Fix KeyError when solving after insert_node (non-contiguous element ids) (#308)

[gaoflow]

Summary

Fixes #308.

insert_node() splits an element by appending two new elements and then removing the original via remove_element(). This leaves system.element_map with non-contiguous keys — e.g. after inserting a node into element 2 of a 2-element model, the map holds {1, 3, 4}.

assemble_system_matrix iterated positionally:

for i in range(len(system.element_map)):
    element = system.element_map[i + 1]

which assumes the ids are exactly 1..N. With a gap, element_map[2] no longer exists and validation/solving raises KeyError: 2.

This positional assumption is isolated to this one loop — everywhere else in the codebase indexes element_map by an actual element id, and add_element assigns ids from a monotonic counter that is never reused.

Fix

Iterate over the elements directly:

for element in system.element_map.values():

Assembly is an order-independent accumulation into the global system matrix indexed by node id, so this is identical to the old behaviour for the contiguous case and correct when ids have gaps.

Reproduction (raises on master)

from anastruct import SystemElements
s = SystemElements(EA=15000, EI=5000, mesh=101)
s.add_element(location=[[0, 0], [10, 4]])
s.add_element(location=[[10, 4], [10, 0]])
s.add_support_hinged(node_id=1)
s.add_support_hinged(node_id=3)
s.q_load(q=5, element_id=1, direction='element')
s.point_load(node_id=2, Fx=4, Fy=0)
s.insert_node(element_id=2, factor=0.25)
s.point_load(node_id=4, Fx=8, Fy=0)
s.solve()   # KeyError: 2  (assembly.py)

Verification

• New end-to-end test context_insert_node_then_solve reproduces the issue and additionally asserts the solved result is numerically identical to the equivalent model built with the extra node defined up front (same displacements at every shared node). It raises KeyError: 2 on master and passes with this change.
• Full test suite passes (the 6 pre-existing test_plotter failures — 'Plotter' object has no attribute 'plot_structure' — are unrelated to this change and fail identically on master).
• black and mypy clean on the changed files.

Note

This supersedes #309 (from the issue reporter @thisisapple), which proposed the same element_map.values() idea but contains a syntax error (for keyword dropped) and no test. Credit to them for the original diagnosis.

[github-actions]

Thank you so much for contributing to the open-source anaStruct project! Your contribution will help thousands of engineers work more efficiently and accuractely.

Now that you've created your first pull request here, please don't go away; take a look at the bottom of this page for the automated checks that should already be running. If they pass, great! If not, please click on 'Details' and see if you can fix the problem they've identified. Keep in mind that this repository uses the black autoformatter, pylint linter, and mypy type-checking; the most common problems can be fixed by making sure you've installed and run those systems. A maintainer should be along shortly to review your pull request and help get it added to anaStruct!

[brookshsmith]

Thank you @gaoflow for your PR here as well! Again, a simple but needed fix and you've added a good test too. Merged.