How can I configure JPA/Hibernate to store a date/time in the database as UTC (GMT) time zone? Consider this annotated JPA entity:
public class Event {
#Id
public int id;
#Temporal(TemporalType.TIMESTAMP)
public java.util.Date date;
}
If the date is 2008-Feb-03 9:30am Pacific Standard Time (PST), then I want the UTC time of 2008-Feb-03 5:30pm stored in the database. Likewise, when the date is retrieved from the database, I want it interpreted as UTC. So in this case 530pm is 530pm UTC. When it's displayed it will be formatted as 9:30am PST.
Since Hibernate 5.2, you can now force the UTC time zone by adding the following configuration property into the properties.xml JPA configuration file:
<property name="hibernate.jdbc.time_zone" value="UTC"/>
If you're using Spring Boot, then add this property to your application.properties file:
spring.jpa.properties.hibernate.jdbc.time_zone=UTC
To the best of my knowledge, you need to put your entire Java app in UTC timezone (so that Hibernate will store dates in UTC), and you'll need to convert to whatever timezone desired when you display stuff (at least we do it this way).
At startup, we do:
TimeZone.setDefault(TimeZone.getTimeZone("Etc/UTC"));
And set the desired timezone to the DateFormat:
fmt.setTimeZone(TimeZone.getTimeZone("Europe/Budapest"))
Hibernate is ignorant of time zone stuff in Dates (because there isn't any), but it's actually the JDBC layer that's causing problems. ResultSet.getTimestamp and PreparedStatement.setTimestamp both say in their docs that they transform dates to/from the current JVM timezone by default when reading and writing from/to the database.
I came up with a solution to this in Hibernate 3.5 by subclassing org.hibernate.type.TimestampType that forces these JDBC methods to use UTC instead of the local time zone:
public class UtcTimestampType extends TimestampType {
private static final long serialVersionUID = 8088663383676984635L;
private static final TimeZone UTC = TimeZone.getTimeZone("UTC");
#Override
public Object get(ResultSet rs, String name) throws SQLException {
return rs.getTimestamp(name, Calendar.getInstance(UTC));
}
#Override
public void set(PreparedStatement st, Object value, int index) throws SQLException {
Timestamp ts;
if(value instanceof Timestamp) {
ts = (Timestamp) value;
} else {
ts = new Timestamp(((java.util.Date) value).getTime());
}
st.setTimestamp(index, ts, Calendar.getInstance(UTC));
}
}
The same thing should be done to fix TimeType and DateType if you use those types. The downside is you'll have to manually specify that these types are to be used instead of the defaults on every Date field in your POJOs (and also breaks pure JPA compatibility), unless someone knows of a more general override method.
UPDATE: Hibernate 3.6 has changed the types API. In 3.6, I wrote a class UtcTimestampTypeDescriptor to implement this.
public class UtcTimestampTypeDescriptor extends TimestampTypeDescriptor {
public static final UtcTimestampTypeDescriptor INSTANCE = new UtcTimestampTypeDescriptor();
private static final TimeZone UTC = TimeZone.getTimeZone("UTC");
public <X> ValueBinder<X> getBinder(final JavaTypeDescriptor<X> javaTypeDescriptor) {
return new BasicBinder<X>( javaTypeDescriptor, this ) {
#Override
protected void doBind(PreparedStatement st, X value, int index, WrapperOptions options) throws SQLException {
st.setTimestamp( index, javaTypeDescriptor.unwrap( value, Timestamp.class, options ), Calendar.getInstance(UTC) );
}
};
}
public <X> ValueExtractor<X> getExtractor(final JavaTypeDescriptor<X> javaTypeDescriptor) {
return new BasicExtractor<X>( javaTypeDescriptor, this ) {
#Override
protected X doExtract(ResultSet rs, String name, WrapperOptions options) throws SQLException {
return javaTypeDescriptor.wrap( rs.getTimestamp( name, Calendar.getInstance(UTC) ), options );
}
};
}
}
Now when the app starts, if you set TimestampTypeDescriptor.INSTANCE to an instance of UtcTimestampTypeDescriptor, all timestamps will be stored and treated as being in UTC without having to change the annotations on POJOs. [I haven't tested this yet]
With Spring Boot JPA, use the below code in your application.properties file and obviously you can modify timezone to your choice
spring.jpa.properties.hibernate.jdbc.time_zone = UTC
Then in your Entity class file,
#Column
private LocalDateTime created;
Adding an answer that's completely based on and indebted to divestoclimb with a hint from Shaun Stone. Just wanted to spell it out in detail since it's a common problem and the solution is a bit confusing.
This is using Hibernate 4.1.4.Final, though I suspect anything after 3.6 will work.
First, create divestoclimb's UtcTimestampTypeDescriptor
public class UtcTimestampTypeDescriptor extends TimestampTypeDescriptor {
public static final UtcTimestampTypeDescriptor INSTANCE = new UtcTimestampTypeDescriptor();
private static final TimeZone UTC = TimeZone.getTimeZone("UTC");
public <X> ValueBinder<X> getBinder(final JavaTypeDescriptor<X> javaTypeDescriptor) {
return new BasicBinder<X>( javaTypeDescriptor, this ) {
#Override
protected void doBind(PreparedStatement st, X value, int index, WrapperOptions options) throws SQLException {
st.setTimestamp( index, javaTypeDescriptor.unwrap( value, Timestamp.class, options ), Calendar.getInstance(UTC) );
}
};
}
public <X> ValueExtractor<X> getExtractor(final JavaTypeDescriptor<X> javaTypeDescriptor) {
return new BasicExtractor<X>( javaTypeDescriptor, this ) {
#Override
protected X doExtract(ResultSet rs, String name, WrapperOptions options) throws SQLException {
return javaTypeDescriptor.wrap( rs.getTimestamp( name, Calendar.getInstance(UTC) ), options );
}
};
}
}
Then create UtcTimestampType, which uses UtcTimestampTypeDescriptor instead of TimestampTypeDescriptor as the SqlTypeDescriptor in the super constructor call but otherwise delegates everything to TimestampType:
public class UtcTimestampType
extends AbstractSingleColumnStandardBasicType<Date>
implements VersionType<Date>, LiteralType<Date> {
public static final UtcTimestampType INSTANCE = new UtcTimestampType();
public UtcTimestampType() {
super( UtcTimestampTypeDescriptor.INSTANCE, JdbcTimestampTypeDescriptor.INSTANCE );
}
public String getName() {
return TimestampType.INSTANCE.getName();
}
#Override
public String[] getRegistrationKeys() {
return TimestampType.INSTANCE.getRegistrationKeys();
}
public Date next(Date current, SessionImplementor session) {
return TimestampType.INSTANCE.next(current, session);
}
public Date seed(SessionImplementor session) {
return TimestampType.INSTANCE.seed(session);
}
public Comparator<Date> getComparator() {
return TimestampType.INSTANCE.getComparator();
}
public String objectToSQLString(Date value, Dialect dialect) throws Exception {
return TimestampType.INSTANCE.objectToSQLString(value, dialect);
}
public Date fromStringValue(String xml) throws HibernateException {
return TimestampType.INSTANCE.fromStringValue(xml);
}
}
Finally, when you initialize your Hibernate configuration, register UtcTimestampType as a type override:
configuration.registerTypeOverride(new UtcTimestampType());
Now timestamps shouldn't be concerned with the JVM's time zone on their way to and from the database. HTH.
You would think this common problem would be taken care of by Hibernate. But its not! There are a few "hacks" to get it right.
The one I use is to store the Date as a Long in the database. So I am always working with milliseconds after 1/1/70. I then have getters and setters on my Class that return/accept only Dates. So the API remains the same. The down side is that I have longs in the database. SO with SQL I can pretty much only do <,>,= comparisons -- not fancy date operators.
Another approach is to user a custom mapping type as described here:
http://www.hibernate.org/100.html
I think the correct way to deal with this is to use a Calendar instead of a Date though. With the Calendar you can set the TimeZone before persisting.
NOTE: Silly stackoverflow won't let me comment, so here is a response to david a.
If you create this object in Chicago:
new Date(0);
Hibernate persists it as "12/31/1969 18:00:00". Dates should be devoid of timezone, so I'm not sure why the adjustment would be made.
There are several timezones in operation here:
Java's Date classes (util and sql), which have implicit timezones
of UTC
The timezone your JVM is running in, and
the default timezone of your database server.
All of these can be different. Hibernate/JPA has a severe design deficiency in that a user cannot easily ensure that timezone information is preserved in the database server (which allows reconstruction of correct times and dates in the JVM).
Without the ability to (easily) store timezone using JPA/Hibernate then information is lost and once information is lost it becomes expensive to construct it (if at all possible).
I would argue that it is better to always store timezone information (should be the default) and users should then have the optional ability to optimize the timezone away (although it only really affects display, there is still an implicit timezone in any date).
Sorry, this post doesn't provide a work-around (that's been answered elsewhere) but it is a rationalization of why always storing timezone information around is important. Unfortunately it seems many Computer Scientists and programming practitioners argue against the need for timezones simply because they don't appreciate the "loss of information" perspective and how that makes things like internationalization very difficult - which is very important these days with web sites accessible by clients and people in your organization as they move around the world.
Please take a look at my project on Sourceforge which has user types for standard SQL Date and Time types as well as JSR 310 and Joda Time. All of the types try to address the offsetting issue. See http://sourceforge.net/projects/usertype/
EDIT: In response to Derek Mahar's question attached to this comment:
"Chris, do your user types work with Hibernate 3 or greater? – Derek Mahar Nov 7 '10 at 12:30"
Yes these types support Hibernate 3.x versions including Hibernate 3.6.
Date is not in any time zone (it is a millisecond office from a defined moment in time same for everyone), but underlying (R)DBs generally store timestamps in political format (year, month, day, hour, minute, second, ...) that is time-zone sensitive.
To be serious, Hibernate MUST be allow being told within some form of mapping that the DB date is in such-and-such timezone so that when it loads or stores it it does not assume its own...
I encountered just the same problem when I wanted to store the dates in the DB as UTC and avoid using varchar and explicit String <-> java.util.Date conversions, or setting my whole Java app in the UTC time zone (because this could lead to another unexpected issues, if the JVM is shared across many applications).
So, there is an open source project DbAssist, which allows you to easily fix the read/write as UTC date from the database. Since you are using JPA Annotations to map the fields in the entity, all you have to do is to include the following dependency to your Maven pom file:
<dependency>
<groupId>com.montrosesoftware</groupId>
<artifactId>DbAssist-5.2.2</artifactId>
<version>1.0-RELEASE</version>
</dependency>
Then you apply the fix (for Hibernate + Spring Boot example) by adding #EnableAutoConfiguration annotation before the Spring application class. For other setups installation instructions and more use examples, just refer to the project's github.
The good thing is that you don't have to modify the entities at all; you can leave their java.util.Date fields as they are.
5.2.2 has to correspond to the Hibernate version you are using. I am not sure, which version you are using in your project, but the full list of provided fixes is available on the wiki page of the project's github. The reason why the fix is different for various Hibernate versions is because Hibernate creators changed the API a couple of times between the releases.
Internally, the fix uses hints from divestoclimb, Shane and a few other sources in order to create a custom UtcDateType. Then it maps the standard java.util.Date with the custom UtcDateType which handles all the necessary time zone handling.
The mapping of the types is achieved using #Typedef annotation in the provided package-info.java file.
#TypeDef(name = "UtcDateType", defaultForType = Date.class, typeClass = UtcDateType.class),
package com.montrosesoftware.dbassist.types;
You can find an article here which explains why such a time shift occurs at all and what are the approaches to solve it.
Hibernate does not allow for specifying time zones by annotation or any other means. If you use Calendar instead of date, you can implement a workaround using HIbernate property AccessType and implementing the mapping yourself. The more advanced solution is to implement a custom UserType to map your Date or Calendar. Both solutions are explained in my blog post here: http://www.joobik.com/2010/11/mapping-dates-and-time-zones-with.html
Here is the complete checklist for storing dates in your database in a proper way:
first, make sure your OS timezone is st properly: either set TZ env variable in your Dockerfile or start your Java application with -Duser.timezone=“UTC”.
As mentioned earlier: also set the timezone for JPA with hibernate.jdbc.time_zone = UTC
Use OffsetDateTime for storing dates as UTC in your entities. Don’t use Date, Calendar & ZonedDateTime !
I'm using javaee-api 6.0.
I have an entity bean with a java.util.Date field called updated.
public class Tariff implements Serializable {
private Date updated
And I have a REST service.
#GET
#Path("/example")
public Response getTariff() {
return Response.status(200).entity(new Records(createExampleTariff())).build();
}
When I call my REST service it returns the date as a number.
{"records":{"description":"OTHER","message":"Nothing to say","status":"OK", "updated":1475822878961},"status":"ok"}
Anyone know how I can work past this, without using a DTO.
Reason:
The date is always stored as the number of milliseconds since January 1, 1970, 00:00:00 GMT represented by this Date object.
Solution:
You can convert the Date format using org.codehaus.jackson.map.JsonSerializer. by writing a class JsonDateSerializer with that conversion logic.
#JsonSerialize(using=JsonDateSerializer.class)
public Date getDate() {
return date;
}
you can get detailed explanation here
The Date class is in fact the wrapper for a number, the number being the specified number of milliseconds since the standard base time known as "the epoch", namely January 1, 1970, 00:00:00 GMT.
You need to specify what you mean by "work past this". Either you send a number or a string. For a string, you can use SimpleDateFormat to produce the string you want.
I would still go with returing the timestamp so it remains simple. If you really need to send the date in human readable format you can create a class that extends org.codehaus.jackson.map.JsonSerializer and implement the serialize method as you want. Then anotate the date getter like this
#JsonSerialize(using=JsonDateSerializer.class)
public Date getDate() {
return date;
}
Check this article for more info
I'm currently struggling with Java JSON deserialization with Jackson in the following way:
I want to process and deserialize a JSON response I get back from a webservice and convert the response into POJOs with the help of Jackson. This works fine most of the time, as long as the response I'm getting contains JSON attributes in the correct format.
As however the webservice and the delivered data is out of my control, I can not rely on the data always being in the correct format.
Let me give you an example:
In my POJO, there's a java.util.Date field, and the JSON response contains a property holding a datetime string. Jackson will try to parse the string and convert it into a Date. If the date format matches the ObjectMapper's dateformat (ObjectMapper.setDateFormat(...)), everything is fine. If the format is different, I get an InvalidFormatException.
The problem now is, the dateformat sent from the service can differ. I can get dates formatted like 2014-11-02T00:00:00Z, but I can also get dates formatted like 2014-11 (identifying just a single month instead of an entire datetime).
I know, I can write a custom deserializer which could take care of this exact case and handle datestrings with different date formats correctly. But as this would only fix issues with Dates, but not with potential other datatypes, I'm searching for a more general approach.
(What happens e.g. if I expect a Double and receive an alphanumerical string?)
What I would like is to have the possibility to ignore all cases in which an InvalidFormatException happens and define a default value (like null) to the respective POJO field.
And it would be really valuable for me, if despite an invalid dateformat being returned or any other InvalidFormatException happening, the rest of the JSON would still be deserialized into the POJO.
Is this in any way possible with Jackson?
Thank you for reading my question till the end and I would be grateful for any pointers in the right direction.
Not sure if this is best practice, I have little experience with Jackson.
You can add a DeserializationProblemHandler to the ObjectMapper to specify what happens when the deserializer encounters a weird String or weird number.
In your case you could set the handler such that when encountering an unrecognized format, instead of throwing an InvalidFormatException, it just returns null:
ObjectMapper objectMapper = new ObjectMapper();
objectMapper.addHandler(new DeserializationProblemHandler() {
#Override
public Object handleWeirdStringValue(DeserializationContext ctxt, Class<?> targetType, String valueToConvert, String failureMsg) throws IOException {
return null;
}
#Override
public Object handleWeirdNumberValue(DeserializationContext ctxt, Class<?> targetType, Number valueToConvert, String failureMsg) throws IOException {
return null;
}
});
I'm using Spring MVC with JSR303 to do my input validation.
A form I've created has a couple of date fields that are bound to Date objects within the object backing the form. I'm using JSR303 to do the validation for the Date using #Future. I'm also using #DateTimeFormat(pattern="dd/MM/yyyy"), (I know it's not validation).
How do I validate the date format of the String on the form? If I leave the other required fields blank (#NotEmpty) and enter a non-valid date in the form 'dd/MM/yy' it gets converted to 'dd/MM/yyyy' on re-presentation (e.g. 12/03/12 is re-presented as 12/03/0012). Which means I will get duff data in my system. If I enter "aaa" for I get a conversion Exception. Correctly formatted Strings get converted to Date objects.
Additionally should the 'required field' annotation for Date fields be #NotNull or #NotEmpty?
Many thanks in advance for any advice provided.
Thanks Ralph. I did some further digging around and came up with this (Which goes in my form controller):
#InitBinder
public void initBinder(WebDataBinder binder) {
String format = "dd/MM/yyyy";
SimpleDateFormat dateFormat = new SimpleDateFormat(format);
dateFormat.setLenient(false);
CustomDateEditor customDateEditor = new CustomDateEditor(dateFormat,true,format.length());
binder.registerCustomEditor(Date.class, customDateEditor);
}
With the properties file having the following key:
typeMismatch.java.util.Date : Some nice calm reassuring message to assist all negligent users
Maybe there are some other ways to do this but this will do for now.
You can not do this with JSR303, because the validation runs on the already poplulated (form baching) object.
So you need to implement your own custom converter, that is a bit more strickt than the one shipped with spring.
#See Spring Reference: Chapter 6.5 Spring 3 Type Conversion
I have some dates I have to represent in an XML file in a format different than the one JAXB uses by default.
So, I've written some static methods to print and parse the required format:
public static String toDateTime(XMLGregorianCalendar d) {
return new SimpleDateFormat("yyyy-MM-dd HH:mm:ss")
.format(d.toGregorianCalendar().getTime());
}
public static XMLGregorianCalendar parseDateTime(String s)
throws DatatypeConfigurationException, ParseException {
GregorianCalendar cal = new GregorianCalendar();
cal.setTime(new SimpleDateFormat("yyyy-MM-dd HH:mm:ss")
.parse(s));
return DatatypeFactory.newInstance().newXMLGregorianCalendar(cal);
}
And I've written a .xjb file specifying to use these parse and print methods:
<jxb:bindings
node="//xs:element[#name='date']">
<jxb:javaType name="javax.xml.datatype.XMLGregorianCalendar"
parseMethod="MyClass.parseDateTime"
printMethod="MyClass.toDateTime"
/>
</jxb:bindings>
(The XML schema specifies the element as an xsd:dateTime).
The xjc tool is generating an adapter that calls my methods:
public XMLGregorianCalendar unmarshal(String value) {
return (MyClass.parseDateTime(value));
}
public String marshal(XMLGregorianCalendar value) {
return (MyClass.toDateTime(value));
}
My problem is that the generated adapter won't compile. The unmarshal method is declared not to throw exceptions, but MyClass.parseDateTime does. It seems perfectly reasonable that it should, since the date might be invalid. And indeed, the unmarshal method of java.xml.bind.annotation.adapters.XmlAdapter that the generated adapter overrides declares that it throws Exception.
I couldn't find any way to declare in the xjb file which exceptions the parseMethod throws. It occurred to me that maybe I should create my own adapter class, but I couldn't find any way to declare in the xjb that my custom adapter should be used. I would very much like to be able to generate the classes from the xsd and xjb with xjc and not have to modify them afterwards.
What is the best way to resolve this situation?
You should handle both checked exceptions in your code. For an invalid date-time, you have to decide whether to leave it as null (and maybe emit a warning), or blow up and kill the whole parsing attempt (in that case, throw a RuntimeException wrapping the original ParseException).
In the case of DatatypeConfigurationException, this is one of the JDK APIs that is badly designed, throwing checked exceptions when there's nothing your code can do. Just rethrow it as RuntimeException.