Repository navigation
feat(unit-tests): add cute-dbt dogfooding unit_tests on dim_payers + mart_dq_summary - #290
Merged
Merged
Conversation
…mart_dq_summary Add 3 unit_tests exercising the UNION-bearing mart models in this playground: - `test_dim_payers_injects_unknown_sentinel` — verifies dim_payers' unknown-member sentinel (payer_key = -1) survives the UNION ALL with the sequenced source rows. sql given + dict expect. - `test_mart_dq_summary_combines_encounter_and_medication_metrics` — verifies mart_dq_summary's UNION ALL combining encounter + medication data-quality metrics. sql given + csv expect. - `test_mart_dq_summary_zero_quarantined_when_all_valid` — verifies zero quarantined_count and 0.00 quarantine_rate_pct when all upstream rows pass DQ validation. sql given + dict expect. All three tests use sql format for GIVEN inputs (no upstream introspection needed → pass standalone without seeded Synthea data) and vary EXPECT formats across dict + csv to demonstrate the dbt 1.8+ unit-test fixture format surface. These tests serve dual purposes: 1. Real assertions on the playground's UNION-bearing mart models. 2. Source material for cute-dbt (https://github.com/breezy-bays-labs/cute-dbt) to dogfood its unit-test explorer against multi-model, multi-test, UNION-arm, empty-state diff scenarios (cute-dbt#39). Tests verified locally with dbt-core 1.11.11: $ dbt test --select test_type:unit PASS=3 WARN=0 ERROR=0 SKIP=0 NO-OP=0 TOTAL=3 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
7 of 11 tasks
cmbays
added a commit
to breezy-bays-labs/cute-dbt
that referenced
this pull request
May 25, 2026
…2cb38 cmbays/dbt-playground#290 squash-merged at 602cb38 (2026-05-25 04:58Z). Update both playground fixture entries' origin_url from the feat-branch HEAD dc1b08f to the merge commit. The fixture sha256s are unchanged — the manifests themselves are identical; only the provenance pointer moves to the stable merge commit. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
cmbays
added a commit
to breezy-bays-labs/cute-dbt
that referenced
this pull request
May 25, 2026
… unit_test format coverage (#63) * #39 — feat(examples): dogfooded playground fixture + matrix CI gate + unit_test format coverage Adds the richer cute-dbt example sourced from cmbays/dbt-playground#290 — a real dbt project with synthetic Synthea healthcare data. The fixture pair captures three modified models exercising in one report: - Multi-model in-scope cascade: mart_dq_summary, dim_payers, and int_dq_quarantine__encounters are all modified. - UNION arm rendering in two distinct patterns (encounter + medication metric UNION ALL; unknown-sentinel UNION ALL). - Multi-test-per-model: mart_dq_summary carries 2 unit tests; dim_payers carries 1. - Empty-state card: int_dq_quarantine__encounters is in scope but carries no unit tests targeting it. - dbt unit_test fixture-format diversity: the 3 unit tests in the playground span sql `given` + mixed dict/csv `expect` formats. Changes: - tests/fixtures/playground-{current,baseline}.json + MANIFEST.toml provenance entries (synthetic_only=true, origin=dbt-playground, sha256, license, description). - examples/playground-report.html committed (3.6MB) rendered from the fixture pair via the cute-dbt CLI. - .github/workflows/ci.yml example-report-up-to-date refactored to a matrix (jaffle-shop + playground) with a stable aggregator job presenting the existing branch-protection check name. Adding new examples now only requires adding a matrix row. - features/unit_test_format_coverage.feature + 4 BDD scenarios asserting cute-dbt renders unit_tests authored in dict / csv / sql formats uniformly. Feature count bumped 6 → 7 in ci.yml + lefthook (atomic mirror update). - book/src/examples.md + examples/README.md updated with the new playground example. Cross-repo coordination: - cmbays/dbt-playground#290 (private) adds the 3 unit_tests this fixture pair captures. The MANIFEST.toml origin_url pins to that commit SHA for provenance audit. Follow-ups tracked but out of scope: - dbt-autofix sweep on playground for fusion compatibility (separate playground PR). - Fusion-produced cross-engine fixture in cute-dbt. - Cross-join demo model (playground has none today). Verified locally: - cargo nextest run: 324 passed, 1 skipped - cargo test --test bdd: 7 features, 36 scenarios, 202 steps all passing (4 new scenarios) - cargo fmt --check, cargo clippy --all-targets -- -D warnings: clean - resource-ref lint: 17 passed (both jaffle-shop + playground HTML) - fixture-manifest-listed: all 7 fixtures listed with matching sha256 Closes #39. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * test(zero-egress): extend resource-ref + headless gates to every committed example The PR-D commit added examples/playground-report.html but left the zero-egress audit gates hardcoded to examples/jaffle-shop-report.html. That meant the new example shipped without: - the secondary structural lint (`tests/resource_ref_lint.rs`) - the PRIMARY runtime proof (`tests/headless_zero_egress.rs`) — the load-bearing auditability test that opens the report in real Chromium with DNS denied and asserts zero `Network.requestWillBeSent` events for http/https/ws/wss. Both tests are now keyed on a single `COMMITTED_EXAMPLES` array. Adding a new examples/<name>-report.html requires only appending its filename there (same shape as the .github/workflows/ci.yml matrix added in the parent commit). The headless test loops over examples inside a single Chrome instance (fresh tab per example, separate event capture per example) so the additional runtime cost is one extra tab, not an extra Chrome launch. Verified locally: - cargo test --test resource_ref_lint: 17 passed - cargo test --test headless_zero_egress -- --ignored: 1 passed (covers both jaffle-shop AND playground in 7.74s on a single launch) Surfaced by advisor pre-PR-open audit — exactly the kind of silent audit gap PR-Cβ taught us about. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fixtures: pin playground origin_url to playground#290 merge commit 602cb38 cmbays/dbt-playground#290 squash-merged at 602cb38 (2026-05-25 04:58Z). Update both playground fixture entries' origin_url from the feat-branch HEAD dc1b08f to the merge commit. The fixture sha256s are unchanged — the manifests themselves are identical; only the provenance pointer moves to the stable merge commit. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> * fixup(tests): dedupe COMMITTED_EXAMPLES + structural empty-state assertion (Gemini disposition) Addresses 3 medium-severity findings from gemini-code-assist on PR #63. G-1 (resolved) — `COMMITTED_EXAMPLES` was duplicated in `tests/headless_zero_egress.rs` and `tests/resource_ref_lint.rs`. Moved the array to `tests/common/mod.rs` as `pub const`, along with the `example_path(filename)` helper. Both gates now reference a single source of truth — adding a new committed example only requires appending its filename in one place. G-2 + G-3 (resolved) — the `that model's section indicates zero unit tests are wired` step matched a loose union of keyword variations (`html.contains("0 unit tests wired") || .contains("No unit tests") || .contains("no unit tests")`). The assertion would pass if the copy appeared anywhere on the page, not just on the named model's card. Rewrote all 4 `Then` steps to parse the embedded `<script id="cute-dbt-data">` JSON payload via `tl` + `serde_json` and assert structurally: - "contains the unit test X" → find X across all models' `tests` arrays; stash the owning model name in `World.last_named_model`. - "that unit test names the target model X" → compare the stashed owner to X (no double regex match needed). - "contains a section for the model X" → look up the model in the payload by name; stash the name. - "that model's section indicates zero unit tests are wired" → assert the stashed model's `tests` array is empty. Per-model assertions now correctly scope to the named model, not the whole document. Verified locally: - cargo test --test bdd: 7 features, 36 scenarios, 202 steps passing - cargo test --test resource_ref_lint: 17 passed - cargo test --test headless_zero_egress -- --ignored: 1 passed (covers both jaffle-shop + playground in one Chrome launch) - cargo fmt --check, cargo clippy --all-targets -- -D warnings: clean - cargo nextest run: 324 passed, 1 skipped Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
1 task done
cmbays
added a commit
that referenced
this pull request
May 25, 2026
…s the 3 cute-dbt dogfooding tests (#291) Follow-up to #290 (cute-dbt#39 dogfooding). Surfaced visually: cmbays/cute-dbt's renderer currently shows "Fixture uses format: sql — not yet visualized" for the sql-format mocks, AND I had previously chosen sql for ALL three tests to avoid the upstream-introspection prerequisite for dict/csv. That choice defeated the format-coverage demonstration the tests were meant to provide. This PR re-authors the 3 tests so each test uses a different fixture format end-to-end: | Test | Given | Expect | Demonstrates | |---|---|---|---| | test_dim_payers_injects_unknown_sentinel | dict | dict | Key-value mock for ~22 columns | | test_mart_dq_summary_combines_encounter_and_medication_metrics | csv | csv | Tabular boolean grid | | test_mart_dq_summary_zero_quarantined_when_all_valid | sql | dict | Inline SELECT with explicit casts | The dict and csv `given` blocks introspect the upstream model's schema (to NULL-fill unspecified columns), which requires the upstream relation to exist in the warehouse. Adds a `## Unit tests (cute-dbt format coverage)` section to `dbt_project/README.md` documenting the persistent-DuckDB target prerequisite + the exact 3-command workflow to run them. All 3 tests pass locally against dbt-core 1.11.11 with a persistent DuckDB target seeded via `run-operation load_synthea_sources`: $ dbt test --select "test_type:unit" --target unit_test PASS=3 WARN=0 ERROR=0 SKIP=0 NO-OP=0 TOTAL=3 Closes the cute-dbt-side gap that motivated cute-dbt#66 (fixture format rendering for csv + sql). cute-dbt's follow-up PR ships the matching renderer fix + regenerated playground fixture against this PR's merge commit. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds 3 unit_tests exercising the playground's UNION-bearing mart
models. Doubles as source material for cute-dbt
dogfooding (see cute-dbt#39)
— the manifest output from this branch is the input fixture for
cute-dbt's first richer example report against real dbt models.
Changes
unit_tests:block todbt_project/models/marts/core/_core__models.ymlwith
test_dim_payers_injects_unknown_sentinel.unit_tests:block todbt_project/models/marts/analytics/_analytics__models.ymlwith
test_mart_dq_summary_combines_encounter_and_medication_metricsand
test_mart_dq_summary_zero_quarantined_when_all_valid.Total: +154 lines across 2 YAML files. No SQL model changes, no
new fixtures, no Python.
Related Issues
Related to cute-dbt#39 (cross-repo fixture-source dogfooding).
Type of Change
Testing
All 3 tests verified locally with dbt-core 1.11.11 + dbt-duckdb 1.10.1:
All tests use
format: sqlforgiveninputs so they pass standalonewithout requiring seeded Synthea source data (no
dbt buildofstaging models required).
expectformats vary acrossdict+csvto demonstrate the dbt 1.8+ unit-test fixture-format surface.
Specific Test Cases
test_dim_payers_injects_unknown_sentinel— dim_payersUNION-ALLs an unknown-member sentinel row (
payer_key = -1) withthe sequenced source rows. Verifies the sentinel survives the
union and source rows are surrogate-keyed starting at 1.
test_mart_dq_summary_combines_encounter_and_medication_metrics— mart_dq_summary UNION-ALLs metrics from
stg_synthea__encountersand
stg_synthea__medications. Mocks 2 encounter rows (oneinvalid) + 2 medication rows (both invalid) and asserts per-entity
counts + quarantine rate.
test_mart_dq_summary_zero_quarantined_when_all_valid— theinverse: zero quarantined_count + 0.00 quarantine_rate_pct when
every upstream row passes DQ validation. Uses dict expect to
sidestep non-deterministic
_generated_at(current_timestamp).Manual Testing
UI-specific test items (Chrome/Firefox/Safari/localStorage) are N/A —
this PR adds dbt YAML only, no application code changes.
Notes
2.0.0-preview currently blocks parse on the playground's deprecated
test-args YAML (~166 unrelated occurrences); migrating those is a
separate dbt-autofix sweep tracked outside this PR.
givenmocks areself-contained inline SELECTs —
dbt build --emptyis NOTrequired first. Run
dbt test --select test_type:unitafterdbt depsand a singledbt parse.🤖 Generated with Claude Code