Exstreme is an implementation of a Stream Push data structure in the way of a runnable graph where all the nodes must be connected and process a message and pass the result to next node(s)
The package can be installed as:
-
Add exstreme to your list of dependencies in
mix.exs
:def deps do [{:exstreme, "~> 0.0.3"}] end
-
Check the documentation: available in Hex
A graph is a data structure that contains nodes connected between them, this graphs must start with only one node and can finish in many nodes, all the nodes in the graph must be connected, for example:
n3
|
n1 - n2 - b1
|
n4
The information of a graph is:
:name
- Name assigned for the graph, if you don't assign a name this will be generated:nodes
- The nodes with their parameters as a keyword list:connections
- The nodes and their connections
The nodes can be of three types:
-
Common
- it represents a node that is connected to can be connected to another node and can receive a message from another node, represented by the n letter. -
Broadcast
- it is a node that can broadcast a message to multiple nodes, represented by the b letter. -
Funnel
- it receives messages from a group of nodes and sends it to the next, represented by the f letter.
A graph could look like this:
n3
| |
n1 - n2 - b1 f1 - n5
| |
n4
It works this way:
- n1 passes the message to n2
- n2 passes the message to b1
- b1 broadcasts the message to n3 and n4
- f1 receives the message from n3 and n4 and packages them as one and sends that to n5
- n5 process the message received from f1
How to create a graph:
graph = GraphCreator.create_graph("name")
{graph, n1} = GraphCreator.create_node(graph, params)
{graph, n2} = GraphCreator.create_node(graph, params)
GraphCreator.add_connection(graph, n1, n2)
A complex one(this is the one for graph above):
graph = GraphCreator.create_graph("name")
{graph, n1} = GraphCreator.create_node(graph, params)
{graph, n2} = GraphCreator.create_node(graph, params)
{graph, b1} = GraphCreator.create_broadcast(graph, params_broadcast)
{graph, n3} = GraphCreator.create_node(graph, params)
{graph, n4} = GraphCreator.create_node(graph, params)
{graph, f1} = GraphCreator.create_funnel(graph, params_funnel)
{graph, n5} = GraphCreator.create_node(graph, params)
graph
|> GraphCreator.add_connection(n1, n2)
|> GraphCreator.add_connection(n2, b1)
|> GraphCreator.add_connection(b1, n3)
|> GraphCreator.add_connection(b1, n4)
|> GraphCreator.add_connection(n3, f1)
|> GraphCreator.add_connection(n4, f1)
|> GraphCreator.add_connection(f1, n5)
The nodes in the graph are named like this if the name of the graph is "demo":
n1
- :n_demo_1n2
- :n_demo_2b1
- :b_demo_1f1
- :f_demo_1
The node params must have a function that is the one called every time a message arrives to the node. The function receives a tuple where the first parameter is the message and the second one the node data, it must return a tuple with :ok and the new message.
params = [func: fn({msg, node_data}) -> {:ok, new_msg} end]
We build a graph after we create it, like this:
graph_built = GraphBuilder.build(graph)
The name of the supervisor is the name of the graph so you can get the pid for the supervisor:
pid =
graph_built.name
|> String.to_atom
|> Process.whereis
Also we can get the pid for the nodes:
Enum.each(graph_built.nodes, fn({nid, params}) ->
pid = Process.whereis(nid)
end)
And we can connect a process to the graph and receive the output of the processing:
[start_node] = Graph.find_start_node(graph_built)
[last_node] = Graph.find_last_node(graph_built)
:ok = GenServer.call(last_node, {:connect, self})
GenServer.cast(start_node, {:next, self, {:sum, 0}})
If I try to build another graph with the same I'll get an error because there can't be two process with the same name.