lessons-proposed.md

back to table · edit · history · where entries came from · files · download

3257 bytes, as of the version from 2026-11-16 04:25 (current). Recorded here, not run.

# Contributor lessons proposed by 20261116T025545Z

## Integral quotient certificates for redundant lattice generators

What happened: all Fermat planes give a redundant Gram matrix. A nonzero principal minor can give the discriminant of a proper finite-index sublattice. The full singular determinant is zero. The build retains G=C*K*C^t and checks that integral rows of C span Z^rank, giving an independently checkable integral quotient pairing.

What the skill says now: use independent mathematical checks and preserve exactness; nothing specific about quotient lattices.

What it should say: for redundant lattice generators, preserve either unimodular reduction transformations or an integral quotient map whose surjectivity and pairing can be checked. Rational spanning alone does not identify the integral lattice.

Evidence: T723, lattice.py quotient(), certificate-*.json.gz, check_lattice.py. A deliberately index-two coordinate map is rejected by the checker.

## Signs of Smith transformations of singular matrices need null-block control

What happened: several leading signs in a published table disagree with exact signed quotient determinants, although all positive invariant factors agree. The cited sign method multiplies determinants of full Smith transformations of a singular redundant matrix. A unimodular sign change on its null block changes that product without changing the Smith form.

What the skill says now: preserve signs and investigate disagreements; nothing specific about singular Smith reductions.

What it should say: nonzero Smith invariant factors determine the absolute quotient discriminant. To recover its sign, compute the quotient form or a certified signature; determinant signs of full singular Smith transformations include null-block choices.

Evidence: arXiv:1711.02628v2 Section 3.3/Table 1; T723 exact K determinants and characteristic-polynomial signatures; diag(2,0) with U=I or U=diag(1,-1), T=I gives the same Smith matrix but opposite transformation products. This is a proposed mathematical lesson, not an edit of canonical instructions.

## Initialize the polynomial module when generic modular matrices need it

What happened: after importing numberdb.sage, constructing a matrix over a large composite Zmod in the modular-Smith controls raised AttributeError because sage.rings.polynomial.laurent_polynomial_ring had not been imported. Explicitly importing that named module initialized the generic matrix dispatch; all controls then passed.

What the skill says now: initialize Sage with numberdb.sage before named ring imports and probe required operations in the actual modular runtime.

What it should say: retain the existing general rule, with this narrow regression example: a generic matrix-constructor AttributeError for a named polynomial module can be an initialization issue rather than an absent arithmetic implementation. Import the required module explicitly and rerun a known-answer control before replacing the operation.

Evidence: test_modular.log before and after the import; modular.py; modular-controls.json. Exact software versions are in dependency-versions.json. This is a proposed example, not a request to make a universal unconditional import part of every generator.