Repository navigation
feat(analysis): Symbolic engine for formula manipulation #4
Description
Scientific Value
Parse formula AST, derive contrasts, prove BH family size, human-readable report, DOI bundle symbolic derivation.
Scope Deferred
- Formula parser ~ + * : I() (1|batch) future
- Contrast derivation k levels -> k*(k-1)/2
- Provenance hash AST
Difficulty: Very Hard
Risks: Complexity, scope creep to mixed models, security injection, R NSE
See full spec in docs/milestones/02-deferred-issues.md
Activity
feat(proofs): Agda formal verification of the numeric core (issue #1) (…
hyperpolymath commentedon Oct 16, 2026
Triage, 2026-10-15 (backlog sort). Lane: A2-dependent — now status:blocked (it lacked the label; the dependency was always real): blocked behind #2's decision, which is itself blocked behind #1's deep statistical validation. No change in substance; the label now says what the body implied.
Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe vibe@mistral.ai
docs(milestones): restore the deferred-features spec the backlog issu…
hyperpolymath commentedon Oct 16, 2026
Addressed, 2026-10-16 — in the only way the rules allow.
The full spec this issue points at is restored verbatim from the pre-rewrite history ([
docs/milestones/02-deferred-issues.md](https://github.com/hyperpolymath/MetaManifold-WebUI/blob/main/docs/milestones/02-deferred-issues.md), PR #177) — the reference had dangled since the history rewrite. The spec carries this issue's full analysis: the formula-parser scope, contrast derivation, AST-hashed provenance, the security analysis (symbol whitelisting,system/eval/parse` refusal), the acceptance criteria, and the risk register.The gate, stated here so a reader of this issue alone cannot miss it: work on this issue is held by #2's mandatory hold, in the owner's own words: do not proceed with the symbolic engine until the statistics layer is fully and deeply tested; this issue records future intent and does not authorise implementation, engine integration, experimental user-facing exposure, or a symbolic dependency added in anticipation of later approval. The hold opens only when #1 meets its acceptance criteria (independent references, simulated and representative biological data, adversarial cases, negative controls, mutation/fuzz evidence) and the owner explicitly authorises opening this stage, with links to the validation evidence. No automatic unblocking — a closed #1 or a passing suite is not sufficient.
When the gate legitimately opens, the first permitted work is #2's step 1, in order: (1) identify the concrete user problems and assess whether an engine is necessary at all against safer alternatives (numeric methods, AD, offline review); (2) compare candidate engines (the spec names
Symbolics.jlvs custom; Rterms.formulavs pure-Julia parsing is the other axis) on correctness, assumptions, licensing, offline bundling, x86-64/ARM64; (3) the design and threat model, with the constrained expression language and domain/assumption tracking; (4) validation fixtures before integration. Nothing here starts before that.Until then: this issue is blocked — recorded, specified, gated, and not built.
Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe vibe@mistral.ai
ci(hygiene): measure branch drift and enforce the two pull-request ru…