Why does Java have transient fields?
Solution 1
The transient
keyword in Java is used to indicate that a field should not be part of the serialization (which means saved, like to a file) process.
From the Java Language Specification, Java SE 7 Edition, Section 8.3.1.3. transient
Fields:
Variables may be marked
transient
to indicate that they are not part of the persistent state of an object.
For example, you may have fields that are derived from other fields, and should only be done so programmatically, rather than having the state be persisted via serialization.
Here's a GalleryImage
class which contains an image and a thumbnail derived from the image:
class GalleryImage implements Serializable
{
private Image image;
private transient Image thumbnailImage;
private void generateThumbnail()
{
// Generate thumbnail.
}
private void readObject(ObjectInputStream inputStream)
throws IOException, ClassNotFoundException
{
inputStream.defaultReadObject();
generateThumbnail();
}
}
In this example, the thumbnailImage
is a thumbnail image that is generated by invoking the generateThumbnail
method.
The thumbnailImage
field is marked as transient
, so only the original image
is serialized rather than persisting both the original image and the thumbnail image. This means that less storage would be needed to save the serialized object. (Of course, this may or may not be desirable depending on the requirements of the system -- this is just an example.)
At the time of deserialization, the readObject
method is called to perform any operations necessary to restore the state of the object back to the state at which the serialization occurred. Here, the thumbnail needs to be generated, so the readObject
method is overridden so that the thumbnail will be generated by calling the generateThumbnail
method.
For additional information, the article Discover the secrets of the Java Serialization API (which was originally available on the Sun Developer Network) has a section which discusses the use of and presents a scenario where the transient
keyword is used to prevent serialization of certain fields.
Solution 2
Before understanding the transient
keyword, one has to understand the concept of serialization. If the reader knows about serialization, please skip the first point.
What is serialization?
Serialization is the process of making the object's state persistent. That means the state of the object is converted into a stream of bytes to be used for persisting (e.g. storing bytes in a file) or transferring (e.g. sending bytes across a network). In the same way, we can use the deserialization to bring back the object's state from bytes. This is one of the important concepts in Java programming because serialization is mostly used in networking programming. The objects that need to be transmitted through the network have to be converted into bytes. For that purpose, every class or interface must implement the Serializable
interface. It is a marker interface without any methods.
Now what is the transient
keyword and its purpose?
By default, all of object's variables get converted into a persistent state. In some cases, you may want to avoid persisting some variables because you don't have the need to persist those variables. So you can declare those variables as transient
. If the variable is declared as transient
, then it will not be persisted. That is the main purpose of the transient
keyword.
I want to explain the above two points with the following example (borrowed from this article):
package javabeat.samples; import java.io.FileInputStream; import java.io.FileOutputStream; import java.io.IOException; import java.io.ObjectInputStream; import java.io.ObjectOutputStream; import java.io.Serializable; class NameStore implements Serializable{ private String firstName; private transient String middleName; private String lastName; public NameStore (String fName, String mName, String lName){ this.firstName = fName; this.middleName = mName; this.lastName = lName; } public String toString(){ StringBuffer sb = new StringBuffer(40); sb.append("First Name : "); sb.append(this.firstName); sb.append("Middle Name : "); sb.append(this.middleName); sb.append("Last Name : "); sb.append(this.lastName); return sb.toString(); } } public class TransientExample{ public static void main(String args[]) throws Exception { NameStore nameStore = new NameStore("Steve", "Middle","Jobs"); ObjectOutputStream o = new ObjectOutputStream(new FileOutputStream("nameStore")); // writing to object o.writeObject(nameStore); o.close(); // reading from object ObjectInputStream in = new ObjectInputStream(new FileInputStream("nameStore")); NameStore nameStore1 = (NameStore)in.readObject(); System.out.println(nameStore1); } }
And the output will be the following:
First Name : Steve Middle Name : null Last Name : Jobs
Middle Name is declared as transient
, so it will not be stored in the persistent storage.
Solution 3
To allow you to define variables that you don't want to serialize.
In an object you may have information that you don't want to serialize/persist (perhaps a reference to a parent factory object), or perhaps it doesn't make sense to serialize. Marking these as 'transient' means the serialization mechanism will ignore these fields.
Solution 4
Why are transient fields needed in Java?
The transient
keyword gives you some control over the serialization process and allows you to exclude some object properties from this process. The serialization process is used to persist Java objects, mostly so that their states can be preserved while they are transferred or inactive. Sometimes, it makes sense not to serialize certain attributes of an object.
Which fields should you mark transient?
Now that we know the purpose of the transient
keyword and transient fields, it's important to know which fields to mark transient. Static fields aren't serialized either, so the corresponding keyword would also do the trick. But this might ruin your class design; this is where the transient
keyword comes to the rescue. I try not to allow fields whose values can be derived from others to be serialized, so I mark them transient. If you have a field called interest
whose value can be calculated from other fields (principal
, rate
& time
), there is no need to serialize it.
Another good example is with article word counts. If you are saving an entire article, there's really no need to save the word count, because it can be computed when article gets "deserialized." Or think about loggers; Logger
instances almost never need to be serialized, so they can be made transient.
Solution 5
A transient
variable is a variable that isn't included when the class is serialized.
One example of when this might be useful that comes to mind is, variables that make only sense in the context of a specific object instance and which become invalid once you have serialized and deserialized the object. In that case it is useful to have those variables become null
instead so that you can re-initialize them with useful data when needed.
Animesh
Updated on July 08, 2022Comments
-
Animesh almost 2 years
Why does Java have transient fields?
-
Joachim Sauer over 14 yearsI think that
transient
wouldn't be a keyword if it were designed at this time. They'd probably use an annotation. -
Elazar Leibovich over 13 yearsBut why is it a keyword, and not an annotation
@DoNotSerialize
? -
Peter Wippermann about 13 yearsI guess, this is owned to a time when there were no annotations in Java.
-
caleb over 12 yearsI find it odd that serializable is internal to Java. It can be implemented as an interface or abstract class that requires users to override the read and write methods.
-
Krishna over 11 yearsThis example is taken from this code, you can read it here:javabeat.net/2009/02/what-is-transient-keyword-in-java
-
Mohammad Jafar Mashhadi almost 11 years
readObject
method is called automatically or we should call it? -
TFennis almost 11 yearsCorrect me if I'm wrong, but Thread is not serializable so it will be skipped anyways?
-
Mike Adler almost 11 years@MJafar: readObject is usually chained into deserialization mechanisms and thus called automatically. Furthermore, in many cases you do not need to override it - the default implementation does the trick.
-
A.H. over 10 years@TFennis: If a serializable class
A
references a not serializable classB
(likeThread
in your example), thenA
must either mark the reference astransient
XOR must override the default serialization process in order to do something reasonable withB
XOR assume that only serializable subclasses ofB
are actually referenced (so the actual subclass must take care for their "bad" parentB
) XOR accept that the serialization will fail. In only one case (marked as transient)B
is automatically and silently skipped. -
Admin over 10 years@caleb probably because dealing with binary formats yourself is incredibly painful in Java due to the lack of unsigned integers.
-
Danny Gloudemans over 10 yearsPlease provide more information when you give an answer.
-
Hugo over 10 years@ElazarLeibovich I think is because this keyword is older than the annotations: en.wikipedia.org/wiki/Java_annotation#History
-
glen3b about 10 yearsThere are use cases other than sensitive data in which you may not want to serialize a field. For example, you probably would never want to serialize a
Thread
(credit to @A.H. for example), in which case you would mark it as transient. However, a thread is not sensitive data in and of itself, it just makes no logical sense to serialize it (and it is not serializable). -
sproketboy almost 10 yearsNote that XML serialization in Java ignores this keyword for some reason.
-
Garcia Hurtado over 9 yearsThis part strikes me as odd and possibly confusing: "That means the state of the object is converted into a stream of bytes and stored in a file". It seems to me that most of the time serialization does not involve writing to a file (case in point: the networking examples that follow)
-
zatenzu over 9 yearsWhy I can set transient on an instance variable of a class when this class doesn't implement the Serializable interface? This makes no sense...
-
user207421 about 9 yearsYour 'simple sentence' is merely a tautology. It explains nothing. You'd be better off without it.
-
user207421 about 9 years@TFennis No, it will cause an exception.
-
user207421 about 9 years@glen3b That case isn't excluded by this answer. It certainly is needed as things stand in the case the poster mentioned.
-
Novaterata about 8 years@zatenzu A child class could implement Serializable
-
Holger about 8 years@novaterata: when a
Serializable
child class extends a non-Serializable
class, the child class becomes responsible for storing and restoring all relevant state of its superclass(es) (and the superclass must have an accessible no-arg constructor). Therefore,transient
modifiers on the non-Serializable
superclass’ fields still have no relevance then. But the language specification doesn’t forbid usingtransient
for storage mechanisms other than Serialization. -
Sarkhan over 7 yearswhy Serializable interface is market don't include readResolve,writeObject etc. because sub class and also super class's readResolve,writeObject etc. methods must be invoked by jvm.If it were overridable then super class's readResolve method could not be invoked
-
supercat about 7 years@A.H.: Why XOR? I would think code that did any combination of those things would work, and some combinations might be useful (e.g. overriding the default serialization process may be useful even if only serializable subclasses of B are referenced, and vice versa).
-
user207421 almost 7 years@caleb It isn't 'internal to Java', it is part of an API, and it is implemented as an interface. Your point escapes me.
-
tgkprog almost 7 yearsThe default serializer can access parent classes using unsafe operations/ using native code. So the transient keyword is used when you use the default java serialization. Useful to send class state in RMI or in a custom binary over REST/http implementation. Faster than json/ xml transformers.
-
Raphael over 6 yearsThe example is a bad one, since the middle name is clearly not a transient property.
-
sleske about 6 years@caleb: After many years, Oracle agrees with you - they want to drop serialization: infoworld.com/article/3275924/java/…
-
Arefe over 5 years@Raphael For me, the example is helpful and at least explains the concept. Would you provide any better example if you are aware off?
-
Raphael over 5 years@Yoda We could have a reference to a
NameFormatter
used to drivetoString()
. Or anything else related to configuration, or viewing, or business logic, ... as opposed to data. -
Arefe over 5 yearsThis is a good explanation where the field should be
transient
-
Tarun over 5 years@Raphael - A practical example would be In a LinkedList size could be a transient variable as it could be recalculated by after deserializing the object.
-
Tarun over 5 yearsinterest field and word counts are good examples of transient fields.
-
Mohammed Siddiq over 5 yearsAnother good use case: If your object has components like sockets and if you would want to serialize then what happens to the socket? If would persist, after you deserialize what would the socket be holding? It makes sense to make that socket object as
transient
-
Mateen over 5 yearsyeah, some thing like "password or crediCardPin" field members of a class.
-
kapad about 5 years@Holger what are the storage mechanisms other than serialization?
-
Holger about 5 years@kapad some 3rd party persistence mechanism. Might be entirely hypothetical. But they are allowed by the specification.
-
Ngọc Hy over 3 yearsIs it correct that in case of dedecting whether an object has changed after some operations, this concept is one way to go?
-
Zoe stands with Ukraine about 3 yearsThe only thing left out of which fields should be marked transient is classes that literally can't be serialized, for any reason. As already mentioned, it may be a Socket, or some other type of session storage, or simply a class that doesn't allow serialization - the point being, aside times where it's not necessary to serialize a field, there's times where it's actively disallowed, and
transient
becomes a requirement for the class of interest to be serialized. Also,Logger
instances tend to be static, and therefore don't need to betransient
in the first place -
aderchox almost 3 yearsWhy would an object be transferred through the network? Can you give an example?