Skip to content

Documentables Model

Documentables represent data that is parsed from sources. Think of this data model as of something that could be seen or produced by a compiler frontend, it's not far off from the truth.

By default, documentables are parsed from Descriptor (for Kotlin) and Psi (for Java) models. Code-wise, you can have a look at following classes:

  • DefaultDescriptorToDocumentableTranslator - responsible for Kotlin -> Documentable mapping
  • DefaultPsiToDocumentableTranslator - responsible for Java -> Documentable mapping

Upon creation, it's a collection of trees, each with DModule as root.

Take some arbitrary Kotlin source code that is located within the same module:

// Package 1
class Clazz(val property: String) {
    fun function(parameter: String) {}
}

fun topLevelFunction() {}

// Package 2
enum class Enum { }

val topLevelProperty: String

This would be represented roughly as the following Documentable tree:

flowchart TD
    DModule --> firstPackage[DPackage]
    firstPackage --> DClass
    firstPackage --> toplevelfunction[DFunction] 
    DClass --> DProperty
    DClass --> DFunction
    DFunction --> DParameter
    DModule --> secondPackage[DPackage]
    secondPackage --> DEnum
    secondPackage --> secondPackageProperty[DProperty]

At later stages of transformation, all trees are folded into one (by DocumentableMerger).

Documentable

The main building block of documentables model is Documentable class. It's the base class for all more specific types that represent elements of parsed sources with mostly self-explanatory names (DFunction, DPackage, DProperty, etc) . DClasslike is the base class for class-like documentables such as DClass, DEnum, DAnnotation, etc.

The contents of each documentable normally represent what you would see in source code. For instance, if you open DClass, you should find that it contains references to functions, properties, companion object, constructors and so on. DEnum should have references to enum entries, and DPackage can have references to both classlikes and top-level functions and properties (Kotlin-specific).

Here's an example of a documentable:

data class DClass(
    val dri: DRI,
    val name: String,
    val constructors: List<DFunction>,
    val functions: List<DFunction>,
    val properties: List<DProperty>,
    val classlikes: List<DClasslike>,
    val sources: SourceSetDependent<DocumentableSource>,
    val visibility: SourceSetDependent<Visibility>,
    val companion: DObject?,
    val generics: List<DTypeParameter>,
    val supertypes: SourceSetDependent<List<TypeConstructorWithKind>>,
    val documentation: SourceSetDependent<DocumentationNode>,
    val expectPresentInSet: DokkaSourceSet?,
    val modifier: SourceSetDependent<Modifier>,
    val sourceSets: Set<DokkaSourceSet>,
    val isExpectActual: Boolean,
    val extra: PropertyContainer<DClass> = PropertyContainer.empty()
) : DClasslike(), WithAbstraction, WithCompanion, WithConstructors,
    WithGenerics, WithSupertypes, WithExtraProperties<DClass>

There are three non-documentable classes that important for this model:

  • DRI
  • SourceSetDependent
  • ExtraProperty.

DRI

DRI stans for Dokka Resource Identifier - a unique value that identifies a specific Documentable. All references and relations between documentables (other than direct ownership) are described using DRI.

For example, DFunction with a parameter of type Foo has only Foo's DRI, not the actual reference to Foo's Documentable object.

Example

For an example of how a DRI can look like, let's take the limitedParallelism function from kotlinx.coroutines:

package kotlinx.coroutines

import ...

public abstract class MainCoroutineDispatcher : CoroutineDispatcher() {

    override fun limitedParallelism(parallelism: Int): CoroutineDispatcher {
        ...
    }
}

If we were to re-create the DRI of this function in code, it would look something like this:

DRI(
    packageName = "kotlinx.coroutines",
    classNames = "MainCoroutineDispatcher",
    callable = Callable(
        name = "limitedParallelism",
        receiver = null,
        params = listOf(
            TypeConstructor(
                fullyQualifiedName = "kotlin.Int",
                params = emptyList()
            )
        )
    ),
    target = PointingToDeclaration,
    extra = null
)

If you format it as String, it would look like this:

kotlinx.coroutines/MainCoroutineDispatcher/limitedParallelism/#kotlin.Int/PointingToDeclaration/

SourceSetDependent

SourceSetDependent helps handling multiplatform data by associating platform-specific data (declared with either expect or actual modifier) with particular source sets.

This comes in handy if expect/actual declarations differ. For instance, the default value for actual might differ from that declared in expect, or code comments written for expect might be different from what's written for actual.

Under the hood, it's a typealias to a Map:

typealias SourceSetDependent<T> = Map<DokkaSourceSet, T>

ExtraProperty

ExtraProperty is used to store any additional information that falls outside of the regular model. It is highly recommended to use extras to provide any additional information when creating custom Dokka plugins.

This element is a bit more complex, so you can read more about how to use it in a separate section.


Documentation model

Documentation model is used alongside Documentables to store data obtained by parsing code comments (such as KDoc/Javadoc).

DocTag

DocTag describes a specific documentation syntax element.

It's universal across source languages. For instance, DocTag B is the same for **bold** in Kotlin and <b>bold</b> in Java.

However, some DocTag elements are specific to a certain language, there are many such examples for Java because it allows HTML tags inside Javadoc comments, some of which are simply not possible to reproduce with Markdown.

DocTag elements can be deeply nested with other DocTag children elements.

Examples:

data class H1(
    override val children: List<DocTag> = emptyList(),
    override val params: Map<String, String> = emptyMap()
) : DocTag()

data class H2(
    override val children: List<DocTag> = emptyList(),
    override val params: Map<String, String> = emptyMap()
) : DocTag()

data class Strikethrough(
    override val children: List<DocTag> = emptyList(),
    override val params: Map<String, String> = emptyMap()
) : DocTag()

data class Strong(
    override val children: List<DocTag> = emptyList(),
    override val params: Map<String, String> = emptyMap()
) : DocTag()

data class CodeBlock(
    override val children: List<DocTag> = emptyList(),
    override val params: Map<String, String> = emptyMap()
) : Code()

TagWrapper

TagWrapper describes the whole comment description or a specific comment tag. For example: @see / @author / @return.

Since each such section may contain formatted text inside of it, each TagWrapper has DocTag children.

/**
 * @author **Ben Affleck*
 * @return nothing, except _sometimes_ it may throw an [Error]
 */
fun foo() {}

DocumentationNode

DocumentationNode acts as a container for multiple TagWrapper elements for a specific Documentable, usually used like this:

data class DFunction(
    ...
    val documentation: SourceSetDependent<DocumentationNode>,
    ...
)
Back to top