Skip to content

docs: close the 3.0.0 loose ends — a test, the notebook, C-99's wording, #237 - #302

Merged
Polichinel merged 1 commit into
developmentfrom
docs/3.0.0-loose-ends
Nov 2, 2026
Merged

Polichinel merged 1 commit into
developmentfrom
docs/3.0.0-loose-ends

Conversation

@Polichinel

Copy link
Copy Markdown
Contributor

Loose ends from a falsification audit of "everything in this repo is done and published". No package code changes, so no release.

  • tests/test_reconciliation_scale.py: it asserted the old 2.x rule (empty grids stay zero) through ReconciliationModule, over countries with no empty-grid draw, so the assertion was empty. It now checks that every draw sums to its country, on the countries that actually have empty draws. It fails when the donor fill is disabled; I checked that with a mutation.
  • notebooks/03_reconciliation.ipynb: a new §4b shows donor fill (filled_draws, METHOD_DONOR_FILL). §2 now says the all-zero edge belongs to the low-level function only. The whole notebook executes.
  • C-99: "0 unplaced" holds by construction, so it isn't evidence the mass lands in the right cells. The entry also notes that adoption in the forecast lane is pending (views-datafactory#527, views-models#663).
  • GOVERNANCE.md: the conformance floor is not a safe dependency pin; floor at the release that introduced the behaviour you rely on. Closes CONFORMANCE_FLOOR reads as a safe dependency floor, but estimator semantics changed above it #237.

Gate: ruff, format, mypy (also on 3.11), lint-imports, 100% coverage, validate_docs, check_doc_refs, check_arch_tree and the examples all pass.

Closes #237

🤖 Generated with Claude Code

https://claude.ai/code/session_01Q77JMFJbxEabjGdAh8jHWA

…ng, #237

Found by a falsification audit of "everything in this repo is done":

- tests/test_reconciliation_scale.py asserted the retired 2.x rule (empty
  grids stay zero) through ReconciliationModule, over a selection that held
  no empty-grid draw, so it asserted nothing. It now checks the 3.0.0 rule
  (every draw sums to its country) on countries that do have such draws,
  and fails when the fill is disabled.
- notebooks/03_reconciliation.ipynb never mentioned donor fill. It gains a
  short §4b showing a filled draw, and §2 says the all-zero edge is the
  low-level function's, not the module's.
- C-99 said "verified (0 unplaced)". It now says that figure holds by
  construction, and that adoption in the forecast lane is still pending
  (views-datafactory#527, views-models#663).
- GOVERNANCE.md: the conformance floor is not a safe dependency pin. Closes #237.

No package code changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Q77JMFJbxEabjGdAh8jHWA
@Polichinel
Polichinel merged commit 176ef2d into development Nov 2, 2026
11 checks passed
@Polichinel
Polichinel deleted the docs/3.0.0-loose-ends branch November 2, 2026 02:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

CONFORMANCE_FLOOR reads as a safe dependency floor, but estimator semantics changed above it

1 participant