build(deps): a quantity that changes by regime is stated once, so ten of the ladder's recorded differences from PyPSA are gone

[FBumann]

Prompt: "292 merged." — the corpus adopting cases:, which this follows.

Note

The following content was generated by AI.

Stacked on #1464, which teaches both engines cases:. This is the corpus arriving:
math-spec v0.0.0-alpha.51 → v0.0.0-alpha.52, where the PyPSA file states each split
constraint once.

Ten recorded differences from PyPSA are gone — every one of the regime splits, which
were the only class with a route to closing. All 15 rungs still MATCH, and five of them
now carry no structural difference at all.

	before	after
recorded deviations	32	22
PyPSA names carrying one	26	19
rungs with a region difference	7	4

rung_02_storage         2 region -> 0        rung_08_modular_big_m   6 region -> 0
rung_03_expansion       9 region -> 3        rung_12_linearized_uc   4 region -> 0
rung_04_ramps           4 region -> 0        rung_05_global_limits   4 region -> 2
rung_07_commitment      4 region -> 0

What the remaining 22 are

None is a language gap, and none can be closed by stating the file differently.

• 4 are checked rather than excused. The negated: entries (fix(differential): a dual the file writes negated is held to the negative of PyPSA's rather than excused #1444) assert
  ours + theirs == 0 at max_abs_diff 0.0. They are proofs living in the same file.
• 5 are the sense-splits — one block per ==, <=, >= where PyPSA writes one row
  per labelled constraint whatever its sense. cases: is value-level and cannot reach an
  operator; that is
  math-spec#239, closed
  not planned.
• 13 are PyPSA's own bookkeeping, which the file is right not to imitate: the
  objective_constant fixed variable (3), the capacity floor and cap copied once per
  scenario over one variable (5), a binary bounded at 1 and capped by a row (4), and one
  degenerate dual.

The rung projection keeps a cased quantity whole

projection.py cuts each rung's file to what that rung builds. It cannot cut inside a
cased quantity: doing so decides from the names a region reads whether that region
applies, which only the region's mask is entitled to say — and a mask reads the other way
as often as not, not committable being true exactly where committable is absent. A
first cut that dropped a region on that basis lost a term worth 75 on rung 3 and was
caught by the projection gate, which holds each cut to the full file's objective.

Kept whole, a cased quantity states what the file states, and the declarations it names
are kept whether or not the rung fills them — as the file itself does. The visible
consequence is that three tables (Generator_status_initial,
Generator_ramp_limit_start_up, Generator_ramp_limit_shut_down) are now first declared
on rung 3 rather than rung 7, the ramp allowance's boundary region naming them there
even though its mask is empty.

Verified

• PyPSA parity's own runner against the corpus at v0.0.0-alpha.52: all 15 rungs
  MATCH, exit 0, and the committed certificate, projections and tables are what a re-run
  produces. The gate that reports a reason no rung needs is what named the ten stale
  entries, so their removal is measured rather than asserted.
• pixi run lint / format-check clean; pyrefly 0 errors, both legs.
• pixi run test — 3461 passed, 254 skipped, 94 failed, every failure xpress (the
  expired local licence, matched by file and parameter id). The three extra skips against
  feat(language): a quantity defined by cases may mix a constant, a parameter and a variable across its regions #1464 are test_merge_markers on the three tables that moved rung.
• tools.ladder --check clean; the index and all 15 rung pages regenerated.
• Not run: pixi run test-bench — no bench/ file changed.

Deliberately not done

• No deviation was retired by hand. The runner named all ten; nothing else was touched.
• The sense-splits are left recorded, feat!: absence propagates and drops the row, plus defined(v) #239 being closed not planned upstream.
• The projection is not made tighter than "keep a cased quantity whole" — a tighter cut
  would have to reason about masks, which is what made the first attempt wrong.

[codspeed]

Merging this PR will not alter performance

✅ 16 untouched benchmarks
⏩ 42 skipped benchmarks¹

---

_{Comparing build/pin-alpha-52 (66f9b14) with build/pin-alpha-51 (2a5b0db)}

Footnotes

1. 42 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[read-the-docs-community]

Documentation build overview

📚 lpspec | 🛠️ Build #34314123 | 📁 Comparing 66f9b14 against latest (7a5f015)

  🔍 Preview build  

10 files changed · ± 10 modified

± Modified

• about/architecture/index.html
• examples/pypsa_ladder/index.html
• examples/pypsa_ladder/rung_02_storage/index.html
• examples/pypsa_ladder/rung_03_expansion/index.html
• examples/pypsa_ladder/rung_04_ramps/index.html
• examples/pypsa_ladder/rung_05_global_constraints/index.html
• examples/pypsa_ladder/rung_06_kvl/index.html
• examples/pypsa_ladder/rung_07_commitment/index.html
• examples/pypsa_ladder/rung_08_modular_big_m/index.html
• examples/pypsa_ladder/rung_12_linearized_uc/index.html