fix(language): an at that joins on a dimension it also consumes is refused, as a sum already was

[FBumann]

Prompt: Continue bumping to the head of #566

Note

The following content was generated by AI.

math-spec moves from 0.0.0-alpha.105 to c25984e, the head of energy-models/mathspec#566. at()'s joined-column check asked whether a joined dimension was produced, which the landing check already refuses, rather than whether it was consumed — so at(load, by=diag, over=rep, into=generator) loaded and landed with the joined dimension gone. sum already refused that shape.

The pin is a commit, not a tag, because math-spec#566 is open and stacked on two more unmerged PRs. This is an integration branch rather than a pin to merge on: it wants a tag once that stack lands.

Carrying the refusal is a port of the relation node API, which no consumer can tell apart.

The API that moved

Walk is now Direction, and Partition is split off as its own class rather than a direction with nothing consumed or produced. The accessors that inverted between the two operators are gone, so a node's ends are read off its direction and an At reads the direction the other way round at the four sites that care.

was	is
program.Walk	program.Direction
GroupSum.walk, At.walk	.direction
GroupSum.over, .into	.direction.consumed_dims, .produced_dims
At.over, .into	.direction.produced_dims, .consumed_dims
.joined, .relation	.direction.joined_dims, .name
partition Walk(rel, consumed, produced, joined)	Partition(rel, along, group, joined)
partition .consumed, .produced, .consumed_dims	.along, .group, .along_dim

67 call sites across seven modules in both lanes: relational/engines/polars/{compiler,relations,reindex}.py and linopy/{builder,loader,where}.py.

Coverage moved

• test_by_takes_a_relation asserts position(by=)'s new refusals. A parameter and an undeclared name now take the message every operator's by= gives; a dimension is told that by= takes the map out of it.
• test_a_call_walks_one_table_and_says_so follows "walks one table" to "reads one table".
• GEN_BUS_WALK is GEN_BUS_DIRECTION.

The thirteen generated pages are tools.gallery_math and tools.ladder re-run. The diff is one legend line per page — "the order shift walks" becoming "the order shift steps along".

Gates

Run through uv rather than pixi, which is not installed in this environment — a departure from the pixi run check default. uv sync --group dev --extra linopy --extra gurobi --extra xpress matches what pixi's default environment installs. Re-run after the rebase onto main.

gate	result
pytest -q -n auto	4016 passed, 251 skipped, 1 xfailed, 4 failed
ruff check .	All checks passed
ruff format --check .	326 files already formatted
pyrefly check	2 errors

The four failures and both type errors are one thing, and it is not this PR's: math-spec#566 also adds ExpressionComparisonNode and ArithmeticComparisonNode, which neither lane evaluates yet. They go green in the PR stacked on this one, which is where math-spec#566's own body says that evaluator belongs.

Not run: test-bench, the sweep, and the docs build. The typst gallery cases skip for want of the binary, as they do in CI.

Deliberately not done: the mask expression evaluator, which is the stacked PR. No mutation table — this PR adds no correctness guard, it moves call sites onto a renamed API.

🤖 Generated with Claude Code

https://claude.ai/code/session_01NHXujoUG77G2SoDbhau5mS

---

Generated by Claude Code

[read-the-docs-community]

Documentation build overview

📚 lpspec | 🛠️ Build #34652809 | 📁 Comparing 4fc4b9e against latest (556b558)

  🔍 Preview build  

13 files changed · ± 13 modified

± Modified

• examples/piecewise_lp/index.html
• examples/pypsa_mixed_cycling/index.html
• examples/pypsa_spill/index.html
• examples/pypsa_storage/index.html
• examples/pypsa_store/index.html
• examples/pypsa_unit_commitment/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
• and 3 more...

[codspeed]

Merging this PR will not alter performance

✅ 24 untouched benchmarks
⏩ 58 skipped benchmarks¹

---

_{Comparing claude/happy-maxwell-9ldgqa-direction (4fc4b9e) with main (556b558)}

Footnotes

1. 58 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. ↩