Extended Messaging Intermediate Language (EMIL)


The original MIL cannot be used directly because Pousse-Café:

In addition to above items, it is desirable that complex processing chains can be represented in a structured way in order to enable a quick overview of the process as a whole. Therefore, it should be possible to represent the fact that an event is consumed by an aggregate root and in turn triggers the issuance of another which is itself consumed by a factory in order create a new aggregate etc. Unfortunately, MIL, while not preventing this explicitly, does not enable it in an unambiguous way unless you take indentation as a way of detecting the nesting of events issuance.

Therefore, an extension of MIL called EMIL (Extended MIL) and which takes PC’s specificity into account as well as a need to represent complex processes in a structured way is proposed.

Aggregate update

Below is a first example of EMIL.

process InvoiceManagement

SendCustomerInvoice? -> Ru{SendRunner}
        :CustomerInvoiceSent! -> . [Communication]
        :InvoiceAged! -> Ru{UpdateRunner}

process X statement tells the name of the process being described. If there is not focus on a particular process, * can be used instead (i.e. all processes are described). In the following, EMIL snippets will be shown so this statement will be ignored. It is however mandatory in full EMIL descriptions.

X{...} is an operator used to qualify a name. In above example, the Ru prefix tells that SendRunner is a runner. The other supported prefixes are Re for repositories, F for factories and P for processes (see examples below).

The [...] operator is added in order to enrich MIL with notes. The semantics behind notes depend on the context (see below for more details).

@Invoice[send] represents message listener send of Invoice’s aggregate root. As there is a single message listener per message per aggregate root, this information could actually be optional, but it is useful to tell how the aggregate’s state is altered by the consumption of the message if the message listener name is “intention revealing” enough.

The [...] after ` -> . expression tells the consumedByExternals list of ProducesEvent` annotation.

It is interesting to note that @InvoiceStatistics[update] replaces somehow the * operator in the original MIL by actually describing a state change.

The :. token represents the end of a list of issued events.

Aggregate creation

Next is an example of aggregate creation described with EMIL:

CreateInvoice? -> F{InvoiceFactory}[createNew]
        :InvoiceCreated! -> . [CommunicationSystem]

As explicited above, the ProducesEvent annotation is on aggregate root’s onAdd method.

Above aggregate creations is “required” i.e. it will unconditionally be executed. However, there are cases where creation is optional (e.g. if the aggregate already exists). For this purpose, cardinality operators are introduced. # tells that a creation is optional (i.e. the set of created aggregates contains one or no elements). At the same time, there are situations where several aggregates may be created. The operator + tells that 0 to N instances of an aggregate may be created.

Below two examples, one for optional aggregate creation and one for multiple aggregates creation:

NewReceivableRegistered! [AccountSystem] -> F{InvoiceFactory}[createIfNotAlreadyPresent]#
NewReceivablesRegistered! [AccountSystem] -> F{InvoiceFactory}[createMissing]+

Aggregate deletion

Below is an example of aggregate deletion described with EMIL. In above examples, all event issuances are “required”. However, there are cases where issuance is optional. For this purpose, the # operator may also be used:

DeleteInvoice? -> Re{InvoiceRepository}[removeIfExists]
        :InvoiceDeleted!# -> . [CommunicationSystem]

As explicited above, the ProducesEvent annotation is on aggregate root’s onDelete method.

Reference external systems

In PC, a process may be triggered by an event coming from an external sub-system. In that case, a note can be used to name the origin of the event as shown below:

ReceivableOverdue! [AccountingSystem] -> Ru{MarkOverdueRunner}
        :InvoiceMarkedOverdue! -> . [CommunicationSystem]

The note after the top event being consumed gives the value of consumesFromExternal attribute of MessageListener annotation.

Reference other processes

It may happen that, when focusing on a given process, a message is actually handled by another process. But as the focus is set on the given process, its consumption is not explicitly represented. Instead, the other process is represented as a component which consumes a message e.g.

CancelInvoice! -> Ru{CancelRunner}
        :InvoiceCancelled! -> P{CustomerManagement}

In above example, CustomerManagement is another process of the domain (i.e. not an external system) which contains at least one listener consuming event InvoiceCancelled.

Convert EMIL to MIL

As EMIL is only an extension of MIL, it is easy to come back on an original MIL syntax by applying the following string transformations:


The latest grammar of EMIL is available here. It is written in the form of an ANTLR4 grammar, see ANTLR documentation to learn the syntax (which is rather similar to EBNF).

Back and forth between EMIL and code

Pousse-Café comes with a Maven plugin which gives access to a set of tools enabling to output EMIL based on source code and update existing code based on an EMIL representation of a process.

The most interesting one is provided through the update-process goal. Running this goal in your project opens an editor containing either an empty EMIL process (i.e. only the header with no consumption), either the representation of an existing process. You may then update the process. Once done, just save your changes and exit the editor, the code will be updated accordingly. If no change was brought to the initial content, code generation will be skipped.

Pousse-Café’s code generation tool will:

However, Pousse-Café’s code generation tool will not generate any data model (i.e. attributes) nor actual behavior (i.e. listeners body). Comments starting with TODO are put where code should be completed. Please refer to the reference guide for details about generated code.