Options must affect the requested computation

An option can change the mathematical input, select a computational method, or control storage and checking. Knowing which effect it has lets you decide whether two results should agree. This guide follows those choices from finite encoding through algebra and module queries; the finite-encoding introduction describes the underlying object whose spaces and maps those operations use.

Use the options object belonging to the task. The ingestion guide describes filtration construction, grid representation, field selection and JSON replay. The following contracts apply to the other option owners. These option types are available from TamerOp.Options and the curated TamerOp.Advanced surface.

Finite encoding

EncodingOptions(poset_kind=:dense) selects a dense finite poset through both direct owner calls and workflow calls. Omitting the explicit poset_kind keyword uses the value in the options object; an explicit keyword overrides it. Geometric poset-only operations do not use a coefficient field. Operations that produce a fringe or module use opts.field as the target field.

A Zn flange call without explicit options preserves the flange's field. Passing an EncodingOptions object explicitly selects its field, whose default is QQ. Changing characteristic can change ranks and homology; it is not an invariance claim. Rational coefficients whose denominators vanish in the target field are rejected by the field conversion contract.

strict_eps belongs to general PL geometry. The default nothing certifies strict feasibility over the rationals, including arbitrarily narrow cells. An explicit positive value selects a fixed feasibility margin and can omit cells; point membership still uses the original strict boundaries. Result provenance records feasibility=:exact_rational or :fixed_margin. Zn and axis-aligned box encoders reject a nondefault value instead of accepting an ineffective setting. Derived Zn box computations honor their encoding field and region budget, and reject controls that do not apply to that construction.

Derived computations

Use DerivedFunctorOptions to select the actual computational model. A fixed projective or injective computation rejects an incompatible model or canonical coordinate choice. canon=:auto resolves to the model actually used; :none requests native coordinates. Unified Ext retains its projective/injective coordinate comparison capability. Tor has its own :first/:second models and rejects inapplicable canonical choices.

For example, use ExtLongExactSequenceFirst(...; opts=DerivedFunctorOptions(...)) for the canonical keyword surface. Long exact sequences and derived Kan computations validate their fixed models before building or consulting caches. Negative requested degrees are rejected.

ResolutionOptions(check=true) validates structural resolution data, including cached results, whether or not minimal=true. The structural checks cover term alignment, generator data and differential compositions. They are not independent proofs of exactness of every returned resolution. The finite-poset builders construct minimal resolutions; minimal=true additionally requests the minimality assertion when checking is enabled. check=false explicitly skips those diagnostics, and negative length bounds are rejected.

Use ExtDoubleComplex(M, N; maxlen=...) and ExtSpectralSequence(M, N; maxlen=...) directly. The first constructs the specified bounded bicomplex; the second requires complete resolutions for its Ext abutment.

Module construction and queries

ModuleOptions(cache=cover_cache) supplies cover information to module construction and structure-map queries. The cache must belong to the represented poset. Supplied caches and conflicting cache arguments are validated even for identity maps and empty batches.

check_sizes=false is a trusted constructor-only opt-out. Structure-map queries always validate indices and comparability and reject that opt-out. Default constructor validation also covers reused map storage.

Finite-fringe constructors accept scalar, checking and storage arguments directly.