feat(program): the windowing verdict says which way a row reads and what only data or the driver can resolve

[FBumann]

Prompt: "Can you just do a stacked PR fixing those?" — the three points lpspec's driver raised against #248.

Note

The following content was generated by AI.

Three changes to Separability, read from the consumer that will call it (lpspec.strategy.EachWindow). The shape and placement of #248 are unchanged.

1. halo becomes behind and ahead. A translation reads behind for a positive offset (t - offset) and ahead for a negative one; a trailing window reads behind by width - 1. A rolling-horizon driver supplies the two differently — lookahead is rows it solves and does not keep, history is rows it carries from the window before — and abs(reach) could not say which it was in.
2. Reasons are typed by what a driver can do about them, as three mappings rather than one. coupled (a sum in a constraint, a grouping, a wrap, a set, an at()) no window satisfies. undecided names the parameter or lookup whose values set the reach — a named offset or width, a partition — and a driver holding the data computes it; the same shape as coupled refused PyPSA's link delay, which rolls routinely. restarts names a position() a window restarts, which a rolling horizon seeding its opening state means. windowable is false while anything is coupled or undecided; a restart does not count against it.
3. The claim is narrowed to what the walk certifies: every row builds complete inside some window. It cannot say the windowed answer is the whole-horizon one — a store carried over one row windows cleanly and a rolling solve of it is still a different answer — and the docstrings and reading.md said it could.

Two more, from the follow-up questions ("can we say what would make it windowable?", "isn't separability also what a scenario sweep asks?"):

4. Each coupled entry names the change that would lift it, after a dash: a horizon total becomes a rolling sum_back(within=n), a wrap an opening-state seed at position(t) == 0, a grouping is windowed along the dimension it groups into, a set needs a window holding it whole. Named and not applied — every one changes what the model means, and math-spec's Advice already takes that line. at() moves from coupled to undecided, naming the lookup whose values set its reach.
5. independent answers the other driver: whether each coordinate builds on its own — windowable, reading nothing behind or ahead, counting no position. A scenario sweep asks it before solving one coordinate per slice, and it is what licenses solving slices in any order or at once. A restart counts against it, unlike for windowable: with one coordinate per slice a position() holds everywhere, which changes what the mask means.

Verified

pixi run lint (lefthook: prettier, ruff), pixi run typecheck (0 errors), pixi run test: 931 passed. reading.md gains one executed line (the remedy) and the page test's claim count follows. Not run: mkdocs build --strict, compile-tex.

Coverage moved: the six coupling cases split into two structural (coupled), three data-decided (undecided, asserting the name), and one restart (asserting it stays windowable); a negative-offset case, the at() case, four independent cases and the remedy wording are new.

🤖 Generated with Claude Code

https://claude.ai/code/session_01FPBsP3UM6esCBREVmjMEhZ

[read-the-docs-community]

Documentation build overview

📚 math-spec | 🛠️ Build #34348039 | 📁 Comparing ffc6ae5 against latest (711a917)

  🔍 Preview build  

20 files changed · + 1 added · ± 19 modified

+ Added

• reference/math_spec/where_parser/index.html

± Modified

• index.html
• CHANGELOG/index.html
• about/ceiling/index.html
• examples/pypsa/index.html
• reference/typeset/index.html
• reference/language/expressions/index.html
• reference/language/reading/index.html
• reference/math_spec/boundedness/index.html
• reference/math_spec/dimensions/index.html
• reference/math_spec/errors/index.html
• and 9 more...

[FBumann]

Superseded: the branch is rebased onto main (alpha.67) carrying #248's commit, since #248 predates the Mask move and GitHub refuses to retarget a stacked PR. New PR follows.