A composite API is a sequence of tasks that runs synchronously and which execution path is not persisted.
Composite APIs are best described by comparing them with processes :
- Both processes and composite APIS are a sequence of tasks
- Composite APIs run synchronously unlike processes : the result of triggering of a Composite API is the result of the execution, not the request that will contain the result of the execution once it is finished.
- Each composite API can be triggered by calling its API.
- A composite API execution is never paused, which implies that a composite API cannot contains manual tasks nor subprocesses tasks unlike a process.
- A composite API execution path is not persisted unlike a process which enables it to run more quickly.
Triggering a Composite API
A composite API can be triggered in the following ways:
- As a task in a process.
- As a listener in a web interface.
- Directly via its exposed API (from a third-party).
A composite API may be used :
- To speed up the execution of a sequence of tasks.
- To improve listeners performance in web interfaces.
- To compose an API by creating a sequence of web services calls that is meaningful.
The design of a composite API includes the design of the sequence of tasks like in a process and the design of its exposed API.
Contrary to a process, the corresponding REST API of a Composite API can be configured. You can configure :
Method: the HTTP verb.
Accept media type: the media type of the input.
Result media type: the media type of the output.
It is important to note that a Composite API which is triggered from a web interface must be configured with its
Accept media type set to
A good practice is to choose the HTTP verb that matches what the composite API does :
- GET if it only retrieves information
- PUT if it modifies something
- POST if it creates something
- DELETE it it deletes something
Debugging and Maintenance
Since the composite API execution path is not persisted, the test and debug should be thought very differently from what is done with processes.
Retrieving the error
An unhandled error launched during the execution of a Composite API will appear in the customer logs which have been introduced with the
mogador release. They can be accessed in the Logs Application.
When the execution of a composite API fails, the error is logged in the
Custom Logs with the following information :
- Date : Date and time of the error
- Log Level : An unhandled error that aborts the execution of a Composite API will always be
- Message : The error message
- Process ID / Composite API ID : ID of the Composite API
- Project ID : ID of the context project. For a composite API belonging to a project B, subproject of a project A, that is launched by a process or a web interface from project A, the context project will be project A.
- Request ID : Unique ID of the Composite API execution.
In addition and depending on how the Composite API was launched, the error message can also be retrieved in the execution context
- Composite API launched from a process : The error message is returned to the request
- Composite API launched from a web interface : The error message is returned in a
Throwing your own errors
The platform provides a freemarker method
throw that enables you to throw an error and abort the execution of the Composite API.
This method enables you to define a error message, that will be returned into the execution context as seen above.
Composite API can handle errors exactly in the same way as process by using error handlers in their design. Error handlers enable you to capture errors and deal with them.
The Logs Application can also be used to track the execution of composite API by using the
P_log function within the composite API.
This method can be quite useful for debugging.
(@see Composite API Example)