Decision tree diagram type

[ibirothe]

This extends tree diagrams by providing a label for each parent-child-relationship, representing a decision.

Example:

[kharybdys]

We have a need for these labels. I would be open to making a PR for it.

I can see two distinct approaches:

• The Node changes its children data structure from list to dict. The key becomes the label on that edge. The connect function in the TreeDiagram is preferably changed to accept a label (default None) so the new subclass doesn't need to reimplement the connect function, only the draw_connections function.
• The Node is changed to not only keep track of its tree_parent but also a tree_parent_label, with a default of None. This would allow integration into the existing TreeDiagram class instead of requiring a new class.

I'm leaning towards the second approach. Both because of not needing a new class but also because your current approach for drawing the connections is object to parent, not object to children.

Please let me know your thoughts.

[kharybdys]

As "proof" for the validity of integrating it into the current TreeDiagram. This is my current code that I intend to replace with the support from the library when available:

from drawpyo.diagram_types import NodeObject, TreeDiagram

class DecisionNodeObject(NodeObject):
  def __init__(self, parent_edge_label: str | None, **kwargs):
    self.parent_edge_label = parent_edge_label
    super().__init__(**kwargs)

class DecisionTreeDiagram(TreeDiagram):
  def connect(self, source: NodeObject, target: NodeObject):
    super().connect(source, target)
    if isinstance(target, DecisionNodeObject):
      self.links[-1].label = target.parent_edge_label

Ofcourse this implementation is extremely fragile. But the subclass of NodeObject is unneeded, and the subclass of TreeDiagram can also be avoided.

This doesn't do anything more than allowing for labels along the parent-child edges