Circular imports and the standard fix

Question

Accepted Answer

Here is what happens step by step: Python starts importing A, creates a partial module object in sys.modules, then begins executing A's top-level code. A's code hits 'import B', so Python starts importing B and adding it to sys.modules. B's top-level code runs and hits 'import A'. Python sees A is already in sys.modules, so it returns the partial object - A's names that were defined before the 'import B' line exist, but anything after it does not. If B tries to use a name from A that hasn't been defined yet, you get AttributeError. The standard fix is to defer the import: move 'import A' inside the specific function in B that needs it. By the time that function is called, both modules are fully loaded. The deeper fix is usually to restructure so the dependency only flows one way, or extract the shared code into a third module that neither A nor B imports.