Concept Taxonomy

This taxonomy organizes the 130 mountainash-rules concepts into 9 categories aligned with the package's architecture layers. Updated 2026-09-02 (source 7d0e3dc) to add two categories — Hit Policies and Batch Evaluation — for capabilities added since the original 90-concept graph (41d5846, 2026-06-03).

Categories

FOUND — Foundation Concepts

Prerequisites and external knowledge: ternary logic, sentinel value patterns, match strategy concepts, Pydantic model validation, vectorized evaluation principles, backend-agnostic design patterns, and the mountainash expression and relation libraries that the rules engine builds upon.

DIM — Dimension Model

The dimension metadata layer: the MatchStrategy enum with its 12 strategy types (EXACT, EXACT_KEY, NOT_EQUAL, RANGE, GREATER_THAN, LESS_THAN, PREFIX, SUFFIX, CONTAINS, REGEX, CONTEXT_REGEX, SET_MEMBERSHIP, SET_EXCLUSION — note REGEX split into per-row REGEX and metadata-literal CONTEXT_REGEX), the DimensionRole enum (CONSTRAINT vs CONTEXT_KEY), the DataType enum with temporal (date/datetime) sentinels, bool-dimension ternary comparison, the Dimension Pydantic model with field resolution and validation, the DimensionsMetadata collection, and its YAML round-trip serialization.

EXPR — Expression Rules Engine

The single-pass filter evaluation engine: DimensionCompiler that translates Dimension metadata into backend-agnostic expression templates (including the shared in-band set-wildcard sentinel and its normalization), the ExpressionRulesEngine class with its construction paths, and the single-pass evaluation pipeline (context binding, dimension expression application, survival computation, specificity scoring, rank assignment, column cleanup).

POLICY — Hit Policies

The selection-semantics layer applied to survivor sets: the HitPolicy enum (collect, unique, first, priority, any, rule_order), SelectionInfo (what a result needs to re-apply a policy post-hoc), HitPolicyViolationError for UNIQUE/ANY breaches, ordering-key computation, and cardinality application (truncating to a single survivor). Configurable at the table level via DimensionsMetadata's hit-policy fields, or per evaluation call.

RESULT — Expression Engine Results

Result accessors for the ExpressionRulesEngine: RuleResult wrapper with survivors, best_match, count, active_dimensions, the per-rule explain method, at_least filtering, select (re-applying a hit policy post-hoc), and the engine-level ExplainResult / explain() for scoring every rule against a context without survival filtering.

BATCH — Batch Evaluation

Scoring many contexts against the rules table in a single vectorized pass: evaluate_batch(), context preparation and backend conforming, the cross-join evaluation strategy, per-context-id ranking (window-function free), opt-in chunked evaluation, and the BatchRuleResult wrapper with its per-context accessors.

ACCUM — Accumulator Engine

The combinatorial lattice-building engine: AccumulatorCompiler with compatible/coalesce/NA-flag expressions (including set-membership/set-exclusion support), AccumulatorEngine with its build process (partition, prime assignment, anchor creation, level expansion with canonical ordering, frontier filter for dominated combination removal, apply-phase caching), and prime number encoding for unique combination identity with an explicit per-partition rule cap.

LATT — Accumulator Lattice & Results

The lattice data structure and accumulator result accessors: Lattice class with combinations, partition key, coalesced columns, NA flags, combination depth, is_composed, and save/load snapshot persistence; AccumulatorResult extending RuleResult with accumulated aggregates (sum, min, max, product), provenance (prime products), and depth accessors.

SUPP — Supporting Modules

Cross-cutting concerns: the Aggregate Pydantic model and its four operations, partition key filtering for CONTEXT_KEY dimensions, build_all for multi-partition lattice construction, apply_auto for automatic partition-key-based lattice selection, and the LatticeIndex ternary-partition router (with its AmbiguousPartitionError and EXACT_KEY-powered routing) that generalizes partition selection beyond apply_auto's single-lattice case.

Taxonomy Summary Table

TaxonomyID	Category Name	Concept Range	Count
FOUND	Foundation Concepts	1-10	10
DIM	Dimension Model	11-30, 91-96	26
EXPR	Expression Rules Engine	31-48, 98-99	20
POLICY	Hit Policies	97, 100-108	10
RESULT	Expression Engine Results	49-58, 109-111	13
BATCH	Batch Evaluation	112-119	8
ACCUM	Accumulator Engine	59-76, 120-123	22
LATT	Accumulator Lattice & Results	77-86, 124-127	14
SUPP	Supporting Modules	87-90, 128-130	7