-
Notifications
You must be signed in to change notification settings - Fork 38.2k
Code Style
This document defines the coding standards for source files in the Spring Framework. It is primarily intended for the Spring Framework team but can be used as a reference by contributors.
The structure of this document is based on the Google Java Style reference and is a work in progress.
Source files must be encoded using UTF-8
.
- Indentation uses tabs (not spaces)
- Unix (LF), not DOS (CRLF) line endings
- Eliminate all trailing whitespace
- On Linux, Mac, etc.:
find . -type f -name "*.java" -exec perl -p -i -e "s/[ \t]$//g" {} \;
- On Linux, Mac, etc.:
A source file consists of the following, in this exact order:
- License
- Package statement
- Import statements
- Exactly one top-level class
Exactly one blank line separates each of the above sections.
Each source file must specify the following license at the very top of the file:
/*
* Copyright 2002-2023 the original author or authors.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* https://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
Always check the date range in the license header. For example, if you've modified a file in 2022 whose header still reads:
* Copyright 2002-2018 the original author or authors.
Then be sure to update it to 2023 accordingly:
* Copyright 2002-2023 the original author or authors.
The import statements are structured as follow:
- import
java.*
- blank line
- import
javax.*
- import
jakarta.*
- blank line
- import all other imports
- blank line
- import
org.springframework.*
- blank line
- import static all other imports
Static imports should not be used in production code, but they should be used in test code, especially for things like import static org.assertj.core.api.Assertions.assertThat;
.
Although static imports are generally forbidden in production code, the following are use cases for which static imports are permissible.
- constants (including enum constants): such as those in
java.nio.charset.StandardCharsets
ororg.springframework.core.annotation.MergedAnnotations.SearchStrategy
- static factory methods for third-party DSLs: such as the methods in
org.junit.platform.engine.discovery.DiscoverySelectors
when used with the JUnit PlatformLauncher
API
Wildcard imports such as import java.util.*
or import static org.assertj.core.api.Assertions.*
are forbidden, even in test code.
The following governs how the elements of a source file are organized:
- static fields
- normal fields
- constructors
- (private) methods called from constructors
- static factory methods
- JavaBean properties (i.e., getters and setters)
- method implementations coming from interfaces
- private or protected templates that get called from method implementations coming from interfaces
- other methods
-
equals
,hashCode
, andtoString
Note that private or protected methods called from method implementations should be placed immediately below the methods where they're used. In other words if there 3 interface method implementations with 3 private methods (one used from each), then the order of methods should include 1 interface and 1 private method in sequence, not 3 interface and then 3 private methods at the bottom.
Above all, the organization of the code should feel natural.
Braces mostly follow the Kernighan and Ritchie style (a.k.a., "Egyptian brackets") for nonempty blocks and block-like constructs:
- No line break before the opening brace but prefixed by a single space
- Line break after the opening brace
- Line break before the closing brace
- Line break after the closing brace if that brace terminates a statement or the body of a method, constructor, or named class
- Line break before else, catch and finally statements
Example:
return new MyClass() {
@Override
public void method() {
if (condition()) {
something();
}
else {
try {
alternative();
}
catch (ProblemException ex) {
recover();
}
}
}
};
90 characters is the preferred line length we aim for. In some cases the preferred length can be achieved by refactoring code slightly. In other cases it's just not possible.
90 is not a hard limit. Lines between 90-105 are perfectly acceptable in many cases where it aids readability and where wrapping has the opposite effect of reducing readability. This is a judgement call and it's also important to seek consistency. Many times you can learn by looking how specific situations are handled in other parts of the code.
Lines between 105-120 are allowed but discouraged and should be few.
No lines should exceed 120 characters.
The one big exception to the above line wrapping rules is Javadoc where we aim to wrap around 80 characters for maximum readability in all kinds of contexts -- for example, reading on Github, on your phone, etc.
When wrapping a lengthy expression, 90 characters is the length at which we aim to wrap. Place separator symbols at the end of the current line rather than at the beginning of the next line. In this context, the following are considered separator symbols: ,
, +
, ?
, :
, &&
, ||
. For example:
if (thisLengthyMethodCall(param1, param2) && anotherCheck() &&
yetAnotherCheck()) {
// ....
}
Add two blank lines before the following elements:
-
static {}
block - Fields
- Constructors
- Inner classes
Add one blank line after a method signature that is multiline, i.e.
@Override
protected Object invoke(FooBarOperationContext context,
AnotherSuperLongName name) {
// code here
}
For inner-classes, extra blank lines around fields and constructors are typically not added as the inner class is already separated by 2 lines, unless the inner class is more substantial in which case the 2 extra lines could still help with readability.
Try as much as possible to put the implements
, extends
section of a class declaration on the same line as the class itself.
Order the classes so that the most important comes first.
Constant names use CONSTANT_CASE
: all uppercase letters, with words separated by underscores.
Every constant is a static final
field, but not all static final
fields are constants. Constant case should therefore be chosen only if the field is really a constant.
Example:
// Constants
private static final Object NULL_HOLDER = new NullHolder();
public static final int DEFAULT_PORT = -1;
// Not constants
private static final ThreadLocal<Executor> executorHolder = new ThreadLocal<Executor>();
private static final Set<String> internalAnnotationAttributes = new HashSet<String>();
Avoid using single characters as variable names. For instance prefer Method method
to Method m
.
- A file should look like it was crafted by a single author, not like a history of changes.
- Don't artificially spread things out that belong together.
Choose wisely where to add a new setter method; it should not be simply added at the end of the list. Perhaps the setter is related to another setter or relates to a group. In that case it should be placed near related methods.
- Setter order should reflect order of importance, not historical order.
- Ordering of fields and setters should be consistent.
Wrap the ternary operator within parentheses -- for example, return (foo != null ? foo : "default");
.
Also make sure that the not null condition comes first.
Use the org.springframework.util.Assert.notNull
static method to check that a method argument is not null
and throw an IllegalArgumentException
otherwise. Format the exception message so that the name of the parameter comes first with its first character capitalized, followed by "must not be null". For instance:
public void handle(Event event) {
Assert.notNull(event, "Event must not be null");
//...
}
For other use cases, use the org.springframework.util.Assert.state
static method to ensure a variable is not null
and throw an IllegalStateException
otherwise. Format the exception message so that the identifier related to the variable (name, type or description) comes first with its first character capitalized, followed by "must not be null". For instance:
//...
Event event = ...
Assert.state(event != null, "Event must not be null");
//...
The null-safety of Spring Framework APIs and fields must be specified using related annotations, and consistency is enforced at build-time via NullAway.
Packages are expected to define non-null as the default in package-info.java
:
@NonNullApi
@NonNullFields
package org.springframework.mypackage;
And nullable fields, return values and parameters are expected to be specified explicitly using @Nullable
:
import org.springframework.lang.Nullable;
public class MyClass {
@Nullable
private Class<?> nullableType;
public void setNullableType(@Nullable Class<?> nullableType) {
this.nullableType = nullableType;
}
@Nullable
public Class<?> getNullableType() {
return this.nullableType;
}
}
org.springframework.lang.Contract
annotation can be used to specify some aspects of the method behavior depending on the arguments that are then taken in account by NullAway. See for example how @Contract
is used in Spring Framework codebase in the Assert
class:
public abstract class Assert {
// ...
@Contract("false, _ -> fail")
public static void state(boolean expression, String message) {
if (!expression) {
throw new IllegalStateException(message);
}
}
// ...
@Contract("null, _ -> fail")
public static void notNull(@Nullable Object object, String message) {
if (object == null) {
throw new IllegalArgumentException(message);
}
}
// ...
}
Those contract allows NullAway static analysis to understand that after an invocation of Assert.state(value != null, "Value must not be null")
or Assert.notNull(value, "Value must not be null")
, a nullable value
can be considered as non nullable.
Related guidelines:
- When overriding a method, null-safety annotations of the super method need to be specified on the overridden method unless you want to override null-safety.
- Use
@SuppressWarnings("NullAway")
when NullAway triggers irrelevant errors, which can happen due to NullAway bugs or missing features or when the analysis is not able to prove null-safety (can happen with lambdas).
Always add @Override
on methods overriding or implementing a method declared in a super type.
-
@since
should be added to every new class with the version of the framework in which it was introduced. -
@since
should be added to any new public and protected methods of an existing class.
A class that is only a collection of static utility methods must be named with a Utils
suffix, must have a private
default constructor, and must be abstract
. Making the class abstract
and providing a private
default constructor prevent anyone from instantiating it. For example:
public abstract MyUtils {
private MyUtils() {
/* prevent instantiation */
}
// static utility methods
}
A field of a class should always be referenced using this
. A method of a class, however, should never be referenced using this
.
The use of var
for variable declarations (local variable type inference) is not permitted. Instead, declare variables using the concrete type or interface (where applicable).
The following template summarizes typical Javadoc usage for a method.
/**
* Parse the specified {@link Element} and register the resulting
* {@link BeanDefinition BeanDefinition(s)}.
* <p>Implementations must return the primary {@link BeanDefinition} that results
* from the parsing if they will ever be used in a nested fashion (for example as
* an inner tag in a {@code <property/>} tag). Implementations may return
* {@code null} if they will <strong>not</strong> be used in a nested fashion.
* @param element the element that is to be parsed into one or more {@link BeanDefinition BeanDefinitions}
* @param parserContext the object encapsulating the current state of the parsing process;
* provides access to a {@link org.springframework.beans.factory.support.BeanDefinitionRegistry}
* @return the primary {@link BeanDefinition}
*/
BeanDefinition parse(Element element, ParserContext parserContext);
In particular, please note:
- Use an imperative style (i.e. Return and not Returns) for the first sentence.
- No blank lines between the description and the parameter descriptions.
- If the description is defined with multiple paragraphs, start each of them with
<p>
. - If a parameter description needs to be wrapped, do not indent subsequent lines (see
parserContext
).
The Javadoc of a class has some extra rules that are illustrated by the sample below:
/*
* Interface used by the {@link DefaultBeanDefinitionDocumentReader} to handle custom,
* top-level (directly under {@code <beans/>}) tags.
*
* <p>Implementations are free to turn the metadata in the custom tag into as many
* {@link BeanDefinition BeanDefinitions} as required.
*
* <p>The parser locates a {@link BeanDefinitionParser} from the associated
* {@link NamespaceHandler} for the namespace in which the custom tag resides.
*
* @author Rob Harrop
* @since 2.0
* @see NamespaceHandler
* @see AbstractBeanDefinitionParser
*/
- Each class must have a
@since
tag with the version in which the class was introduced. - The order of tags for type-level Javadoc is:
@author
,@since
,@param
,@see
,@deprecated
. - The order of tags for constructor-level, method-level, and field-level Javadoc is:
@param
,@return
,@throws
,@since
,@see
,@deprecated
. - In contrast to constructor-level, method-level, and field-level Javadoc, the paragraphs of a class description are separated by blank lines.
The following are additional general rules to apply when writing Javadoc:
- Use
{@code}
to wrap code statements or values such asnull
. - If a type is only referenced by a
{@link}
element, use the fully qualified name in order to avoid an unnecessaryimport
declaration.
Tests must be written using JUnit Jupiter (a.k.a., JUnit 5).
The only exceptions to the above rule are test classes in the spring-test
module that specifically test Spring's integration with JUnit 4 and TestNG.
Each test class name must end with a Tests
suffix.
Use AssertJ for assertions.
Use the BDD Mockito support.