The Introspector is a component that introspects the current local
topology of an Axon system - that is what components there are and how
they are wired up.
It continually outputs any changes that occur to the topology.
Example Usage
Introspect and display whats going on inside the system:
MyComplexSystem().activate()
pipeline( Introspector(),
AxonVisualiserServer(noServer=True),
)
More detail
Once activated, this component introspects the current local topology
of an Axon system.
Local? This component examines its scheduler to find components and
postmen. It then examines them to determine their inboxes and outboxes
and the linkages between them. In effect, it determines the current
topology of the system.
If this component is not active, then it will see no scheduler and
will report nothing.
What is output is how the topology changes. Immediately after
activation, the topology is assumed to be empty, so the first set of
changes describes adding nodes and linkages to the topology to build up
the current state of it.
Subsequent output just describes the changes - adding or deleting
linkages and nodes as appropriate.
Nodes in the topology represent components and postboxes. A linkage
between a component node and a postbox node expresses the fact that that
postbox belongs to that component. A linkage between two postboxes
represents a linkage in the Axon system, from one component to
another.
This topology change data is output as string containing one or more
lines. It is output through the "outbox" outbox. Each line may be one of
the following:
- "DEL ALL"
- the first thing sent immediately after activation - to ensure that
the receiver of this data understand that we are starting from
nothing
- "ADD NODE <id> <name> randompos component"
- "ADD NODE <id> <name> randompos inbox"
- "ADD NODE <id> <name> randompos outbox"
- an instruction to add a node to the topology, representing a
component, inbox or outbox. <id> is a unique identifier.
<name> is a 'friendly' textual label for the node.
- "DEL NODE <id>"
- an instruction to delete a node, specified by its unique id
- "ADD LINK <id1> <id2>"
- an instruction to add a link between the two identified nodes. The
link is deemed to be directional, from <id1> to <id2>
- "DEL LINK <id1> <id2>"
- an instruction to delete any link between the two identified nodes.
Again, the directionality is from <id1> to <id2>.
the <id> and <name> fields may be encapsulated in double
quote marks ("). This will definitely be so if they contain space
characters.
If there are no topology changes then nothing is output.
This component ignores anything arriving at its "inbox" inbox.
If a shutdownMicroprocess message is received on the "control" inbox,
it is sent on to the "signal" outbox and the component will
terminate.
How does it work?
Every execution timeslice, Introspector queries its scheduler to
obtain a list of all components. It then queries the postoffice in each
component to build a picture of all linkages between components. It also
builds a list of all inboxes and outboxes on each component.
This is mapped to a list of nodes and linkages. Nodes being
components and postboxes; and linkages being what postboxes belong to
what components, and what postboxes are linked to what postboxes.
This is compared against the nodes and linkages from the previous
cycle of processing to determine what has changed. The changes are then
output as a sequence of "ADD NODE", "DEL NODE", "ADD LINK" and "DEL
LINK" commands.
Introspector() -> new Introspector component.
Outputs topology (change) data describing what components there are,
and how they are wired inside the running Axon system.
Returns the current set of components, postboxes and interpostbox
linkages.
Main loop.
