As an alternative to static blocks, Oracle docs are suggesting to call a method(s) and the example is using a variable assignment:
public static varType myVar = initializeClassVariable();
The advantage of private static methods is that they can be reused later if you need to reinitialize the class variable.
But if I don't need (and want to avoid unused) extra variable and also return statement in my static block, what is a better way to call static code?
Calling a static method in the constructor is wrong design for executing static code once (constructor can be private for utility class) for static block
public MyClass() {
MyClass.initializeClassVariable();
}
So is the only improvement is reducing variable access level to private ?
private static varType myVar = initializeClassVariable();
Or a better approach is to keep static block and add the method there?
static {
initializeClassVariable();
}
The "alternative to static blocks" is about initializing a single static field.
Example:
class A {
static Map<String, Integer> romans;
static {
romans = new HashMap<>();
romans.put("I", 1);
romans.put("V", 5);
romans.put("X", 10);
}
}
Alternative:
class A {
static Map<String, Integer> romans = initRomans();
private static Map<String, Integer> initRomans() {
Map<String, Integer> r = new HashMap<>();
r.put("I", 1);
r.put("V", 5);
r.put("X", 10);
return r;
}
}
As the article says, with this code you can reset the static field.
public static void resetRomans() {
romans = initRomans();
}
If your code does something else, then the "alternative" isn't applicable, and you write the code in a static initializer block.
class A {
static {
Manager.register(A.class);
}
}
I have the following Java code:
public class Something {
static {
new Something();
}
public static final int[] EMPTY_INT_ARRAY = new int[0];
}
I'm using FindBugs to look for code errors, but the following error is never raised:
SI: Static initializer creates instance before all static final fields
assigned (SI_INSTANCE_BEFORE_FINALS_ASSIGNED)
The class's static initializer creates an instance of the class before
all of the static final fields are assigned.
Is this the correct case that should demonstrate this issue? why is it not picking up this code issue?
Not tried, but I think an example would be:
public class Something {
static {
new Something();
}
public static final int CONST = 42;
}
As lysergic-acid has found out, this error has a lower severity in FindBugs (16). You have to manually lower the min threshold to make it show.
I have different Exception category Enum as below
public enum GSBBCacheCategory {
SEARCH(9001),
UPDATE_PERSECURITY(9002),
CROSS_REFERENCING_PERSECURITY(9003),
METADATA_SEARCH(9004),
REMOVEALL(9005),
UPDATE_BACKOFFICE(9002);
private int exceptionCode;
GSBBCacheCategory(int exceptionCode)
{
this.exceptionCode = exceptionCode;
}
public int getExceptionCode()
{
return exceptionCode;
}
}
public enum GSBBEncryptionCategory {
.
.
.
}
I want to provide one place to access these Enum in client code. Presently I achieved this as below
public class GSBBExceptionCodes
{
public static GSBBDecryptionCategory decryptionCategory;
public static GSBBCacheCategory cacheCategory;
}
Now to access exception code I have do something like below
public static void main(String[] args) {
System.out.println(GSBBExceptionCodes.decryptionCategory.ERRORCODE_DECRYPTION_FAILURE);
System.out.println(GSBBExceptionCodes.cacheCategory.UPDATE_PERSECURITY);
}
Which says “The static field GSBBDecryptionCategory.ERRORCODE_DECRYPTION_FAILURE should be accessed in a static way”
Is it possible to achieve above without any warning?
There are two ways to reference a static member (either a field or a method). One is WhateverClass.theField, and the other is someInstance.theField where someInstance has a compile-time type of WhateverClass. The former is much clearer, and so your IDE is helpfully telling you to use it instead of the latter.
The reason it's better is that referencing a static member by an instance makes it look like the method has something to do with that instance, which it doesn't. Here's a real-life example:
Thread myThread = getMyThread();
myThread.start();
myThread.sleep(5000);
At first blush, it looks like you're asking myThread to sleep for 5 seconds (5000 milliseconds), but that's not what you're doing at all. You're asking the current thread to sleep, because that last line is exactly the same as invoking Thread.sleep(5000). That second example is much more clearly a static method.
Or, here's another example. Let's say your static fields were mutable.
public class Foo {
public static int value = 1;
}
(This public static mutable field is a bad idea for other reasons, but simplifies the example). Now let's say you do:
Foo one = new Foo();
Foo two = new Foo();
one.value = 2;
two.value = 3;
System.out.println(one.value);
System.out.println(two.value);
Kinda looks like that should print "2" and then "3", right? But no -- it'll print "3", "3" because both assignments to .value are in fact to the same, static field. It's just an optical illusion that the one or two instances have anything to do with anything.
Imho, the ability to reference static members from instances is a misfeature. But it's there, so you should avoid it. Which is what the compiler is trying to suggest you do.
Try this :
public static void main(String[] args) {
System.out.println(GSBBDecryptionCategory.ERRORCODE_DECRYPTION_FAILURE);
System.out.println(GSBBCacheCategory.UPDATE_PERSECURITY);
}
You are now accessing the field in a static way which should remove the warning.
It sounds like instead of having these as public static fields, they should be inner classes:
public class GSBBExceptionCodes {
public enum GSBBCacheCategory {
...
}
}
I have a class holding a large a mount of generated constants as such:
public class Constants extends SomeBaseClass {
// init() is defined in some base class...
public static final XXX KEY1 = init(...);
public static final XXX KEY2 = init(...);
public static final XXX KEY3 = init(...);
// ...
public static final XXX KEY2000 = init(...);
}
When the number of generated constants is very high, this results in a static initialiser that is larger than the upper limit for Java method sizes (i.e. > 64kb), resulting in a compiler error. One solution is to create several "block initialisation methods" for blocks that can be guaranteed to produce less than 64kb of byte-code, such that they fit into a method:
public class Constants extends SomeBaseClass {
public static XXX KEY1;
public static XXX KEY2;
public static XXX KEY3;
// ...
public static XXX KEY2000;
static {
initialise0001To1000();
initialise1001To2000();
}
private static void initialise0001To1000() {
KEY1 = init(...);
KEY2 = init(...);
KEY3 = init(...);
// ...
}
private static void initialise1001To2000() {
// ...
KEY2000 = init(...);
}
}
The drawback of this is that I can no longer declare the constants as final, because they are now no longer initialised directly in the static initialiser.
My question is, how can I circumvent that compiler / JVM limitation in a way that I can still generate static final constants?
One option would be to use inheritance - have a series of classes Constants1, Constants2, ..., ConstantsN that all define the constants, then have each one inherit from the previous one. The final class Constants can then directly inherit from the last of them. This also lets you mark everything final.
Out of curiosity, how did you end up with a file so large that you couldn't fit the initialization code into the 64KB limit?
Hope this helps!
I finally went for a solution involving nested classes. This was suggested in a comment to this answer here by user Loadmaster. Nested classes have two advantages:
They allow for hiding these workaround implementation details from the outside world by being private nested classes
They allow for keeping constants final
But they also have a disadvantage compared to templatetypedef's solution:
I will run into the same problem again with much larger numbers of constants
Right now, however, this seems to be the most suitable solution:
public class Constants {
public static XXX KEY1 = Constants1.KEY1;
public static XXX KEY2 = Constants1.KEY2;
public static XXX KEY3 = Constants1.KEY3;
// ...
public static XXX KEY2000 = Constants2.KEY2000;
// Nested class holding 1000 constants
private static class Constants1 extends SomeBaseClass {
KEY1 = init(...);
KEY2 = init(...);
KEY3 = init(...);
// ...
}
// Nested class holding the next 1000 constants
private static class Constants2 extends SomeBaseClass {
// ...
KEY2000 = init(...);
}
// Keep generating nested classes for more constants...
private static class Constants3 ... {}
}
Wouldn't this work? NO. See comments for why this kind of answer will not solve the problem.
It would allow you to keep the static variables final and it would be easier to autogenerate. However java collapses all the static init blocks into one giant static init block and thus the problem is not solved.
public class Constants extends SomeBaseClass {
// init() is defined in some base class...
public static final XXX KEY1 ;
static
{
KEY1 = init(...);
}
public static final XXX KEY2 ;
static
{
KEY2 = init(...);
}
public static final XXX KEY3 ;
static
{
KEY3 = init(...);
}
// ...
}
You can have as many static initialization blocks as you want.
As far as I understood the "static initialization block" is used to set values of static field if it cannot be done in one line.
But I do not understand why we need a special block for that. For example we declare a field as static (without a value assignment). And then write several lines of the code which generate and assign a value to the above declared static field.
Why do we need this lines in a special block like: static {...}?
The non-static block:
{
// Do Something...
}
Gets called every time an instance of the class is constructed. The static block only gets called once, when the class itself is initialized, no matter how many objects of that type you create.
Example:
public class Test {
static{
System.out.println("Static");
}
{
System.out.println("Non-static block");
}
public static void main(String[] args) {
Test t = new Test();
Test t2 = new Test();
}
}
This prints:
Static
Non-static block
Non-static block
If they weren't in a static initialization block, where would they be? How would you declare a variable which was only meant to be local for the purposes of initialization, and distinguish it from a field? For example, how would you want to write:
public class Foo {
private static final int widgets;
static {
int first = Widgets.getFirstCount();
int second = Widgets.getSecondCount();
// Imagine more complex logic here which really used first/second
widgets = first + second;
}
}
If first and second weren't in a block, they'd look like fields. If they were in a block without static in front of it, that would count as an instance initialization block instead of a static initialization block, so it would be executed once per constructed instance rather than once in total.
Now in this particular case, you could use a static method instead:
public class Foo {
private static final int widgets = getWidgets();
static int getWidgets() {
int first = Widgets.getFirstCount();
int second = Widgets.getSecondCount();
// Imagine more complex logic here which really used first/second
return first + second;
}
}
... but that doesn't work when there are multiple variables you wish to assign within the same block, or none (e.g. if you just want to log something - or maybe initialize a native library).
Here's an example:
private static final HashMap<String, String> MAP = new HashMap<String, String>();
static {
MAP.put("banana", "honey");
MAP.put("peanut butter", "jelly");
MAP.put("rice", "beans");
}
The code in the "static" section(s) will be executed at class load time, before any instances of the class are constructed (and before any static methods are called from elsewhere). That way you can make sure that the class resources are all ready to use.
It's also possible to have non-static initializer blocks. Those act like extensions to the set of constructor methods defined for the class. They look just like static initializer blocks, except the keyword "static" is left off.
It's also useful when you actually don't want to assign the value to anything, such as loading some class only once during runtime.
E.g.
static {
try {
Class.forName("com.example.jdbc.Driver");
} catch (ClassNotFoundException e) {
throw new ExceptionInInitializerError("Cannot load JDBC driver.", e);
}
}
Hey, there's another benefit, you can use it to handle exceptions. Imagine that getStuff() here throws an Exception which really belongs in a catch block:
private static Object stuff = getStuff(); // Won't compile: unhandled exception.
then a static initializer is useful here. You can handle the exception there.
Another example is to do stuff afterwards which can't be done during assigning:
private static Properties config = new Properties();
static {
try {
config.load(Thread.currentThread().getClassLoader().getResourceAsStream("config.properties");
} catch (IOException e) {
throw new ExceptionInInitializerError("Cannot load properties file.", e);
}
}
To come back to the JDBC driver example, any decent JDBC driver itself also makes use of the static initializer to register itself in the DriverManager. Also see this and this answer.
I would say static block is just syntactic sugar. There is nothing you could do with static block and not with anything else.
To re-use some examples posted here.
This piece of code could be re-written without using static initialiser.
Method #1: With static
private static final HashMap<String, String> MAP;
static {
MAP.put("banana", "honey");
MAP.put("peanut butter", "jelly");
MAP.put("rice", "beans");
}
Method #2: Without static
private static final HashMap<String, String> MAP = getMap();
private static HashMap<String, String> getMap()
{
HashMap<String, String> ret = new HashMap<>();
ret.put("banana", "honey");
ret.put("peanut butter", "jelly");
ret.put("rice", "beans");
return ret;
}
There are a few actual reasons that it is required to exist:
initializing static final members whose initialization might throw an exception
initializing static final members with calculated values
People tend to use static {} blocks as a convenient way to initialize things that the class depends on within the runtime as well - such as ensuring that particular class is loaded (e.g., JDBC drivers). That can be done in other ways; however, the two things that I mention above can only be done with a construct like the static {} block.
You can execute bits of code once for a class before an object is constructed in the static blocks.
E.g.
class A {
static int var1 = 6;
static int var2 = 9;
static int var3;
static long var4;
static Date date1;
static Date date2;
static {
date1 = new Date();
for(int cnt = 0; cnt < var2; cnt++){
var3 += var1;
}
System.out.println("End first static init: " + new Date());
}
}
It is a common misconception to think that a static block has only access to static fields. For this I would like to show below piece of code that I quite often use in real-life projects (copied partially from another answer in a slightly different context):
public enum Language {
ENGLISH("eng", "en", "en_GB", "en_US"),
GERMAN("de", "ge"),
CROATIAN("hr", "cro"),
RUSSIAN("ru"),
BELGIAN("be",";-)");
static final private Map<String,Language> ALIAS_MAP = new HashMap<String,Language>();
static {
for (Language l:Language.values()) {
// ignoring the case by normalizing to uppercase
ALIAS_MAP.put(l.name().toUpperCase(),l);
for (String alias:l.aliases) ALIAS_MAP.put(alias.toUpperCase(),l);
}
}
static public boolean has(String value) {
// ignoring the case by normalizing to uppercase
return ALIAS_MAP.containsKey(value.toUpper());
}
static public Language fromString(String value) {
if (value == null) throw new NullPointerException("alias null");
Language l = ALIAS_MAP.get(value);
if (l == null) throw new IllegalArgumentException("Not an alias: "+value);
return l;
}
private List<String> aliases;
private Language(String... aliases) {
this.aliases = Arrays.asList(aliases);
}
}
Here the initializer is used to maintain an index (ALIAS_MAP), to map a set of aliases back to the original enum type. It is intended as an extension to the built-in valueOf method provided by the Enum itself.
As you can see, the static initializer accesses even the private field aliases. It is important to understand that the static block already has access to the Enum value instances (e.g. ENGLISH). This is because the order of initialization and execution in the case of Enum types, just as if the static private fields have been initialized with instances before the static blocks have been called:
The Enum constants which are implicit static fields. This requires the Enum constructor and instance blocks, and instance initialization to occur first as well.
static block and initialization of static fields in the order of occurrence.
This out-of-order initialization (constructor before static block) is important to note. It also happens when we initialize static fields with the instances similarly to a Singleton (simplifications made):
public class Foo {
static { System.out.println("Static Block 1"); }
public static final Foo FOO = new Foo();
static { System.out.println("Static Block 2"); }
public Foo() { System.out.println("Constructor"); }
static public void main(String p[]) {
System.out.println("In Main");
new Foo();
}
}
What we see is the following output:
Static Block 1
Constructor
Static Block 2
In Main
Constructor
Clear is that the static initialization actually can happen before the constructor, and even after:
Simply accessing Foo in the main method, causes the class to be loaded and the static initialization to start. But as part of the Static initialization we again call the constructors for the static fields, after which it resumes static initialization, and completes the constructor called from within the main method. Rather complex situation for which I hope that in normal coding we would not have to deal with.
For more info on this see the book "Effective Java".
So you have a static field (it's also called "class variable" because it belongs to the class rather than to an instance of the class; in other words it's associated with the class rather than with any object) and you want to initialize it. So if you do NOT want to create an instance of this class and you want to manipulate this static field, you can do it in three ways:
1- Just initialize it when you declare the variable:
static int x = 3;
2- Have a static initializing block:
static int x;
static {
x=3;
}
3- Have a class method (static method) that accesses the class variable and initializes it:
this is the alternative to the above static block; you can write a private static method:
public static int x=initializeX();
private static int initializeX(){
return 3;
}
Now why would you use static initializing block instead of static methods?
It's really up to what you need in your program. But you have to know that static initializing block is called once and the only advantage of the class method is that they can be reused later if you need to reinitialize the class variable.
let's say you have a complex array in your program. You initialize it (using for loop for example) and then the values in this array will change throughout the program but then at some point you want to reinitialize it (go back to the initial value). In this case you can call the private static method. In case you do not need in your program to reinitialize the values, you can just use the static block and no need for a static method since you're not gonna use it later in the program.
Note: the static blocks are called in the order they appear in the code.
Example 1:
class A{
public static int a =f();
// this is a static method
private static int f(){
return 3;
}
// this is a static block
static {
a=5;
}
public static void main(String args[]) {
// As I mentioned, you do not need to create an instance of the class to use the class variable
System.out.print(A.a); // this will print 5
}
}
Example 2:
class A{
static {
a=5;
}
public static int a =f();
private static int f(){
return 3;
}
public static void main(String args[]) {
System.out.print(A.a); // this will print 3
}
}
If your static variables need to be set at runtime then a static {...} block is very helpful.
For example, if you need to set the static member to a value which is stored in a config file or database.
Also useful when you want to add values to a static Map member as you can't add these values in the initial member declaration.
It is important to understand that classes are instantiated from java.class.Class during runtime. That is when static blocks are executed, which allows you to execute code without instantiating a class:
public class Main {
private static int myInt;
static {
myInt = 1;
System.out.println("myInt is 1");
}
// needed only to run this class
public static void main(String[] args) {
}
}
The result is myInt is 1 printed to the console.
As supplementary, like #Pointy said
The code in the "static" section(s) will be executed at class load
time, before any instances of the class are constructed (and before
any static methods are called from elsewhere).
It's supposed to add System.loadLibrary("I_am_native_library") into static block.
static{
System.loadLibrary("I_am_a_library");
}
It will guarantee no native method be called before the related library is loaded into memory.
According to loadLibrary from oracle:
If this method is called more than once with the same library name,
the second and subsequent calls are ignored.
So quite unexpectedly, putting System.loadLibrary is not used to avoid library be loaded multi-times.
static block is used for any technology to initialize static data member in dynamic way,or we can say for the dynamic initialization of static data member static block is being used..Because for non static data member initialization we have constructor but we do not have any place where we can dynamically initialize static data member
Eg:-class Solution{
// static int x=10;
static int x;
static{
try{
x=System.out.println();
}
catch(Exception e){}
}
}
class Solution1{
public static void main(String a[]){
System.out.println(Solution.x);
}
}
Now my static int x will initialize dynamically ..Bcoz when compiler will go to Solution.x it will load Solution Class and static block load at class loading time..So we can able to dynamically initialize that static data member..
}
static int B,H;
static boolean flag = true;
static{
Scanner scan = new Scanner(System.in);
B = scan.nextInt();
scan.nextLine();
H = scan.nextInt();
if(B < 0 || H < 0){
flag = false;
System.out.println("java.lang.Exception: Breadth and height must be positive");
}
}