Documentation
¶
Overview ¶
Package dwrules is the paradigm-aware DW (data-warehouse) rule family: a SEPARATE two-argument Rule family from dbrules.Rule (schema-only), mirroring how internal/core/crossrules solved "a rule needs a fact beyond the bare schema" (ADR 0029) without mutating dbrules.Rule. A dwrules.Rule reasons over db.Schema PLUS the paradigm.Classification computed by internal/core/paradigm — both neutral inputs, so this package imports ONLY internal/core/db, internal/core/findings, internal/core/paradigm and internal/core/surface, never a provider, never a sensor (ADR 0033).
S1 shipped this package as an inert skeleton (All() empty) so RunWith could prove the merge mechanism before any real rule existed. S2 fills it with the star-schema + SCD family — DW-001 (fact without a dimension FK), DW-002 (dimension without a surrogate key), DW-005 (facts present, no time dimension), DW-010 (SCD-2 without a currency index) and DW-011 (mixed SCD strategies) — without touching that seam. DW-021 (columnar index, S3) and DW-020 (partitioning, S4) are NOT here yet and remain declared-not-covered.
Every rule in this package is SURFACE-only (ADR 0017). A warehouse-modelling choice is a design judgment, not a structurally undeniable defect: an intentional degenerate fact, a natural-key dimension carried over from a source system, or a deliberately mixed SCD policy are all legitimate. So a DW rule states the observed shape and the agent decides — it never affirms.
Every rule also reads table ROLES from the Classification and never re-derives them. A rule that re-guessed roles would fork the S1 heuristic and drift from it (locked decision A5).
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func OwnedCategories ¶
func OwnedCategories() []string
OwnedCategories are the baseline Item categories the DW rules produce — one per rule, in All() order. The DB sensor appends them to its own OwnedCategories() (design §2f): DW rules run INSIDE the sensor, so their categories are the SENSOR's, unlike crossrules, whose categories are unioned in the MCP adapter because the cross runs outside any sensor.
A category that is emitted but NOT declared here can never be baselined or pruned (ADR 0019) — it is the one way to corrupt an existing baseline — so this list is test-locked against All() rather than maintained by memory.
func Run ¶
func Run(s *db.Schema, cls *paradigm.Classification) ([]findings.Finding, []findings.SurfaceItem)
Run executes the production rule set (All()) over the schema and its classification.
func RunWith ¶
func RunWith(s *db.Schema, cls *paradigm.Classification, rules []Rule) ([]findings.Finding, []findings.SurfaceItem)
RunWith executes an EXPLICIT rule set over the schema and classification — the injectable core of Run. Production calls Run (which passes All()); the seam-gate test passes an empty/nil set so it proves the MERGE mechanism (that appending the runner's output leaves the caller's result byte-identical) independently of which rules All() happens to hold — otherwise the first real rule would put the seam gate red. A nil schema OR a nil classification yields nothing, mirroring dbrules.Run/ crossrules.RunWith — unreachable via the production sensor today (it always builds a real Classification), but forward-safety for S2, where every real dwrules.Rule dereferences cls (SUGGESTION, S1 review ledger).
Types ¶
type Rule ¶
type Rule interface {
ID() string
Check(*db.Schema, *paradigm.Classification) ([]findings.Finding, []findings.SurfaceItem)
}
Rule is one deterministic check over BOTH neutral inputs — the schema AND its paradigm/role classification. It is a DISTINCT family from dbrules.Rule (schema-only): a DW rule needs the paradigm fact, so it has its own runner. Like dbrules and crossrules, a rule emits affirmations (Findings, Confidence 1.0) and/or surface (SurfaceItems); it never sees a provider — both inputs are neutral (ADR 0029/0033).
Source Files