Found in the PR #404 round-2 review (2026-07-24), pre-existing: the 2D lumped de-smear assembles P2 line masses (vertex-midpoint-vertex triples), so a P1 scalar/velocity trace has no midpoint keys and dies with a bare KeyError: (x, y) (also hit independently while building the #407 regression tests — the workaround was degree=2). With PR #404 merged, the docstring enumerates 'auto selects lumped for 2D traces', which now over-promises for P1. Fix: assemble the P1 two-node line mass when the trace has no midpoints (mirror of the 3D P1/P2 dispatch #404 added), or raise a legible NotImplementedError.
Underworld development team with AI support from Claude Code
Found in the PR #404 round-2 review (2026-07-24), pre-existing: the 2D lumped de-smear assembles P2 line masses (vertex-midpoint-vertex triples), so a P1 scalar/velocity trace has no midpoint keys and dies with a bare
KeyError: (x, y)(also hit independently while building the #407 regression tests — the workaround was degree=2). With PR #404 merged, the docstring enumerates 'auto selects lumped for 2D traces', which now over-promises for P1. Fix: assemble the P1 two-node line mass when the trace has no midpoints (mirror of the 3D P1/P2 dispatch #404 added), or raise a legible NotImplementedError.Underworld development team with AI support from Claude Code