Uh oh!
There was an error while loading. Please reload this page.
feat(v1): warn on reordered constants under legacy; spec groupers - #837
Merged
FBumann merged 1 commit intoJul 16, 2026
Merged
Conversation
Two follow-ups to the §8-exact work (#831/#834), from an open-items.md audit. Gap 1 — transition-surface hole: after #831 made a pure reorder raise under v1, reordered *constant* operands (`x + array`, `-`, `*`, `/`, rhs) were still silently reindexed by label under legacy with no warning, breaking the "no silent change" guarantee (coeff and merge already warned). Thread a `warn_reorder` flag broadcast_to_coords -> _reindex_reordered_dims, set at the arithmetic const/rhs sites; legacy now emits `_legacy_const_reorder_message` (accurately: reindexed by label, not positional). Legacy result unchanged; no double-warn; construction (bounds/mask, strict=True) stays silent. Groupers — add a strict-alignment paragraph to convention.md §13: a groupby grouper aligns to the grouped dimension by §8 (reorder or set-mismatch raises, never positional), multi-key -> flat group dim + aux coords. Implementation is #830 (on master, lands here on merge). open-items.md records the §8-exact decision and the pending grouper landing. Full suite green; ruff/mypy clean. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Uh oh!
There was an error while loading. Please reload this page.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Note
The following was generated by AI.
Two follow-ups to the §8-exact work (#831/#834), found while auditing
open-items.md.Gap 1 — reordered constants didn't warn under legacy
The "no silent change" guarantee (Stage 1 in
open-items.md) requires every v1 behaviour-change to warn under legacy. After #831 made a pure reorder raise under v1, most paths warned — but reordered constant operands slipped through:coeff * x,x + expr(merge)x + array,x - array,x * array,x / array,expr <= arrayThe const/broadcast path silently reindexes a reordered constant by label under legacy (via
_reindex_reordered_dims), so no mismatch was detected and no warning fired — while v1 raises. This introduced a silent legacy→v1 divergence.Fix: a
warn_reorderflag threadedbroadcast_to_coords → _reindex_reordered_dims, set at the 6 arithmetic const/rhs call sites. When legacy conforms a reordered constant it now emits aLinopySemanticsWarning(new_legacy_const_reorder_message, accurately saying legacy reindexes by label — the const path is not positional like merge). The legacy result is unchanged (still reindex-by-label); coeff/merge keep their single existing warning (no double-warn); bounds/mask=(strict=Trueconstruction) stay silent, since therecoordsis the source of truth. Verified: every reorder path warns exactly once under legacy, raises under v1.Groupers — strict alignment now in the spec
convention.mdcovered operand alignment (§8) but said nothing about grouper alignment. Added a Groupers paragraph under §13: agroupbygrouper aligns to the grouped dimension by §8 — a differing label set or a reorder raises, never positional matching or silent reindex (#827). Also notes multi-key groupers → flatgroupdim + key aux coords (stacked MultiIndex is legacy-only, §11).The implementation is #830 (check-and-raise), already on
master; it reaches this branch when master merges in.open-items.mdnow tracks that landing, records the §8-exact decision, and notes the reorder→raise transition-surface additions.Full suite green; ruff/mypy clean.