[harness] apache/datafusion-python#1484: feat: add AI skill to find and improve the Pythonic interface to functions#6
Closed
sszz01 wants to merge 11 commits intoharness-base-3585c11eedfrom
Closed
Conversation
…uiring lit()
Update 47 functions in functions.py to accept native Python types (int, float,
str) for arguments that are contextually literals, eliminating verbose lit()
wrapping. For example, users can now write split_part(col("a"), ",", 2) instead
of split_part(col("a"), lit(","), lit(2)). All changes are backward compatible.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…ions Update instr and position (aliases of strpos) to accept Expr | str for the substring parameter, matching the updated primary function signature. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Alias functions that delegate to a primary function must have their type hints updated to match, even though coercion logic is only added to the primary. Added a new Step 3 to the implementation workflow for this. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Update SKILL.md to prevent three classes of issues: clarify that float already accepts int per PEP 484 (avoiding redundant int | float that fails ruff PYI041), add backward-compat rule for Category B so existing Expr params aren't removed, and add guidance for inline coercion with many optional nullable params instead of local helpers. Replace regexp_instr's _to_raw() helper with inline coercion matching the pattern used throughout the rest of the file. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…erns Introduce coerce_to_expr() and coerce_to_expr_or_none() in expr.py as the complement to ensure_expr() — where ensure_expr rejects non-Expr values, these helpers wrap them via Expr.literal(). Replaces ~60 inline isinstance checks in functions.py with single-line helper calls, and updates the make-pythonic skill to document the new pattern. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add Technique 1a to detect literal-only arguments in aggregate functions. Unlike scalar UDFs which enforce literals in invoke_with_args(), aggregate functions enforce them in accumulator() via get_scalar_value(), validate_percentile_expr(), or downcast_ref::<Literal>(). Without this technique, the skill would incorrectly classify arguments like approx_percentile_cont's percentile as Category A (Expr | float) when they should be Category B (float only). Updates the decision flow to branch on scalar vs aggregate before checking for literal enforcement. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add Technique 1b to detect literal-only arguments in window functions. Window functions enforce literals in partition_evaluator() via get_scalar_value_from_args() / downcast_ref::<Literal>(), not in invoke_with_args() (scalar) or accumulator() (aggregate). Updates the decision flow to branch on scalar vs aggregate vs window. Known window functions with literal-only arguments: ntile (n), lead/lag (offset, default_value), nth_value (n). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace 7 fragile truthiness checks (x.expr if x else None) with explicit is not None checks to prevent silent None when zero-valued literals are passed. Widen log/power/pow type hints to Expr | int | float with noqa: PYI041 for clarity. Add unit tests for coerce_to_expr helpers and integration tests for pythonic calling conventions. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add FBT003 (boolean positional value) to the per-file-ignores for python/tests/* in pyproject.toml, and remove the 6 now-redundant inline noqa: FBT003 comments across test_expr.py and test_context.py. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replace hardcoded "Known aggregate/window functions with literal-only arguments" lists with instructions to discover them dynamically by searching the upstream crate source. Keeps a few examples as validation anchors so the agent knows its search is working correctly. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
PyThreadState_SetAsyncExc only delivers exceptions when the thread is executing Python bytecode, not while in native (Rust/C) code. The previous test had two issues causing flakiness on Python 3.11: 1. The interrupt fired before df.collect() entered the UDF, while the thread was still in native code where async exceptions are ignored. 2. time.sleep(2.0) is a single C call where async exceptions are not checked — they're only checked between bytecode instructions. Fix by adding a threading.Event so the interrupt waits until the UDF is actually executing Python code, and by sleeping in small increments so the eval loop has opportunities to check for pending exceptions. Co-Authored-By: Claude Opus 4.6 (1M context) <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
2 participants
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.
Automated replay of apache#1484
Base sha:
3585c11eed778810e3317c56c2c25a8cdc29be5bHead sha:
e36eff29a6afb65df4aa2cf5a700f71887f2ca1bThis PR exists only to exercise the deployed LogoMesh App. Do not merge.