Flexible Cluster

[jonioni]

1. Treating Cluster as Node:

   with Cluster("VPC"):
     vpc = VPC("...")
     with Cluster("subnet") as subnet:
       ...
     vpc >> subnet

   Related: Is there a way to have arrows to and from Clusters #17
2. Empty default label:

   with Cluster():
     ...

   Produces empty label instead of "cluster".
3. Node as Cluster Key:

   with Cluster(VPC("...")):
     with Cluster("subnet"):
       ...

   Could be resulting in the same effect as an empty cluster label with an additional node:

   with Cluster(""):
       VPC("...")

   Related: Allow any Node to behave as a Cluster #34, ability to add icons to the cluster label #90

[givanovexpe]

The cluster is different than a node is because a "Cluster" translates to a "subgraph" in graphviz lingo, while "node" and "edge" translate to node and edge respectively. Also when diagrams translate a Cluster into the corresponding graphviz code, it creates a "subgraph" but also a "graph" inside it. So I am not sure this is possible, but I am not expert by far.

[jonioni]

Thanks @givanovexpe . However the proposal here is actually more about python initializer tweaking, and all the features are achievable through current graphviz.

For [1] treating cluster as node, its about being about to point to a subgraph instead of a node. This is achievable by utilizing compound=True diagram attribute and [ltail=subgraphA,lhead= subgraphB] attribute. And for the python part it'd be great if the with Cluster(...) context can be associated with a handle with Cluster(...) as GroupA.

For [2] Empty default label, it's about having a unique Cluster class counter as unique id instead of depending on cluster name for layout and rendering. So that there can be cluster with no name.

For [3] Node as cluster key, it's an enhancement on top of [2] where name can be not limited to :str. In the case of with Cluster(VPC("...")), it'd have same effect as an empty cluster label and a node in the subgraph.