perf: partition elementary_test_results and test_result_rows on created_at

[tlangton3]

What

Two changes, both one-liners:

1. partition_by=elementary.get_partition_by() on elementary_test_results and test_result_rows — the package's existing created_at default, matching how dbt_run_results and dbt_invocations already declare it.
2. run_results.created_at added to the model_run_results view's select list.

Why

These are the two largest tables the package creates, both append-only, both read with a days_back lookback — and neither declares partition_by, so on BigQuery every read scans all history. On one warehouse they are 26M rows / 41GB and 12M rows / 13GB.

created_at rather than detected_at because it is each table's declared meta.timestamp_column, and because it is monotonic with insertion: insert_rows stamps it from edr_current_timestamp() at insert and never reads it from the row payload, so on an append-only table writes always land in the newest partition. A business timestamp could be backdated and scatter writes across old partitions.

The view change

model_run_results selects an explicit column list and omits created_at, so consumers of the view have no prunable column at all: its other timestamps (generated_at, execute_completed_at) are stored as strings, so any predicate on them needs a cast, and a cast defeats pruning. Adding the column is a prerequisite for fixing the CLI's get_models_runs, which reads this view and currently scans the whole table on every report.

The view is rebuilt on every run and this only adds a column, so nothing downstream breaks.

Effect on existing installations

None automatically, and nothing breaks. dbt applies partitioning only when a table is created, so existing tables keep their current layout and their queries remain valid — they simply do not gain pruning until the table is rebuilt.

Worth being explicit that --full-refresh is not the way to get it: these models compile to an empty table and are populated by on-run-end hooks, so a full refresh would discard all history. The package already prevents this — elementary_full_refresh defaults to false — but anyone who has enabled that var should not reach for it here. The safe route is a one-off rebuild that preserves rows:

CREATE TABLE `<project>.<elementary_schema>.elementary_test_results__partitioned`
PARTITION BY timestamp_trunc(created_at, DAY)
AS SELECT * FROM `<project>.<elementary_schema>.elementary_test_results`;
-- verify row counts match, then drop the original and rename this into place

(BigQuery rejects CREATE OR REPLACE across a partitioning change, hence build-and-swap.)

Non-BigQuery adapters

No-op. default__get_partition_by returns none, and bigquery__get_partition_by also returns none when bigquery_disable_partitioning is set, so the existing opt-out is respected.

Testing

sqlfmt reports no changes on all three files. The partition column exists on both models — empty_table.sql declares created_at as a timestamp for each.

I have not run the full integration suite across all five platforms; these are config-only additions that are inert off BigQuery, and I am happy to run the BigQuery suite or make changes if you would like.

Summary by CodeRabbit

• New Features
  • Improved partitioning configuration for test result data processing.
  • Added created_at timestamps to model run results, providing more complete run metadata.

[github-actions]

👋 @tlangton3
Thank you for raising your pull request.
Please make sure to add tests and document all user-facing changes.
You can do this by editing the docs files in the elementary repository.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: d0d4b022-f7cd-49df-b632-685fdac371e5

📥 Commits

Reviewing files that changed from the base of the PR and between 9a7e2b5 and 4f7a1ee.

📒 Files selected for processing (3)

• models/edr/run_results/elementary_test_results.sql
• models/edr/run_results/model_run_results.sql
• models/edr/run_results/test_result_rows.sql

Included review availability: Your plan provides up to 4 included reviews per hour; 3 remain after this review.

---

📝 Walkthrough

Walkthrough

The run-results models now use shared partition configuration for two incremental models. The model_run_results view also exposes created_at.

Changes

Run-results model updates

Layer / File(s)	Summary
Model configuration and view output models/edr/run_results/elementary_test_results.sql, models/edr/run_results/test_result_rows.sql, models/edr/run_results/model_run_results.sql	The two incremental models now use elementary.get_partition_by(). The view now selects run_results.created_at.

Estimated code review effort: 1 (Trivial) | ~5 minutes

Suggested reviewers: haritamar

Merge Risk: ⚪ Minimal · up to 4f7a1

The PR adds BigQuery-aware partitioning to two run-results tables and exposes created_at for filtering. No concrete merge-blocking risk is identified.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the primary performance change: partitioning elementary_test_results and test_result_rows by created_at. It does not mention the related view change, but the title d…
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

_{Comment @coderabbitai help to get the list of available commands.}

[tlangton3]

Companion CLI change: elementary-data/elementary#2351. The view column added here is also what unblocks the get_models_runs fix, which is left out of that PR because it would have no prunable column until this merges.