test: combine #2855 cortex review bot with #2948 snowpark deploy --prune fixe

[sfc-gh-kolszewski]

Pre-review checklist

• If my changes add or modify user-facing interface (commands, flags, output formats), I consulted maintainers and got sign-off beforehand (how).
• I've confirmed that instructions included in README.md are still correct after my changes in the codebase.
• I've added or updated unit tests to verify correctness of my new code.
• I've added or updated integration tests to verify correctness of my new code.
• Manually tested on macOS
• Manually tested on Windows
• I've confirmed my changes are up-to-date with the target branch.
• I've described my changes in RELEASE-NOTES.md (see when and how).
• I've updated documentation if behavior changed.

Changes description

...

[github-actions]

Cortex AI E2E Review

Model: claude-opus-4-6 | Commit: 519342f | Duration: 4m 13s | Reviewed at: 2026-06-14T17:08:45Z

Summary

PR #3106 extends the snow snowpark deploy --prune flag to additionally drop procedures and functions that exist in Snowflake but are no longer declared in the project's snowflake.yml. Previously --prune only cleared stage contents. The new behavior was E2E-verified against a live Snowflake environment: stale objects are correctly dropped, declared objects are preserved, overloaded signatures are handled correctly, and the default --no-prune behavior remains unchanged.

E2E Test Results

Command	Exit Code	Result
snow snowpark deploy --help	0	Updated --prune description shown correctly
snow snowpark deploy --prune --replace (with stale objects)	1*	Prune phase succeeded: STALE_PROC and STALE_FN dropped. *Exit 1 due to pre-existing 'packages' KeyError in procedure creation (unrelated to this PR)
snow snowpark deploy --no-prune --replace (with stale objects)	1*	No prune phase executed, stale objects preserved. *Same pre-existing error
snow snowpark deploy --replace (default, no prune flag)	1*	No prune phase executed (default is --no-prune). *Same pre-existing error
snow snowpark deploy --prune --replace (overloaded function)	1*	Correctly dropped KEEP_FN(NUMBER, NUMBER) overload not declared in project, preserved KEEP_FN(VARCHAR)

*Note: Exit code 1 in all deploy cases is from a pre-existing KeyError: 'packages' in common.py:175 when handling SQL-language procedures (not Python/Snowpark). This is unrelated to the prune feature.

Breaking Changes

Behavioral change to --prune flag (opt-in, non-default):

The --prune flag now has expanded semantics. Previously it only cleared stage contents before uploading artifacts. Now it additionally drops procedures/functions from Snowflake schemas that are not declared in the project definition.

This is a behavioral extension of an existing flag, but:

• The default remains --no-prune (no change to default behavior)
• Users must explicitly pass --prune to trigger the new behavior
• Only schemas that contain at least one declared entity of the same type are affected
• Built-in objects are excluded

Risk assessment: Users who currently use --prune expecting it to ONLY clear stage contents will now also get their undeclared procedures/functions dropped. This is a behavior change for existing --prune users that could be surprising, but it's documented in the help text update and release notes.

Side-Effect Verification

Query	Before --prune	After --prune
SHOW USER PROCEDURES IN SCHEMA ...PUBLIC	KEEP_PROC, STALE_PROC (2 rows)	KEEP_PROC only (1 row)
SHOW USER FUNCTIONS IN SCHEMA ...PUBLIC	KEEP_FN, STALE_FN (2 rows)	KEEP_FN only (1 row)
Overload test: KEEP_FN(VARCHAR) + KEEP_FN(NUMBER,NUMBER)	Both exist	Only KEEP_FN(VARCHAR) retained

All DROP operations used IF EXISTS (safe against race conditions).

Potential Risks

1. Overload safety in production: The prune logic matches by name AND argument types using same_type(). If same_type() has edge cases with complex types (e.g., ARRAY, OBJECT, VARIANT with qualifiers), it could incorrectly drop a valid overload. The existing same_type tests cover common mappings but complex nested types are not exhaustively tested.

2. FQN resolution fallback: When an entity doesn't declare database/schema explicitly, the code falls back to entity.udf_sproc_identifier.identifier_with_arg_types via FQN.from_string(). If this resolution fails or returns unexpected values, it could target the wrong schema.

3. Concurrent deployment: If multiple users deploy to the same schema simultaneously with --prune, one user's valid procedures could be dropped by another user's prune operation if they have different project definitions.

4. The 'packages' KeyError at common.py:175 is a pre-existing bug that surfaces when deploying to schemas with SQL-language procedures, but it's unrelated to this PR.

Verdict

PASS — The new --prune behavior correctly drops undeclared snowpark entities, preserves declared ones (including overload-safe matching), and the default --no-prune behavior is unchanged. The feature is opt-in and well-scoped.