[claude]

Tests only: tests/test_config.py.

What the new tests check

• Exact error text from load (_describe):
  • an unknown top-level key gives the file: has unknown keys ['bogus']. No table path exists, so the message names the file.
  • a value of the wrong kind gives source.magnitudes: 3 is not a dict, using the type's __name__ rather than its repr (dict[str, float]).
• [pulse] sample interval: dt_s = 0.0 and dt_s = -1.0 are refused with pulse: a sample interval of … s is none. A tiny positive interval (1e-6) is accepted.
• [segments] shape: segments = 3 gives segments: 3 is not a table, and segments = { next = 3 } gives segments.next: 3 is not a table.
• Propagation at build time:
  • a two-segment config with no [propagation] table is refused.
  • a predetermined tree that names a segment that doesn't exist is refused (propagation.parents: ['nxt'] are not segments).
  • a valid predetermined tree produces exactly the given parents.

Coverage of rupture_generator/config.py

• Before: 97% (missing lines 122, 422, 681, 700, 851, 860)
• After: 99% (missing line 122)

Line 122 is the NotImplementedError in the private abstract _DepthProfile.field. Every concrete profile overrides it, so it was left uncovered.

Surviving mutants (all in _describe)

Killed by the exact-message tests:

• mutmut_25: the InvalidFieldValue reason becomes None.
• mutmut_26, mutmut_32, mutmut_33: the type is printed as dict[str, float] instead of dict.
• mutmut_43, mutmut_44: the file becomes XXthe fileXX / THE FILE.

Judged equivalent:

• mutmut_5, mutmut_6 (reason = str(error) → None / str(None)): the error passed in is always one of the types the match handles. load and _segments only catch InvalidFieldValue, MissingField, ExtraKeysError or RuptureGeneratorError, so the first loop iteration always overwrites the starting reason.
• mutmut_28, mutmut_31 (getattr(t, '__name__', t) → getattr(t, '__name__', None) / getattr(t, '__name__')): these differ only for a field type with no __name__. On Python 3.14 every type mashumaro reports here has one, including generic aliases, unions and type aliases.

No issues filed.

🤖 Generated with Claude Code

[github-actions]

Benchmarks

739f908 against base 36a8c4a, on AMD EPYC 9V45 96-Core Processor. Times are medians over the rounds. The median change is 2.1%; changes past 5% are in bold.

Kernels

kernel	base	head	change	head IQR	rounds	padded cells
embed[Mw6.5]	29 ms	28.3 ms	-2.5%	303 µs	34	448,000 (+0.0%)
circulant_draw[Mw6.5]	2.72 ms	2.59 ms	-4.5%	153 µs	237
embed[Mw7.5]	90.9 ms	88.4 ms	-2.8%	755 µs	11	1,204,224 (+0.0%)
circulant_draw[Mw7.5]	6.39 ms	6.15 ms	-3.7%	340 µs	139
eikonal_solve	961 ms	1e+03 ms	+4.1%	18 ms	5
synthesise_pulses[impulse]	391 µs	384 µs	-1.8%	17.3 µs	948
synthesise_pulses[liu]	344 ms	347 ms	+0.7%	4.78 ms	5

End to end

Each round is the command line in a fresh interpreter, start-up included. Peak RSS is the largest the child process reached over the rounds.

example	base time	head time	change	base peak RSS	head peak RSS	change
alpine_hope	6.93 s	6.86 s	-1.0%	294 MiB	294 MiB	-0.1%
beavan	2.46 s	2.51 s	+2.1%	232 MiB	232 MiB	+0.0%
colombia	6.28 s	6.39 s	+1.8%	3,167 MiB	3,168 MiB	+0.0%
crustal	1.05 s	1.06 s	+1.7%	245 MiB	244 MiB	-0.2%
landers	2.06 s	2.07 s	+0.5%	385 MiB	385 MiB	-0.1%
northridge	1.06 s	994 ms	-6.5%	291 MiB	290 MiB	-0.2%
two_faults	1.77 s	1.78 s	+0.4%	354 MiB	354 MiB	-0.2%
two_faults_segments	1.05 s	1.08 s	+2.6%	193 MiB	193 MiB	+0.2%
venezuela	12.6 s	12.4 s	-2.1%	5,766 MiB	5,766 MiB	+0.0%