timofeevda / vertx-codegen

Vert.x code generator for asynchronous polyglot APIs

Vert.x API Generation

Render documentation

> mvn clean package -Pdocs
> open target/docs/vertx-codegen/java/index.html

Helper projects

• Codegen CLI: a codegen CLI to help code generating files.
• Codegen starter: a codegen Starter that can be forked to create a new Vert.x code generator

API generator

A code generator consist of an MVEL template declared in a codegen.json descriptor:

{
  "name": "Groovy",
  "generators": [ {
    "kind": "class",
    "filename": "'groovy/' + fqn.replace('io.vertx', 'io.vertx.groovy').replace('.', '/') + '.groovy'",
    "templateFilename": "vertx-groovy/template/groovy.templ"
  } ]
}

• filename is an MVEL expression for the file name, returning null skips the generation.
• templateFilename is the name of the MVEL template to apply
• incremental true when the template performs incremental processing, false or absent otherwise
• kind: can be a String or an Array of Strings each representing a generator type. There are several kinds of generators for different use cases:
  • class : applied on each API classes
  • package : applied on each Java package
  • module : applied on each declared module, a module uniquely identifies an API
  • dataObject: applied on each data object class
  • proxy: applied on each proxy class
  • enum: applied on each enum class annotated with @VertxGen

There can be as many generators as you like.

Generated output

A generator can create 3 different kinds of output: Java classes, resources and anything else

Generated Java classes

A generator declaring a filename that matches a Java FQN followed by .java suffix will have its content generated as a Java class. This class will be automatically compiled by the same compiler (that's a Java compiler feature).

The generated files are handled by the Java compiler (-s option), usually build tools configures the compiler to store them in a specific build location, for instance Maven by default uses the target/generated-sources/annotations directory.

The following generators use it:

• Data object converters
• Service proxy and service handler
• RxJava-ified classes API
• Groovy extension methods API

Generated resources

A generator declaring a filename prefixed by resources/ will have its content generated as a compiler resource. This resource will be stored in the generated sources directory and the generated classes directory.

The generated files are handled by the Java compiler (-s option), usually build tools configures the compiler to store them in a specific build location, for instance Maven by default uses the target/generated-sources/annotations directory.

The following generators use it:

• JavaScript generator
• Ruby generator

Other generated files

Anything else will be stored in the file system using the filename, when the filename is relative (it usually is) the target path will be resolved agains the codegen.output directory.

The following generators use it:

• Ceylon generator
• Scala generator
• Kotlin extension methods

When the codegen.output is not specified, the generated files are discarded.

Relocation

Sometimes you want to have a generator to output its files in another directory, you can do that with the codegen.output.generator-name compiler option:

<codegen.output.data_object_converters>generated</codegen.output.data_object_converters>

The generator will store its content in the codegen.output/generated directory instead as a Java class.

Processor configuration

You can configure the CodeGenProcessor as any Java annotation processor, here is how to do with Maven:

<pluginManagement>
  <plugins>
    <!-- Configure the execution of the compiler to execute the codegen processor -->
    <plugin>
      <artifactId>maven-compiler-plugin</artifactId>
      <version>3.1</version>
      <configuration>
        <source>1.8</source>
        <target>1.8</target>
        <encoding>${project.build.sourceEncoding}</encoding>
        <!-- Important: there are issues with apt and incremental compilation in the maven-compiler-plugin -->
        <useIncrementalCompilation>false</useIncrementalCompilation>
      </configuration>
      <executions>
        <execution>
          <id>default-compile</id>
          <configuration>
            <annotationProcessors>
              <annotationProcessor>io.vertx.codegen.CodeGenProcessor</annotationProcessor>
            </annotationProcessors>
            <compilerArgs>
              <arg>-AoutputDirectory=${project.basedir}/src/main</arg>
            </compilerArgs>
          </configuration>
        </execution>
      </executions>
    </plugin>
  </plugins>
</pluginManagement>

And here is a configuration example for Gradle:

task annotationProcessing(type: JavaCompile, group: 'build') { // codegen
  source = sourceSets.main.java
  classpath = configurations.compile + configurations.compileOnly
  destinationDir = project.file('src/main/generated')
  options.compilerArgs = [
    "-proc:only",
    "-processor", "io.vertx.codegen.CodeGenProcessor",
    "-AoutputDirectory=${project.projectDir}/src/main"
  ]
}

compileJava {
  targetCompatibility = 1.8
  sourceCompatibility = 1.8

  dependsOn annotationProcessing
}

sourceSets {
  main {
    java {
      srcDirs += 'src/main/generated'
    }
  }
}

Besides you can use the processor classified dependency that declares the annotation processor as a META-INF/services/javax.annotation.processing.Processor, if you do so, code generation happens automatically:

<dependency>
  <groupid>io.vertx</groupId>
  <artifactId>vertx-codegen</artifactId>
  <classifier>processor</classifier>
</dependency>

You still need to configure the outputDirectory for generating files non resources/classes as the processors requires this option to know where to place them.

The processor is configured by a few options

• codegen.output : where the non Java classes / non resources are stored
• codegen.output.<generator-name> : relocate the output of to another directory
• codegen.generators : a comma separated list of generators, each expression is a regex, allow to filter undesired generators

API constraints

In order for code generation to work effectively, certain constraints are put on the Java interfaces.

The constraints are

• The API must be described as a set of Java interfaces, classes are not permitted
• Default methods are not permitted
• Nested interfaces are not permitted
• All interfaces to have generation performed on them must be annotated with the io.vertx.codegen.annotations.VertxGen annotation
• Fluent methods (methods which return a reference to this) must be annotated with the io.vertx.codegen.annotations.Fluent annotation
• Data object classes (classes which provide data (e.g. configuration) to methods) must be annotated with the io.vertx.codegen.annotations.DataObject annotation
• Data object classes must provide a constructor which takes a single io.vertx.core.json.JsonObject parameter.
• Methods where the return value must be cached in the API shim must be annotated with the io.vertx.codegen.annotations.CacheReturn annotation
• Only certain types are allowed as parameter or return value types for any API methods (defined below).
• Custom enums should be annotated with @VertxGen, although this is not mandatory to allow the usage of existing Java enums

Permitted types

We define the following set Basic of basic types:

• any primitive type
• any boxed primitive type
• java.lang.String

We define Json as the set of types io.vertx.core.json.JsonObject and io.vertx.core.json.JsonArray

We define DataObject as the set of user defined API types which are defined in its own class and annotated with @DataObject

We define TypeVar as the set of of types variables where the variable is either declared by its generic method or its generic type

We define Api as the set of user defined API types which are defined in its own interface and annotated with @VertxGen

We define JavaType as the set of any Java type that does not belong to Basic, Json, DataObject, TypeVar and Api, e.g java.net.Socket. Methods are not allowed to declare such type by default and can be annotated with @GenIgnore(GenIgnore.PERMITTED_TYPE) to allow them. Such method limit the translation of the method to other languages, so it should be used with care. It is useful to allow method previously annotated with @GenIgnore to be available in code generator like RxJava that can handle Java types.

We define Parameterized as the set of user defined API types which are defined in its own interface and annotated with @VertxGen where the type parameters belong to:

• the type java.lang.Void
• the set Basic
• the set Json
• the set DataObject
• any enum type
• the set Api
• the set TypeVar

The following set Return of types are permitted as return types from any API method:

• void
• the set Basic
• the set Json
• the set DataObject
• any enum type
• java.lang.Throwable
• the set TypeVar
• java.lang.Object
• the set Api
• the set Parameterized
• the set JavaType
• type java.util.List<C> or java.util.Set<C> where C contains
  • the set Basic
  • the set Json
  • any enum type
  • the set Api
  • the set DataObject
  • the set JavaType
  • java.lang.Object
• java.util.Map<String, C> where C contains
  • the set Basic
  • the set Json
  • the set JavaType
  • java.lang.Object

The following set Param of types are permitted as parameters to any API method:

• the set Basic
• the set Json
• the set DataObject
• any enum type
• the type java.lang.Throwable
• the set TypeVar
• java.lang.Object
• the set Api
• the set JavaType
• the set Parameterized
• the type java.lang.Class<T> where <T> is among
  • the set Basic
  • the set Json
  • the set Api
  • the set JavaType
• type java.util.List<C> or java.util.Set<C> where C contains
  • the set Basic
  • the set Json
  • the set DataObject
  • the set Api
  • the set JavaType
  • any enum type
  • java.lang.Object
• type java.util.Map<String, C> where C contains
  • the set Basic
  • the set Json
  • the set Api
  • the set JavaType
  • any enum type
  • java.lang.Object

In addition any `Api method can have as parameter:

• io.vertx.java.core.Handler<io.vertx.java.core.AsyncResult<HA>> where HA contains
  • the set Return where void is interpreted as java.lang.Void minus java.lang.Throwable
• io.vertx.java.core.Handler<H> where H contains
  • the set Return where void is interpreted as java.lang.Void
• java.util.Function<T, R> where T contains Return and R contains Param

Notes:

• Why no support for data object in Map param values ?

Instance Methods

You can declare methods in your interfaces, e.g.

interface MyInterface {

    void doSomething(String foo);

}

Default method works as well

interface MyInterface {

    default String doSomething(String foo) {
      return foo != null ? new StringBuilder(foo).reverse().toString() : null;
    }

}

Static methods

You can declare static methods in your interfaces, e.g.

interface MyInterface {

    static MyInterface newInterface(String foo) {
      return new ....
    }

}

Fields

You can declare fields in your interfaces, e.g.

interface MyInterface {

    int SOME_CONSTANT = 4;

}

Super interfaces

Interfaces can extend other interfaces which also have the @VertxGen annotation.

Concrete/abstract interfaces

Interfaces annotated with @VertxGen can either be concrete or abstract, such information is important for languages not supporting multiple class inheritance like Groovy:

• interfaces annotated with @VertxGen(concrete = false) are meant to be extended by concrete interfaces and can inherit from abstract interfaces only.
• interfaces annotated with @VertxGen or @VertxGen(concrete = true) are implemented directly by Vertx and can inherit at most one other concrete interface and any abstract interface

Ignoring methods

If you do not wish a method to be used for generation you can annotate it with the @GenIgnore annotation.

Modules

Generated types must belong to a module: a java package annotated with @ModuleGen that defines a module. Such file is created in a file package-info.java.

A module must define:

• a name used when generating languages that don't follow Java package naming, like JavaScript or Ruby.
• a groupPackage to define the package of the group used for generating the generated package names (for Groovy, RxJava or Ceylon generation):

@ModuleGen(name = "acme", groupPackage="com.acme")
package com.acme.myservice;

The group package must be a prefix of the annotated module, it defines the naming of the generate packages o for the modules that belongs to the same group, in this case:

• com.acme.groovy... for Groovy API
• com.acme.rxjava... for RxJava API

For this particular com.acme.myservice module we have:

• com.acme.groovy.myservice for Groovy API
• com.acme.rxjava.myservice for RxJava API

Vert.x Apis uses the io.vertx group package and vertx-XYZ name, this naming is exclusively reserved to Vert.x Apis.

NOTE: using Maven coordinates for name and group package is encouraged: the name corresponding to the Maven artifactId and the group package corresponding to the groupId.

Data objects

A data object is a plain Java public class annotated with @DataObject that follows these minimum requirements:

• A constructor with io.vertx.core.json.JsonObject parameter type

Optionally a data object can define a public io.vertx.core.json.JsonObject toJson() method: such method makes the data object convertible to JsonObject, the data object can then be used as an Api return type.

By default, a data object is responsible to decode from Json (via the JsonObject constructor) and encode to Json (via the toJson method).

Data object converter can be generated with @DataObject(generateConverter=true) by Vert.x Core. Such Data object conversion recognize the following types as member of any @DataObject:

• the set Basic
• these specific types
  • io.vertx.core.Buffer
  • java.time.Instant
• the set Json
• any data object class annotated with @DataObject
• type java.util.List<C> where C contains
  • the specific io.vertx.core.Buffer type
  • the set Basic
  • the set Json
  • any @DataObject
  • the Object type : the List<Object> acts like a JsonArray
• type java.util.Map<String, C> where C contains
  • the specific io.vertx.core.Buffer type
  • the set Basic
  • the set Json
  • any @DataObject
  • the Object type : the Map<String, Object> acts like a JsonMap

This is also used for data object cheatsheet generation.

Enums

Enum types can be freely used in an API, custom enum types should be annotated with @VertxGen to allow processing of the enum. This is not mandatory to allow the reuse the existing Java enums.

Enums can be processed for providing more idiomatic APIs in some languages.

Templates

We use MVEL templating to generate APIs.

There should be a single MVEL template for each language API that is to be generated.

The template will be called once for each interface that is annotated with @VertxGen in the Java API. One output file (e.g. one .js file) will be created for each Java interface.

Template variables for class models

The following variables are made available to templates:

• ifaceSimpleName - the simple name of the Java interface
• ifaceFQCN - the fully qualified class name of the Java interface
• ifacePackageName - the name of the Java package the Java interface belongs to
• ifaceComment - the class comment from the Java interface
• concrete - true when the interface is implemented by vert.x useful to decide the generation of a class or interface in the API shim
• helper - a helper class that of type io.vertx.codegen.Helper which contains useful methods for things such as converting CamelCase to underscores.
• methods - a list of MethodInfo objects describing each method in the interface.
• referencedTypes - a list of strings representing the set of user defined types (also annotated with VertxGen) which are referenced from the current interface
• superTypes - a list of TypeInfo representing the set of user defined types which the current interface extends from
• concreteSuperType - the concrete super type or null
• abstractSuperTypes - subset of superTypes which are abstract
• methodMap - this is a Map<String, MethodInfo> - which allows you to look up all methods with a given name
• importedTypes- this is a Set<TypeInfo> containing the types used by this class
• referencedDataObjectTypes- this is a Set<TypeInfo> containing the DataObject types used by this class
• referencedEnumTypes- this is a Set<TypeInfo> containing the Enum types used by this class
• constants- this is a List<ConstantInfo> containing the constants declared by this class

The TypeInfo represents a Java type:

• name. Generates a string of a form suitable for representing this type in source code using qualified names, for instance io.vertx.core.Handler<io.vertx.core.buffer.Buffer>
• simpleName. Generates a string of a form suitable for representing this type in source code using simple names, for instance Handler<Buffer>
• toString. Same as name
• collectImports(Collection<TypeInfo.Class> imports). Collect all imports required by this type

The TypeInfo.Class is a subclass of TypeInfo representing a Java class:

• kind. An enum providing more information about the type
  • STRING, BOXED_PRIMITIVE, PRIMITIVE: basic types
  • JSON_OBJECT, JSON_ARRAY: io.vertx.core.json.JsonObject and io.vertx.core.json.JsonArray
  • THROWABLE: java.lang.Throwable
  • VOID: java.lang.Void
  • OBJECT: java.lang.Object
  • LIST, SET: corresponding java collections
  • API: a type annotated with @VertxGen
  • DATA_OBJECT: a type annotations with @DataObject
  • HANDLER: io.vertx.core.Handler
  • ASYNC_RESULT: io.vertx.core.AsyncResult
  • ENUM: An enum
  • OTHER: anything else

The MethodInfo object has the following fields:

• name. The name of the method
• kind. The method kind
  • HANDLER: last parameter type is io.vertx.core.Handler<T> and a void or fluent return
  • FUTURE: last parameter type is io.vertx.core.Handler<io.vertx.core.AsyncResult<T>> and a void or fluent return
  • OTHER: anything else, i.e a regular method
• returnType. The fully qualified return type (or void) of the method
• fluent. true if the method is fluent (i.e. returns a reference to the interface itself for chaining calls)
• cacheReturn. true if the generated API method should cache return value
• comment. Method comment.
• params. List of ParamInfo objects representing the parameters of the method.
• staticMethod. true if it's a static method.
• typeParams. The list of the type parameters declared by the method
• containingANyJavaType. true when the method declares a type that is any java type

The ParamInfo object has the following fields:

• name. The name of the parameter
• type. The type of the parameter as a TypeInfo
• dataObject. true If the parameter is a data object type.

The ConstantInfo object has the following fields:

• doc. The doc of the field
• name. The name of the field
• type. The type of the field as a TypeInfo

Template variables in data object models

• type - the TypeInfo of the current option
• doc - the Doc object
• concrete" - true when the option is implemented by vert.x - useful to decide the generation of a class or interface in the API shim
• generateConverter - whether a converter should be generated or not
• inheritConverter - whether the generated converter should convert all properties
• properties - a Set<PropertyInfo> of the available properties in this option
• importedTypes - the imported types by the option
• superTypes - a list of TypeInfo representing the set of user defined types which the current interface extends from
• superType - the supertype of this option ???
• jsonifiable - True if the object has a toJson() method
• hasEmptyConstructor - True if the object has an empty constructor

The PropertyInfo object has the following fields:

• name. The name of the property
• type. The type of the property as TypeInfo
• doc - the Doc object
• declared. True if the property is declared by its data object and does not override the same property from an ancestor
• setterMethod. The name of the setter method in Java
• adderMethod. The name of the adder method in Java
• getterMethod. The name of the getter method in Java
• array. True if the property is an array
• adder. True if the property is an adder (addSomething)
• jsonifiable. True if the object can be converted somehow to json

Incremental templating

Incremental templating allows the same template to process several models and create a single result. This is useful when several sources files needs to generate a same file and the output is the result of the models. To achieve incremental processing, a generator must declares "incremental": true in its descriptor.

During the processing phase, the codegen processors collects all the files generated by incremental templates and groups them by file name. Obviously, the filename expression of the generator needs to return an appropriate string.

At the end of the processing phase, templates are invoked for each model, pretty much like the normal templating but with the following differences:

• the variable incrementalIndex gives the sequence number of the current model, starting at 0
• the variable incrementalSize gives the total number of models processed by the template
• the variable session is a map provided that allows the template to maintain state
• the generated content are appended instead of overwritten

For instance the template:

@if{incrementalIndex==0}
<html>\n
  <body>\n
    <ul>\n
@end{}
      <li>@{type.name}</li>\n
@if{incrementalIndex==incrementalSize-1}
    </ul>\n
  </body>\n
</html>\n
@end{}

With codegen.json:

{
  "name": "index",
  "generators": [ {
    "kind": "class",
    "incremental": true,
    "filename": "'index.html'",
    "templateFilename": "html-index.templ"
  } ]
}

Generates an HTML page with the name of all the API classes.

Skipping generation

Sometimes a template can skip the generation of the file. Setting the value of skipFile to true will do that:

@code{skipFile=true}