[pazz]

NestedTree.collapsible, is_collapsed, toggle_collapsed, collapse and expand each looked up the tree managing the position and checked that it can collapse. The new private _on_subtree does that once and replaces _get_subtree_for; each method is one line. Found by an audit for over-engineering.

When NestedTrees are wrapped directly in each other, these methods used to call themselves once per level and failed at about a thousand levels. _on_subtree walks down to the managing tree in a loop, so there is no such limit now. test_collapse_deeply_nested collapses through 1100 levels; it fails on main.

Return values are unchanged: collapsible and is_collapsed default to False, the other three return None. A seeded differential test over 30 random nested trees with all collapse calls gives identical results on main and this branch.

Review findings and resolution

review-correctness

• The first version recursed twice per directly nested NestedTree (method and helper), halving the depth that worked: fixed, by walking in a loop, which removes the recursion altogether; regression test added.
• The private _get_subtree_for is removed: left; private and unused elsewhere.
• Checked without findings: return values, the differential test above, 10000 sibling subtrees, the full suite on urwid 3.0.4 and 2.1.2.

review-design

• String dispatch through getattr is slightly harder to grep than direct calls: left; a helper returning the tree would be the same size, and the reviewer did not require it.
• Whether this deserves its own PR: kept separate, as it is a different module from the other cleanups.

🤖 Generated with Claude Code