Docstring Conventions

Docstring Conventions

Apply Google-style docstrings when writing Python code in this repository.

RST/Sphinx Prohibition

NEVER use RST or Sphinx-style docstrings. This is a hard rule with zero exceptions.

Always use Google-style docstrings as defined in this skill. The following RST/Sphinx patterns are absolutely forbidden:

# CORRECT - Google-style (ALWAYS use this)
def fetch_user(user_id: int, include_inactive: bool = False) -> User:
    """Fetch a user from the database by their unique identifier.

    Args:
        user_id (int): The unique identifier for the user.
        include_inactive (bool): Whether to include deactivated users.

    Returns:
        User: The matching user record.

    Raises:
        UserNotFoundError: If no user matches the given ID.
    """

# INCORRECT - Sphinx/RST-style (NEVER use this)
def fetch_user(user_id: int, include_inactive: bool = False) -> User:
    """Fetch a user from the database by their unique identifier.

    :param user_id: The unique identifier for the user.
    :type user_id: int
    :param include_inactive: Whether to include deactivated users.
    :type include_inactive: bool
    :returns: The matching user record.
    :rtype: User
    :raises UserNotFoundError: If no user matches the given ID.
    """

Required Docstrings

| Element | Format | |---------|---------------------|--------| | Functions | Comprehensive (Args, Returns, Raises, Example) | | Classes | Comprehensive (Attributes, Example) | | Methods | Comprehensive (Args, Returns, Raises) | | Modules | Summary + extended description |

Function Docstring Structure

def function_name(param1: Type1, param2: Type2) -> ReturnType:
    """One-line summary of what the function does.

    Extended description explaining WHY this function exists and WHEN to use it.
    Do not explain HOW it works (the code shows that).

    Args:
        param1 (Type1): Description of first parameter.
        param2 (Type2): Description of second parameter.

    Returns:
        ReturnType: Description of return value.

    Raises:
        ExceptionType: When this exception is raised.

    Examples:
        >>> function_name(value1, value2)
        expected_output
    """

Section Rules

Inline Code Formatting

Use single backticks for inline code references in docstrings. Do NOT use rST-style double backticks.

# CORRECT - single backticks
"""Calculate the distance between `point_a` and `point_b`.

Args:
    point_a (Point): The starting `Point` instance.
    point_b (Point): The ending `Point` instance.

Returns:
    float: Euclidean distance. Returns `0.0` if points are identical.

Raises:
    TypeError: If `point_a` or `point_b` is not a `Point`.
"""

# INCORRECT - rST-style double backticks
"""Calculate the distance between ``point_a`` and ``point_b``.

Args:
    point_a (Point): The starting ``Point`` instance.
"""

# CORRECT - plain backtick references
"""Parse the response into a `SearchResult`.

Delegates to `ResponseParser.parse()` for deserialization.
Call `validate_response()` before using this function.
"""

# INCORRECT - rST role tags
"""Parse the response into a :class:`SearchResult`.

Delegates to :meth:`ResponseParser.parse()` for deserialization.
Call :func:`validate_response` before using this function.
"""

Complete Example

def calculate_similarity(embedding_a: list[float], embedding_b: list[float]) -> float:
    """Calculate cosine similarity between two embeddings.

    Use this for comparing semantic similarity of text chunks when building
    retrieval systems. Values closer to 1.0 indicate higher similarity.

    Args:
        embedding_a (list[float]): First embedding vector.
        embedding_b (list[float]): Second embedding vector.

    Returns:
        float: Cosine similarity score between -1.0 and 1.0.

    Raises:
        ValueError: If vectors have different dimensions.

    Examples:
        >>> calculate_similarity([1.0, 0.0], [1.0, 0.0])
        1.0
    """

Line Length

Docstring lines are allowed to extend to 120 characters. Do not wrap docstring prose at 79 or 88 characters.

Class Docstring Structure

class ClassName:
    """One-line summary of what the class represents.

    Extended description of the class purpose and when to use it.

    Attributes:
        attr1 (Type1): Description of public attribute.
        attr2 (Type2): Description of public attribute.

    Examples:
        >>> obj = ClassName(param)
        >>> obj.method()
        expected_output
    """

    def __init__(self, param: Type) -> None:
        """Initialize the class.

        Args:
            param (Type): Description of parameter.
        """

Module Docstring Structure

Every module file must start with a docstring:

"""One-line summary of module purpose.

Extended description of what the module provides and when to use it.
"""

import ...

Example

"""Embedding utilities for vector similarity search.

This module provides functions for creating, comparing, and manipulating
embeddings used in semantic search and retrieval systems.
"""

from typing import Final
...

Docstrings vs Comments

When to Use Comments

Add comments only when the why isn't obvious:

# CORRECT - explains why this approach was chosen
# Use binary search here because the list is sorted and can have 100k+ items
index = bisect.bisect_left(sorted_items, target)

# CORRECT - explains a non-obvious constraint
# Must check expiry before validation because expired tokens fail silently
if token.is_expired():
    raise TokenExpiredError()

# INCORRECT - explains what (the code already shows this)
# Get the index using binary search
index = bisect.bisect_left(sorted_items, target)

# INCORRECT - obvious from the code
# Increment the counter
counter += 1

Forbidden Patterns

# INCORRECT - no docstring on public function
def fetch_user(user_id: int) -> User:
    return db.query(User).get(user_id)

# INCORRECT - no docstring on private function
def _hash_password(password: str) -> str:
    return hashlib.sha256(password.encode()).hexdigest()

# INCORRECT - docstring explains "how" instead of "why"
def fetch_user(user_id: int) -> User:
    """Query the database for a user by ID and return the User object."""
    return db.query(User).get(user_id)

# INCORRECT - missing Args/Returns sections
def fetch_user(user_id: int) -> User:
    """Fetch a user from the database."""
    return db.query(User).get(user_id)

Doctest Examples

When including examples, make them runnable with doctest:

def add_vectors(a: list[float], b: list[float]) -> list[float]:
    """Add two vectors element-wise.

    Args:
        a (list[float]): First vector.
        b (list[float]): Second vector.

    Returns:
        list[float]: Element-wise sum.

    Examples:
        >>> add_vectors([1.0, 2.0], [3.0, 4.0])
        [4.0, 6.0]
        >>> add_vectors([0.0], [0.0])
        [0.0]
    """

Validation

Run docstring checks via the validate-code skill:

uv run .claude/scripts/validate_code.py --docstring <path>

For doctest examples specifically:

uv run .claude/scripts/validate_code.py --doctest <path>

Converted and distributed by TomeVault — claim your Tome and manage your conversions.