Anytree Module/UniqueDotExporter produces a Tree with redundant edges

Question

so I got the following problem.

I am building a tree that shows dependencies with the Python Module anytree.

Nodes are created with:

Node(""+str(name), parent= parentname)

at the end of my program I want to create a graph with DotExporter.

DotExporter(root, nodeattrfunc=lambda n: 'label="%s"' % (n.name), edgeattrfunc=lambda parent, child: "style=bold").to_picture("filename.png)

The creation of the graph works, but now I got the problem that there are too many redundant edges in big trees.

For example:

A -> B -> C -> D
A -> E -> C -> D

The end graph will be right, but there will be 2 edges/arrows from C to D, but I only want it to show 1.

How do I change that?

Answer 1

Firstly, here is a Minimal Reproducible Example :

from anytree import Node, render

a = Node(name="A", parent=None)
b = Node(name="B", parent=a)
c = Node(name="C", parent=b)
d = Node(name="D", parent=c)
e = Node(name="E", parent=a)
c2 = Node(name="C", parent=e)
d2 = Node(name="D", parent=c)
root = a
print(render.RenderTree(root).by_attr())

A
├── B
│   └── C
│       ├── D
│       └── D
└── E
    └── C

Regarding your question, you will have to remember which Nodes are already created, like so:

from anytree import Node, render


already_seen = {}


def create_node(name, parentname):
    if name in already_seen:
        if already_seen[name].parent.name == parentname:
            # edge already created
            child = already_seen[name]
            return child
        else:
            # the node already exists, but has a new parent
            parent = already_seen[parentname]
            child = already_seen[name]
            parent.children = parent.children + (child,)  # add to parent's children
            return child
    else:  # name never seen
        if parentname in already_seen:
            # new child for a known parent
            parent = already_seen[parentname]
            child = Node(name, parent=parent)
            already_seen[name] = child
            return child
        elif parentname is None:
            # a root
            root = Node(name, None)
            already_seen[name] = root
            return root
        else:
            raise ValueError("unknown parent name " + repr(parentname))


a = create_node(name="A", parentname=None)
b = create_node(name="B", parentname="A")
c = create_node(name="C", parentname="B")
d = create_node(name="D", parentname="C")
e = create_node(name="E", parentname="A")
c2 = create_node(name="C", parentname="E")
d2 = create_node(name="D", parentname="C")

root = a
print(render.RenderTree(root).by_attr())

A
├── B
└── E
    └── C
        └── D

But you have another problem: your graph dependency is not a tree : it is a (directed) graph !
But the library you are using, anytree , as the name implies, can only model trees. So you can not have your C node have both B and E as parent. There is no solution for that other than using another library, which can support (directed) graphs.