Home — Data Types — Provider and Client — Tutorial — Cheat Sheet — Transformations — Generator Tools
Orchestration Flows
Note: The status of the language concepts specified here is Experimental Preview. Concepts and their tool support have been validated, for instance via code generation to multiple target languages (see below). They are rather stable now; many examples and test cases exist. However, they might still change in future versions of MDSL.
Use Cases (When to Specify)
- Business workshop result capturing (for instance, event storming sessions)
- Integration with DDD tools such as Context Mapper
- Rapid prototyping (for instance, yielding event storming results coming in via Context Mapper)
- Chaining of API operations that depend on each other (API call sequencing and branching)
- Integration flows, as input for middleware such as Apache Camel or event simulation tools
News (Jan 13, 2022): The blog post and demo script “Event-Driven Service Design: Five Steps from Event Storming to OpenAPI and Camel Flow” features the MDSL flow modeling concepts in action.
Simple Flow Steps
The following basic flow was generated from the sample story scenario with one of the quick fix transformations in the MDSL tools:
flow Scenario1Flow realizes Scenario1
event something_has_happened triggers command startProcess
command startProcess emits event startProcessCompleted
event CRUDSomeBusinessObjectTrigger triggers command CRUDSomeBusinessObject
event startProcessCompleted triggers command startProcessCompletedProcessor
command CRUDSomeBusinessObject emits event CRUDSomeBusinessObjectCompleted
In addition to the very basic event-command and command-event pairs used above, it is possible to specific event-command-event triples:
event Event1 triggers command Command1 emits event Event2
The above snippet is semantically equivalent to the folllowing two rules:
event Event1 triggers command Command1
command Command1 emits event Event2
Branching and Joining
Events may trigger multiple commands with and, or, xor semantics:
event Event1 triggers command Command1 + Command2
event Event2 triggers command Command1 o Command2
event Event2 triggers command Command1 x Command2
Commands may emit multiple events (and, or, xor):
command StartFlowCommand emits event Event1 + Event2
command Command1 emits event Event1 o Event2
command Command2 emits event Event1 x Event2
Two or more events may be part of a join, modeling the Aggregator pattern:
event Command1Done + Command2Done trigger command JoinedContinuationCommand
Event joining may express that something can only happen when two (or more) commands have been executed previously.
Unlike events, commands cannot be joined in a single rule. Alternative event emission (by two or more commands) can be expressed:
command Command1 emits event Event1
command Command2 emits event Event1
The same works for events:
event Event 1 triggers Command1
event Event 2 triggers Command1
Either Event1 or Event2 (or both) trigger Command1.
Versioning
If they contain a data structure definition (which is optional), both event types and command types can be versioned (just like entire API descriptions and endpoint types and and data types):
event type something_has_happened version "v1" D<string>, CRUDSomeBusinessObjectTrigger
event type startProcessCompleted version "v1.0" {"abc":D<int> , "def":D<int>}
command type startProcess version "v10"
The Version Identifier (such as "v1"
) is a plain string, which may be structured according to the Semantic Versioning conventions. Note that a data element definition such as D<string>
has to ber present; there is no point in versioning a placeholder event/command type.
Grammar Excerpts
Flow and flow steps (with technology binding)
Orchestration:
'flow' name=ID
('realizes' scenario=[IntegrationScenario])? ('type' type=FlowType)?
steps+=FlowStep*
flowBinding=OrchestrationBinding?
;
enum FlowType:
APPLICATION_FLOW
| INTEGRATION_FLOW
| SERVICE_ORCHESTRATION
| API_CALL_SEQUENCING
| EVENT_SOURCING
;
OrchestrationBinding:
'binding' /* 'flow' */ 'to' (eptb+=EndpointTypeBinding|cb+=ChannelBinding)+
('implemented' 'as' ft=FlowTechnology)?
;
FlowTechnology:
'APACHE_CAMEL_ROUTE' | 'SPRING_INTEGRATION_FLOW'
| 'BPMN_PROCESS' | 'BPEL_PROCESS' | 'YET_ANOTHER_FLOW_LANGUAGE'
| 'PLAIN_PROGRAMMING' | STRING
;
Event Types and Command Types
Only event types and command types that have been declared can be used in flows. The syntax for these declarations is similar to that of the data types used in endpoint contracts:
EventTypes:
'event' 'type' events+=EventType (',' events+=EventType)*
;
EventType:
name=ID (content=ElementStructure)? // the domain data reported
('version' svi=SemanticVersioningIdentifier)?
;
CommandTypes:
'command' 'type' commands+=CommandType (',' commands+=CommandType)*
;
CommandType:
name=ID (subject=ElementStructure)? // in/out data or business object/item
('version' svi=SemanticVersioningIdentifier)?
;
The order of the rules does not matter; the control flow is specified partially and implicitly. If commands and events appear in multiple rules, this information is aggregated and assembled in gen model so that generator tools can work on canonical, streamlined models.
See section “Processes and Event/Command Flows” in the CML language reference for more explanations.
Tool Support
A number of the quick fix transformations operate on flows and their steps. The in-flow transformations are:
MDSL element | Transformation (Quick Fix) | Resulting MDSL | Comments (Preconditions, Variants) |
---|---|---|---|
Command in flow step | Add a domain event production step | Let the new step emit a [CommandName]Completed event |
Branching/joining not yet supported in these transformations |
Event in flow step | Add a command invocation step (triggered by event) | Let event trigger a new command [EventName]Processor in a new flow step |
Supports incremental modeling |
Combined step | Split step into a command invocation and an event production | Replaced single combined step with two simple ones | Only works if command is simple |
Events in DEP step | Inject a parallel branch (with join) | Two events, two commands, one aggregator added | Added at end of flow |
Command in CIS step | Inject a choice branch (with join) | Two events, two commands, one aggregator added | Added at end of flow; condition not specified |
Command in DEP step | Merge all single event productions of this command | Consolidates multiple rules into one that emits multiple events | Events are composed with or semantics |
Flows can be used to generate endpoint types (with operations realizing the steps):
MDSL element | Transformation (Quick Fix) | Resulting MDSL | Comments (Preconditions, Variants) |
---|---|---|---|
Flow | Derive endpoint type from application flow | Endpoint added to name of new endpoint, which receives a flow start event |
Operations can be added later |
Command in flow step | Derive an operation to endpoint type that realizes this command | MAP decorator: STATE_TRANSITION_OPERATION |
transitions from ... to ... added to operation |
Event in flow step | Derive an event handler operation for flow step | MAP decorator: EVENT_PROCESSOR |
emitting event added to operation, event added to receives part of endpoint contract |
Freemarker templates allow converting MDSL flows into Apache Camel configurations and into the Sketch Miner story format (which in turn yields BPMN).
Known Limitations
- It is possible to specify incomplete, even invalid models (from a workflow management middlware point of view); results from workshops often yield such imperfect, intermediate models.
- The MDSL Tools do not fully support all flow modeling concepts of MDSL:
- Subprocesses cannot be processed by quick fix transformations and generators.
- Combined event-command-event rules cannot be processed by all tools completely (quick fixes, code generators). For instance, the productivity quick fixes expect the command part of such rules to contain a single simple command only.
- The Context Mapper Language (CML) has support for operations that delegate to service methods in Aggregates; mapping such operations is not supported in the CML to MDSL generator in Context Mapper (as the MDSL grammar does not have such a concept).
- While the models are normalized into a canonic format (see flow gen model export), not all duplicate event emissions and command triggers are detected and may cause invalid gen model content and generator output.
- The flow models do not mandate a single control flow that is guaranteed to terminate.
- Loops can be detected by the MDSL Tool generators, but not all edge cases are covered.
- Many validations could be implemented, for instance single entrance, single exist checks. There is a large body of work in the workflow and BPM community on these topics, which could possibly be leveraged.
Site Navigation
- Language specification:
- Scenarios and stories
- Service endpoint contract types (this page) and data contracts (schemas)
- Bindings and instance-level concepts
- Quick reference, tutorial and tools
- MDSL homepage
Copyright: Olaf Zimmermann, 2018-2022. All rights reserved. See license information.