish)! The timezone used to create the human representation comes from the SimpleDateFormat. All "timezone" information is lost from the Calendar when you convert it back into a java.util.Date by calling getTime().

The code:

Calendar c2 = Calendar.getInstance(TimeZone.getTimeZone("UTC"));
c2.set(year, month, date, hourOfDay, minute, second);
System.out.println(sdf.format(c2.getTime()));

is printing 14/01/2014 10:12:00 because 11AM UTC displayed in Syndey (the timezone of your formatter) is 10PM! (use HH in the format for 24 hour time)

This would print what it seems like you meant to do:

SimpleDateFormat sdf = new SimpleDateFormat("dd/MM/yyyy hh:mm:ss z");
System.out.println(sdf.format(c1.getTime()));

sdf.setTimeZone(TimeZone.getTimeZone("UTC"));
System.out.println(sdf.format(c1.getTime()));

The concept of 'UTC milliseconds' is meaningless. A quantity of milliseconds is just a fixed point in history, it has no timezone associated with it. We add a timezone to it to convert it into human-readable representations.

edit: Yes, the ambiguity of using 'EST' for both (US) Eastern Time and (Australian) Eastern Time has been a pitfall in Java since forever.