Files
initializr/initializr-docs/src/main/asciidoc/metadata-format.adoc
Stephane Nicoll 41f844a3ad Serve version format that is backward compatible
This commit makes sure the metadata format uses a backward compatible
version format even if the new format is used. It also introduces a
new metadata version (2.2) that can be used by clients that support the
new version format.

See gh-1092
2020-06-03 14:32:41 +02:00

181 lines
7.3 KiB
Plaintext

[[metadata-format]]
= Metadata Format
This section describes the hal/json structure of the metadata exposed by the
initializr. Such metadata can be used by third party clients to provide a list of
options and default settings that can be used to request the creation of a project.
A third-party client is advised to set a `User-Agent` header for *each* request
sent to the service. A good structure for a user agent is `clientId/clientVersion`
(i.e. `foo/1.2.0` for the "foo" client and version `1.2.0`).
== Service Capabilities
Any third party client can retrieve the capabilities of the service by issuing a
`GET` on the root URL using the following `Accept` header:
`application/vnd.initializr.v2.2+json`. Please note that the metadata may evolve in a
non backward compatible way in the future so adding this header ensures the service
returns the metadata format you expect.
The following versions are supported:
* `v2` initial version, with support of V1 version format only
* `v2.1` support compatibility range and dependencies links
* `v2.2` (current) support for V1 and V2 version formats.
This is an example output for a service running at `https://start.spring.io`:
.request
include::{snippets}/metadataWithCurrentAcceptHeader/http-request.adoc[]
.response
include::{snippets}/metadataWithCurrentAcceptHeader/http-response.adoc[]
The current capabilities are the following:
* Project dependencies: these are the _starters_ really or actually any dependency
that we might want to add to the project.
* Project types: these define the action that can be invoked on this service and a
description of what it would produce (for instance a zip holding a pre-configured
Maven project). Each type may have one more tags that further define what it
generates.
* Packaging: the kind of projects to generate. This merely gives a hint to the
component responsible to generate the project (for instance, generate an executable
_jar_ project).
* Java version: the supported java versions
* Language: the language to use (e.g. Java)
* Boot version: the Spring Boot version to use
* Additional basic information such as: `groupId`, `artifactId`, `version`, `name`,
`description` and `packageName`.
Each top-level attribute (i.e. capability) has a standard format:
* A `type` attribute that defines the semantic of the attribute (see below).
* A `default` attribute that defines either the default value or the reference to
the default value.
* A `values` attribute that defines the set of acceptable values (if any). This can
be hierarchical (with `values` being held in `values`). Each item in a `values` array
can have an `id`, `name` and `description`).
The following attribute `type` are supported:
* `text`: defines a simple text value with no option.
* `single-select`: defines a simple value to be chosen amongst the specified options.
* `hierarchical-multi-select`: defines a hierarchical set of values (values in
values) with the ability to select multiple values.
* `action`: a special type that defines the attribute defining the action to use.
Each action is defined as a HAL-compliant URL. For instance, the `maven-project` type
templated URL is defined as follows:
.Type link example
include::{snippets}/metadataWithCurrentAcceptHeader/response-fields/_links.maven-project.adoc[]
You can use Spring HATEOAS and the `UriTemplate` helper in particular to generate an
URI from template variables. Note that the variables match the name of top-level
attribute in the metadata document. If you can't parse such URI, the `action`
attribute of each type gives you the root action to invoke on the server. This
requires more manual handling on your end.
=== Project dependencies
A dependency is usually the coordinates of a _starter_ module but it can be just a regular
dependency. A typical dependency structure looks like this:
```json
{
"name": "Display name",
"id": "org.acme.project:project-starter-foo",
"description": "What starter foo does"
}
```
The name is used as a display name to be shown in whatever UI used by the remote
client. The id can be anything, really as the actual dependency definition is
defined through configuration. If no id is defined, a default one is built using the
`groupId` and `artifactId` of the dependency. Note in particular that the version is
**never** used as part of an automatic id.
Each dependency belongs to a group. The idea of the group is to gather similar
dependencies and order them. Here is a value containing the `core` group to
illustrates the feature:
.Dependency group example
include::{snippets}/metadataWithCurrentAcceptHeader/response-fields/dependencies.values.0.adoc[]
Each dependency can have _links_ (in a HAL-compliant format). Links are grouped by
"relations" that provide a semantic to the link. A link can also have a _title_ and
its URI can be templated. At the moment, the only valid parameter is `bootVersion`.
The official relations are:
* `guide`: link to an how-to or guide that explain how to get started
* `reference`: link to a section of a reference guide (documentation)
=== Project types
The `type` element defines what kind of project can be generated and how. For
instance, if the service exposes the capability to generate a Maven project, this
would look like this:
.Project type example
include::{snippets}/metadataWithCurrentAcceptHeader/response-fields/type.values.0.adoc[]
You should not rely on the output format depending that information. Always use the
response headers that define a `Content-Type` and also a `Content-Disposition`
header.
Note that each id has a related HAL-compliant link that can be used to generate a
proper URI based on template variables. The top-level `type` has, as any other
attribute, a `default` attribute that is a hint to select what the service consider
to be a good default.
The `action` attribute defines the endpoint the client should contact to actually
generate a project of that type if you can't use the HAL-compliant url.
The `tags` object is used to categorize the project type and give _hints_ to 3rd
party client. For instance, the _build_ tag defines the build system the project is
going to use and the _format_ tag defines the format of the generated content (i.e.
here a complete project vs. a build file. Note that the `Content-type` header of the
reply provides additional metadata).
=== Packaging
The `packaging` element defines the kind of project that should be generated.
.Packaging example
include::{snippets}/metadataWithCurrentAcceptHeader/response-fields/packaging.values.0.adoc[]
The obvious values for this element are `jar` and `war`.
=== Java version
The `javaVersion` element provides a list of possible java versions for the project:
.Java example
include::{snippets}/metadataWithCurrentAcceptHeader/response-fields/javaVersion.values.0.adoc[]
=== Languages
The `language` element provides a list of possible languages for the project:
.Language example
include::{snippets}/metadataWithCurrentAcceptHeader/response-fields/language.values.0.adoc[]
=== Boot version
The `bootVersion` element provides the list of available boot versions
.Spring Boot version example
include::{snippets}/metadataWithCurrentAcceptHeader/response-fields/bootVersion.values.0.adoc[]
== Defaults
Each top-level element has a `default` attribute that should be used as a hint to
provide the default value in the relevant UI component.