DATE_TRUNC: auto-cast unknown-typed time-dimension expressions

[AivanF]

A SlayerModel time dimension whose sql is a bare string literal (Column(sql="'2025-12-01'", type=TIMESTAMP)) — a real pattern for synthetic/reference-data time spines — generated SQL like DATE_TRUNC('month', '2025-12-01'), which Postgres rejects at execution time with function date_trunc(unknown, unknown) is not unique because the second argument carries no resolvable type and there's no single overload to pick.

This PR fixes the SQL generator's single _build_date_trunc helper to wrap non-column, non-already-cast operands in CAST(... AS TIMESTAMP) before passing them to DATE_TRUNC — bare column references stay untouched (their live DB type is already known, and forcing a cast there could silently strip TIMESTAMPTZ to TIMESTAMP).

Because all DATE_TRUNC emission sites already funnel through that one helper, the fix simultaneously covers regular SELECTs, shifted CTEs, the compare-period builder, and time-bucket window functions. Added a parametrized regression test across postgres / mysql / bigquery / duckdb / snowflake asserting both directions: bare-literal time dims get wrapped, bare columns don't.

Summary by CodeRabbit

• Bug Fixes

  • Enhanced SQL query reliability for date truncation operations across supported database platforms (PostgreSQL, MySQL, BigQuery, DuckDB, Snowflake).

• Tests

  • Added test coverage for date truncation functionality across multiple database systems.

[coderabbitai]

📝 Walkthrough

Walkthrough

SQLGenerator._build_date_trunc now enforces stricter typing by wrapping non-Column, non-Cast expressions in CAST(... AS TIMESTAMP) before generating DATE_TRUNC AST to resolve Postgres planner ambiguity. Docstring expanded and multi-dialect test added.

Changes

DATE_TRUNC Type Casting

Layer / File(s)	Summary
DATE_TRUNC type casting logic slayer/sql/generator.py	Docstring clarifies the cast behavior: non-Column/non-Cast inputs are cast to TIMESTAMP before DATE_TRUNC to disambiguate "unknown" typed operands; bare columns and existing casts are left unchanged. A new guard in _build_date_trunc wraps col_expr with CAST(... AS TIMESTAMP) when neither condition is met.
Multi-dialect test validation tests/test_sql_generator.py	Parameterized regression test across five dialects (postgres, mysql, bigquery, duckdb, snowflake) verifies that bare literal date strings are cast to TIMESTAMP while bare timestamp column references are not cast, using substring assertions on generated SQL.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐰 A rabbit hops through types with care,
Casting dates to timestamp's pair,
Unknown strings now safely caught,
In TIMESTAMP's protective thought,
Five dialects tested, none left bare!

🚥 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 and specifically describes the main change: auto-casting of unknown-typed time-dimension expressions in DATE_TRUNC. It directly matches the primary objective and the core code modification.
Docstring Coverage	✅ Passed	Docstring coverage is 80.00% which is sufficient. The required threshold is 80.00%.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch aivan/2026-05-22-DATE_TRUNC-typing

---

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

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[coderabbitai]

🧹 Nitpick comments (1)

tests/test_sql_generator.py (1)

1926-1926: ⚡ Quick win

Narrow the no-CAST assertion to the DATE_TRUNC operand

assert "CAST(" not in sql.upper() is too broad and can fail on unrelated CASTs added elsewhere in query generation. Assert specifically that the bare-column operand path is uncast (e.g., no CAST(created_at AS ...) near the truncation expression).

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/test_sql_generator.py` at line 1926, The test currently asserts
globally that no "CAST(" appears in sql (assert "CAST(" not in sql.upper()),
which is too broad; instead locate the DATE_TRUNC/DATETRUNC expression in the
generated sql and assert that its operand is the bare column (e.g., not wrapped
in a CAST). Update the assertion to parse the sql string (using a regex) to find
the DATE_TRUNC/DATETRUNC call and then assert that there is no pattern like
CAST\s*\(\s*created_at\s+AS\b (case-insensitive) inside that truncation
expression; operate on the variable sql and keep the check case-insensitive so
unrelated CASTs elsewhere are allowed.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@tests/test_sql_generator.py`:
- Line 1926: The test currently asserts globally that no "CAST(" appears in sql
(assert "CAST(" not in sql.upper()), which is too broad; instead locate the
DATE_TRUNC/DATETRUNC expression in the generated sql and assert that its operand
is the bare column (e.g., not wrapped in a CAST). Update the assertion to parse
the sql string (using a regex) to find the DATE_TRUNC/DATETRUNC call and then
assert that there is no pattern like CAST\s*\(\s*created_at\s+AS\b
(case-insensitive) inside that truncation expression; operate on the variable
sql and keep the check case-insensitive so unrelated CASTs elsewhere are
allowed.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 247ee27c-e68a-4dae-9aff-57c8c083d2ad

📥 Commits

Reviewing files that changed from the base of the PR and between a8b8f2c and aefc25f.

📒 Files selected for processing (2)

• slayer/sql/generator.py
• tests/test_sql_generator.py