recursive_seed_part_inner does not respect provided method

[mkarrmann]

recursive_seed_part and recursive_seed_part_inner accept a parameter method which is supposed to be the funciontion used to generate a single district at a give step. However, this method is not passed to recursive_seed_part_inner's recursive calls to itself, meaning it is only used to generate the first district, but not any subsequent one.

It is easy to reproduce and observe this bug by calling recursive_seed_part with a method which has a noticeable side-effect (e.g. logging), and then noticing that that side-effect only occurs once. However, it probably easy to just eye-ball the code to see the problem:

GerryChain/gerrychain/tree.py

Lines 524 to 534 in 04ff4f9

	def recursive_seed_part_inner(
	graph: nx.Graph,
	num_dists: int,
	pop_target: Union[float, int],
	pop_col: str,
	epsilon: float,
	method: Callable = partial(bipartition_tree, max_attempts=10000),
	node_repeats: int = 1,
	n: Optional[int] = None,
	ceil: None = None,
	) -> List[Set]:

GerryChain/gerrychain/tree.py

Lines 591 to 605 in 04ff4f9

	nodes = method(
	graph.subgraph(remaining_nodes),
	pop_col=pop_col,
	pop_target=pop_target,
	epsilon=epsilon,
	node_repeats=node_repeats
	)
	remaining_nodes -= nodes
	assignment = [nodes] + recursive_seed_part_inner(graph.subgraph(remaining_nodes),
	num_dists - 1,
	pop_target,
	pop_col,
	epsilon,
	n=n,
	ceil=ceil)

GerryChain/gerrychain/tree.py

Lines 621 to 629 in 04ff4f9

	chunk_assignment = recursive_seed_part_inner(
	graph.subgraph(chunk),
	num_dists // num_chunks,
	pop_target,
	pop_col,
	epsilon,
	n=n,
	ceil=ceil
	)

Clearly, method is used exactly once, and will not be used in subsequent calls because it is not passed. Instead, the default of partial(bipartition_tree, max_attempts=10000), will be used.