dr_report_verify()

Checks stored numerical hashes, then optionally recalculates each measure on its exact input release. Supply original approved metric objects: stored code is never evaluated. Verification writes neither reports nor metric registry entries. A result match demonstrates this execution only, not universal reproducibility of external services, SQL random functions or custom code.

Usage

dr_report_verify(lake, id, metrics = NULL, strict_environment = TRUE)

Arguments

lake: Connected lake, configuration or local lake folder.
id: Immutable report id.
metrics: Original metric object or list of metric objects. Definitions must exactly match the saved hash, identity and approval. NULL checks only stored numerical integrity and explicitly reports replay as unavailable.
strict_environment: Skip replay when runtime versions, engine, locale or timezone differ. RNG state is restored separately and excluded from this comparison. With FALSE, replay can match despite environment differences.

Value

A tibble with one row per measure. integrity checks saved values; environment is match, mismatch or unavailable; replay is match, mismatch, unavailable or error. reason explains an unavailable or failed replay without exposing custom computation error messages.

Details

Runtime evidence records R and computation package versions, database engine, locale, timezone and the R random generator state immediately before execution. Replay restores that R generator state temporarily; it does not install old packages or change locale or timezone. Additional packages, external resources, database settings and SQL random generators used by custom computations remain the caller's responsibility. Reports created before runtime capture, and inputs with unsupported classes, have no automatic replay evidence.

Examples

# dr_report_verify(lake, "report.v1", metrics = original_metric)

Documentation source snapshot · View source at 513e09e ↗
Code examples are documented source examples; they are not executed in this website build.