Python circular dependency

Question

I have two classes AST and ASTNode for working with some tree-like data. Both these classes have networkx graph as a field and a node id (integer). For AST this id refers to the root of a tree. These two classes looks pretty similar, but I have splited functionality in a following way:

• I use AST for global functionality like get nodes of some type from a whole tree
• I use ASTNode for local functionality like get direct child of a node and other node properties

I need to be able to shift from AST to ASTNode functionality and backward when needed. Here comes circular dependency. I want to be able:

• to query some ASTNode from AST
• to build a subtree from ASTNode with this node as a root

So this makes ASTNode to know how to create AST. How can I deal with it? Should I use absolute module import? Or it can be done otherwise?

Currently I moved get_subtree(node) function to AST, but this make usage really uncomfortable (you have to pass node and a tree together all the time).

• Yaroslav Kishchenko: So how does the dependencies would look like? ASTBuilder knows how to create AST and ASTNode. And AST and ASTNode use it? Seems like I didn't get the solution. Coud you clarify?
• eparham7861: For AST, is the ID that refers the root of tree considered the overall tree's root node, or the current root node of some subtree?
• Yaroslav Kishchenko: @eparham7861 If AST is original tree build from source code, than it refers to its root. If it is subtree, than to the root of subtree.
• eparham7861: Would it be possible for you to push all of the functionality to a separate class, like ASTLogic or ASTInteractions? That way, instead of the nodes "knowing" how to create the tree, this separate class would be able to hold global functionality (AST specific), "local" functionality (AST->ASTNode specific), and generation functionality (sub-AST specific).