Agent skill

pyrefly-type-coverage

Migrate a file to use stricter Pyrefly type checking with annotations required for all functions, classes, and attributes.

Stars 98,726
Forks 27,372

Install this agent skill to your Project

npx add-skill https://github.com/pytorch/pytorch/tree/main/.claude/skills/pyrefly-type-coverage

SKILL.md

Pyrefly Type Coverage Skill

This skill guides you through improving type coverage in Python files using Pyrefly, Meta's type checker. Follow this systematic process to add proper type annotations to files.

Prerequisites

  • The file you're working on should be in a project with a pyrefly.toml configuration

Step-by-Step Process

Step 1: Remove Ignore Errors Directive

First, locate and remove any pyre-ignore-all-errors comments at the top of the file:

python
# REMOVE lines like these:
# pyre-ignore-all-errors
# pyre-ignore-all-errors[16,21,53,56]
# @lint-ignore-every PYRELINT

These directives suppress type checking for the entire file and must be removed to enable proper type coverage.

Step 2: Add Entry to pyrefly.toml

Add a sub-config entry for stricter type checking. Open pyrefly.toml and add an entry following this pattern:

toml
[[sub-config]]
matches = "path/to/your/file.py"
[sub-config.errors]
implicit-import = false
implicit-any = true

For directory-level coverage:

toml
[[sub-config]]
matches = "path/to/directory/**"
[sub-config.errors]
implicit-import = false
implicit-any = true

You can also enable stricter options as needed:

toml
[[sub-config]]
matches = "path/to/your/file.py"
[sub-config.errors]
implicit-import = false
implicit-any = true
# Uncomment these for stricter checking:
# unannotated-attribute = true
# unannotated-parameter = true
# unannotated-return = true

Step 3: Run Pyrefly to Identify Missing Coverage

Execute the type checker to see all type errors:

bash
pyrefly check <FILENAME>

Example:

bash
pyrefly check torch/_dynamo/utils.py

This will output a list of type errors with line numbers and descriptions. Common error types include:

  • Missing return type annotations
  • Missing parameter type annotations
  • Incompatible types
  • Missing attribute definitions
  • Implicit Any usage

CRITICAL: Your goal is to resolve all errors. If you cannot resolve an error, you can use # pyrefly: ignore[...] to suppress but you should try to resolve the error first

Step 4: Add Type Annotations

Work through each error systematically:

  1. Read the function/code carefully - Understand what the function does
  2. Examine usage patterns - Look at how the function is called to understand expected types
  3. Add appropriate annotations - Add type hints based on your analysis

Common Annotation Patterns

Function signatures:

python
# Before
def process_data(items, callback):
    ...

# After
from collections.abc import Callable
def process_data(items: list[str], callback: Callable[[str], bool]) -> None:
    ...

Class attributes:

python
# Before
class MyClass:
    def __init__(self):
        self.value = None
        self.items = []

# After
class MyClass:
    value: int | None
    items: list[str]

    def __init__(self) -> None:
        self.value = None
        self.items = []

Complex types: CRITICAL: use syntax for Python >3.10 and prefer collections.abc as opposed to typing for better code standards.

Critical: For more advanced/generic types such as TypeAlias, TypeVar, Generic, Protocol, etc. use typing_extensions

python

# Optional values
def get_value(key: str) -> int | None: ...

# Union types
def process(value: str | int) -> str: ...

# Dict and List
def transform(data: dict[str, list[int]]) -> list[str]: ...

# Callable
from collections.abc import Callable
def apply(func: Callable[[int, int], int], a: int, b: int) -> int: ...

# TypeVar for generics
from typing_extensions import TypeVar
T = TypeVar('T')
def first(items: list[T]) -> T: ...

Using # pyre-ignore for specific lines:

If a specific line is difficult to type correctly (e.g., dynamic metaprogramming), you can ignore just that line:

python
# pyrefly: ignore[attr-defined]
result = getattr(obj, dynamic_name)()

CRITICAL: Avoid using # pyre-ignore unless it is necessary. When possible, we can implement stubs, or refactor code to make it more type-safe.

Step 5: Iterate and Verify

After adding annotations:

  1. Re-run pyrefly check to verify errors are resolved:

    bash
    pyrefly check <FILENAME>
    
  2. Fix any new errors that may appear from the annotations you added

  3. Repeat until clean - Continue until pyrefly reports no errors

Step 6: Commit Changes

To keep type coverage PRs manageable, you should commit your change once finished with a file.

Tips for Success

  1. Start with function signatures - Return types and parameter types are usually the highest priority

  2. Use from __future__ import annotations - Add this at the top of the file for forward references:

    python
    from __future__ import annotations
    
  3. Leverage type inference - Pyrefly can infer many types; focus on function boundaries

  4. Check existing type stubs - For external libraries, check if type stubs exist

  5. Use typing_extensions for newer features - For compatibility:

    python
    from typing_extensions import TypeAlias, Self, ParamSpec
    
  6. Document complex types with TypeAlias:

    python
    from typing import Dict, List, TypeAlias
    
    ConfigType: TypeAlias = Dict[str, List[int]]
    
    def process_config(config: ConfigType) -> None: ...
    

Example Workflow

bash
# 1. Open the file and remove pyre-ignore-all-errors
# 2. Add entry to pyrefly.toml

# 3. Check initial errors
pyrefly check torch/my_module.py

# 4. Add annotations iteratively

# 5. Re-check after changes
pyrefly check torch/my_module.py

# 6. Repeat until clean

Expand your agent's capabilities with these related and highly-rated skills.

pytorch/pytorch

aoti-debug

Debug AOTInductor (AOTI) errors and crashes. Use when encountering AOTI segfaults, device mismatch errors, constant loading failures, or runtime errors from aot_compile, aot_load, aoti_compile_and_package, or aoti_load_package.

98,726 27,372
Explore
pytorch/pytorch

add-uint-support

Add unsigned integer (uint) type support to PyTorch operators by updating AT_DISPATCH macros. Use when adding support for uint16, uint32, uint64 types to operators, kernels, or when user mentions enabling unsigned types, barebones unsigned types, or uint support.

98,726 27,372
Explore
pytorch/pytorch

at-dispatch-v2

Convert PyTorch AT_DISPATCH macros to AT_DISPATCH_V2 format in ATen C++ code. Use when porting AT_DISPATCH_ALL_TYPES_AND*, AT_DISPATCH_FLOATING_TYPES*, or other dispatch macros to the new v2 API. For ATen kernel files, CUDA kernels, and native operator implementations.

98,726 27,372
Explore
pytorch/pytorch

metal-kernel

Write Metal/MPS kernels for PyTorch operators. Use when adding MPS device support to operators, implementing Metal shaders, or porting CUDA kernels to Apple Silicon. Covers native_functions.yaml dispatch, host-side operators, and Metal kernel implementation.

98,726 27,372
Explore
pytorch/pytorch

scrub-issue

Fetch, analyze, reproduce, and minimize GitHub issue reproductions. Use when asked to check if an issue reproduces, minimize a repro, analyze a bug report, or scrub/triage an issue for reproducibility.

98,726 27,372
Explore
pytorch/pytorch

pr-review

Review PyTorch pull requests for code quality, test coverage, security, and backward compatibility. Use when reviewing PRs, when asked to review code changes, or when the user mentions "review PR", "code review", or "check this PR".

98,726 27,372
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results