fix: [AI-7675] accept `reused`/unknown run statuses in run_results v1-v5 parsers by Sourabhchrs93 · Pull Request #108 · AltimateAI/datapilot-cli

Sourabhchrs93 · 2026-07-22T10:58:26Z

Summary

Closes the residual of AI-7675 finding #2: the reused-status shim shipped for run_results v6 in 0.3.2 (AI-7435, PR #106) was never applied to v1–v5. Any run status outside success/error/skipped still raises a ValidationError on those schema versions, and the backend ingestion worker silently drops the entire run_results.json (batch stays green — same silent-drop failure mode as AI-7435/AI-7856).

Changes

Exact mirror of the v6 fix, applied to run_results_v{1..5}.py:

Status becomes str, Enum, gains reused, and gains the _missing_ forward-compat fallback — unknown future dbt statuses parse as real members instead of failing validation.
RunResultOutput.status union reordered to Union[Status1, Status2, Status] so test/freshness statuses (pass/fail/warn/runtime error) keep resolving via the strict enums; the permissive run Status is tried last.
Freshness-specific enums (Status3/Status4 in v3/v4) untouched — matching the v6 fix scope.

Tests

New tests/test_vendor/test_run_results_pre_v6.py, parameterized over v1–v5, mirroring test_run_results_v6.py:

PYTHONPATH=src pytest tests/test_vendor/ -q
36 passed

Also verified against a real sunrun v4 production artifact (run_498571806 from S3): parses to RunResultsV4 with all results intact. Adjacent consumers (tests/clients/altimate/test_utils.py, tests/core/platform/dbt/test_artifact_loaders.py) pass. ruff clean on the new test; zero new violations in the generated vendor files (pre-existing UP007 counts unchanged).

Scope notes

v1/v2/v3 are included alongside v4/v5: identical generated code, identical latent bug — patching only two of five sibling files perpetuates the whack-a-mole this ticket calls out.
This is the point-fix track of AI-7675. The durable dict-first lenient reader (finding feat: datapilot initial commit #1 + design consolidation) is a separate work item and is unaffected by this change.
After release + backend version bump, v1–v5 tenants get the same protection docusign (v6) already has.

🤖 Generated with Claude Code

…-v5 parsers Mirror of the v6 `Status` shim (AI-7435, PR #106) for the pre-v6 schemas, closing the residual of AI-7675 finding #2: any run status outside `success`/`error`/`skipped` (e.g. dbt 2.0 `reused`, or any future status) raised a `ValidationError` in run_results v1-v5 and the ingestion worker silently dropped the entire `run_results.json`. - Make `Status` a `str, Enum`, add `reused`, and add the `_missing_` forward-compat fallback so unknown statuses surface as real members instead of failing validation (identical to the v6 fix). - Reorder `RunResultOutput.status` to `Union[Status1, Status2, Status]` so test/freshness statuses (`pass`/`fail`/`warn`/`runtime error`) keep resolving via the strict enums and the permissive run `Status` is tried last (same reordering as v6). - Add parameterized regression tests over v1-v5 mirroring `test_run_results_v6.py`, including full-file `parse_run_results`. Freshness-specific enums (`Status3`/`Status4` in v3/v4) are untouched, matching the v6 fix scope. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

kilo-code-bot · 2026-07-22T11:11:38Z

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (6 files)

src/vendor/dbt_artifacts_parser/parsers/run_results/run_results_v1.py
src/vendor/dbt_artifacts_parser/parsers/run_results/run_results_v2.py
src/vendor/dbt_artifacts_parser/parsers/run_results/run_results_v3.py
src/vendor/dbt_artifacts_parser/parsers/run_results/run_results_v4.py
src/vendor/dbt_artifacts_parser/parsers/run_results/run_results_v5.py
tests/test_vendor/test_run_results_pre_v6.py

The Status shim (str, Enum base + reused member + _missing_ forward-compat fallback) and the Union[Status1, Status2, Status] reordering are applied identically across all five pre-v6 schema files and match the approved, production v6 reference exactly. The freshness-specific Status3/Status4 enums in v3/v4 are correctly left untouched. The new parameterized test faithfully mirrors test_run_results_v6.py and uses correct schema-version URLs for routing. No downstream consumer in this repo relies on the run-status enum type/identity, so the reordering carries no regression risk.

_{Reviewed by glm-5.2 · Input: 35.7K · Output: 8.4K · Cached: 432.3K}

…#110) `tests/test_vendor/test_run_results_pre_v6.py` (PR #108) was not formatted to the repo's `black` line-length (140), which fails the `pre-commit` gate inside `release.sh`/`tox` and blocked the v0.3.4 PyPI publish in the tag-release workflow. Formatting-only change; tests unchanged semantically (36 pass). This PR carries the `version-bump` label so its merge re-triggers tag-release and publishes 0.3.4 (the previous unpublished v0.3.4 tag is deleted so the workflow can recreate it at this commit). Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

…sources v1-v3 (#114) The dbt Fusion engine serializes source-freshness `status` with its Rust variant names -- `"Pass"` / `"Warn"` / `"Error"` -- while dbt-core and the published `sources/v3.json` schema Fusion itself stamps into the artifact both use the lowercase forms. Every result row therefore failed both members of the `results` union, the whole `sources.json` raised a `ValidationError`, and the ingestion worker silently dropped it. Nine of harvestgroup's ten production environments have zero source-freshness rows as a result. This is the residual of the AI-7675 work: PR #106/#108 added the `_missing_` forward-compat shim to the `run_results` status enums but explicitly left the freshness enums, and the `sources` parsers entirely, untouched. - Make `Status1` a `str, Enum` whose `_missing_` case-folds to the canonical lowercase member first, then falls back to the same forward-compat pseudo-member used by the `run_results` shim for unknown statuses. - Case-folding rather than adding PascalCase members is deliberate: the extractor persists `status.value`, and every dbt-core-backed tenant already writes lowercase into the same table. - Leave the runtime-error-only `Status` enum strict. Fusion has no runtime-error variant, and loosening it would let a row missing a required field fall silently into the field-less branch instead of erroring. - Apply to v1-v3 rather than v3 alone, mirroring how #108 had to follow #106 across the older schemas. Verified against 432 real production artifacts spanning all nine Fusion environments and 13 Fusion builds (preview.190 -> .210), run through parse AND the worker's own `extract_sources`: 432/432 parsed, 10,893 freshness rows extracted, every capitalized input landing as its lowercase counterpart. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

mdesmet approved these changes Jul 22, 2026

View reviewed changes

Sourabhchrs93 merged commit ab70889 into main Jul 22, 2026
31 of 33 checks passed

Sourabhchrs93 mentioned this pull request Jul 22, 2026

chore: [AI-7675] fix test formatting blocking the 0.3.4 release #110

Merged

shreyastelkar mentioned this pull request Aug 20, 2026

fix: [AI-8671] accept dbt Fusion's capitalized freshness statuses in sources v1-v3 #114

Merged

Navigation Menu

Uh oh!

fix: [AI-7675] accept `reused`/unknown run statuses in run_results v1-v5 parsers - #108

fix: [AI-7675] accept `reused`/unknown run statuses in run_results v1-v5 parsers#108
Sourabhchrs93 merged 1 commit into
mainfrom
fix/ai-7675-run-results-v1-v5-status-shim

Sourabhchrs93 commented Jul 22, 2026 •

edited by atlassian Bot

Loading

Uh oh!

kilo-code-bot Bot commented Jul 22, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

Sourabhchrs93 commented Jul 22, 2026 • edited by atlassian Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

Tests

Scope notes

Uh oh!

kilo-code-bot Bot commented Jul 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code Review Summary

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Sourabhchrs93 commented Jul 22, 2026 •

edited by atlassian Bot

Loading

kilo-code-bot Bot commented Jul 22, 2026 •

edited

Loading