[FBumann]

Prompt: Do a pr onto specsolve main now, renaming it
And adjust other related names too.
So it's more intuitive overall

Note

The following content was generated by AI.

Piece is now Term, in _engine/terms.py. A term is linear, quadratic or constant, which retires the words "piece", "term piece", "const piece" and "constant part". The compiler's quadratic flag is now admits_quadratic. Behaviour does not change.

What was renamed

before	after
_engine/pieces.py	_engine/terms.py
Piece	Term
kinds 'term', 'quad', 'const'	'linear', 'quadratic', 'constant'
CompiledExpression.terms, .consts, .quads	.linear, .constant, .quadratic
Compiler.expression(..., quadratic=)	Compiler.expression(..., admits_quadratic=)
map_pieces	map_terms
refuse_a_piece_without_the_dims	refuse_a_term_without_the_dims
join_mul, join_quad, join_pow	times_constant, times_variable, to_the_power
_parameter_piece, _variable_piece, _solved_piece, _dual_piece, _sum_piece, _group_piece, _at_piece, _remap_piece	the same names ending in _term
translate_piece, window_piece	translate_term, window_term

• The flag was renamed after a follow-up question about the names. It is a permission, not a kind of term: it says whether the position may hold a product of two variables. With its old name, a local quadratic list shadowed it. Now the locals in product, quotient and _build_constraint use the plain names linear, quadratic and constant, like the fields of CompiledExpression. Its docstring no longer says that a constraint row cannot hold a product, which was false for a quadratic constraint.
• The flag stays because its assert checks the compiler against mathspec's is_quadratic, and declares_quadratic also decides which constraints the solver capability check sees.
• Lists of frames that were waiting to be concatenated, which were also called pieces, are now frames (and aligned in readback).
• The docstrings, the architecture page (node table, module table and glossary), verification.md and the test names and prose use the new words.
• One user-visible string changed. The a constant summed beside a term is accepted by check and the linopy lane, refused by the engine #1137 refusal now says "a constant term of the expression" instead of "a constant part". The tests match on specsolve cannot build and Declare the parameter over, and both phrases are unchanged.

Verification

On dd1315d, in a uv environment that mirrors the pixi dev and docs features:

• ruff check . and ruff format --check . pass.
• pyrefly check reports 0 errors. On 216fa32, the linopy-reference check also reported 0 errors.
• pytest -n auto: 5524 passed, 663 skipped, 4 xfailed, 4 failed. The 4 failures are the gurobi and xpress cases of test_a_row_with_no_terms_is_not_built_and_is_reported and test_a_solver_hands_back_a_vector_and_not_an_index. These solvers are not installed in this container, and the same 4 fail on unmodified main.
• tests/test_docstring_links.py passes. On 216fa32, tests/test_docs_math.py passed and zensical build --strict found no issues.
• I did not run pixi run check itself, because pixi is not installed here. I ran its parts directly.

Not done

• Unrelated uses of "piece" stay. In _strategy.py and _archive.py, "pieces" means slices of a table. piecewise and the GenX curve's "piece" are language words.
• The local name p for a term, the cval and coeff columns, constant_scalar and the __const {i}__ alias are unchanged. Renaming them would add churn and remove no concept.
• a constant summed beside a term is accepted by check and the linopy lane, refused by the engine #1137, a constant term beside a linear one under a sum along a dimension the constant does not carry, is a separate change. Mathspec's per-node dims (feat(program): each expression node states the dimensions it is read over energy-models/mathspec#899) may let the engine spread the constant term before the sum instead of refusing it.
• Departure from AGENTS.md: a pass that is only a rename normally gets sent back. This one was asked for, and it removes a concept, because "piece" and "term" were two words for one thing.

🤖 Generated with Claude Code

https://claude.ai/code/session_01WjaoQERHiPueVjahe7oKSG

[read-the-docs-community]

Documentation build overview

📚 specsolve | 🛠️ Build #35249363 | 📁 Comparing dd1315d against latest (5a78e3f)

  🔍 Preview build  

2 files changed

± development/architecture/index.html
± development/verification/index.html

[codspeed]

Merging this PR will not alter performance

✅ 104 untouched benchmarks
⏩ 307 skipped benchmarks¹

---

_{Comparing claude/determined-johnson-irmuuz (dd1315d) with main (ecce0f3)²}

Footnotes

1. 307 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. ↩

2. No successful run was found on main (5a78e3f) during the generation of this report, so ecce0f3 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩