H36 Read and verify a report

Use your current course session. Preparation and course runner · R script ↓

H36 Read and verify a report

Goal: Reproduce results after a later delivery.

Prerequisites: H35.

Report values stay bound to the original release. Integrity and business correctness are separate questions.

training_step(
  "H36",
  {
    report_before <- dataraft.metrics::dr_report_read(
      training_lake,
      "training.august.v1",
      values_only = TRUE
    )
    changed_dates <- dated_policies
    changed_dates$cancel_date[1] <- as.Date("2026-08-20")
    cohort_later <- dr_publish(
      cohort_product,
      data = changed_dates,
      to = training_lake
    )
    report_after <- dataraft.metrics::dr_report_read(
      training_lake,
      "training.august.v1",
      values_only = TRUE
    )
    stopifnot(identical(report_before, report_after))
    report_verification <- dataraft.metrics::dr_report_verify(
      training_lake,
      "training.august.v1",
      metrics = portfolio_metrics
    )
    print(report_verification)
    stopifnot(
      all(report_verification$integrity == "match"),
      all(report_verification$replay == "match")
    )
  },
  needs = c("H35"),
  enabled = training_options$lake,
  optional = TRUE
)

Expected result: Historical report values remain identical.

Task: Does successful integrity verification prove the cancellation definition is correct?