Module streamgen.transforms

➡️ useful transformations.

View Source

"""➡️ useful transformations."""

from collections.abc import Callable

from functools import wraps

from typing import Any

import numpy as np

class LabelEncoder:

    """🏷️ index of class-name lookup transform.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> LabelEncoder(["A", "B", "C"])("C")

        array(2, dtype=int64)

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = {class_: idx for idx, class_ in enumerate(classes)}

    def __call__(self, label: str) -> np.ndarray:

        """Standard call method.

        Args:

            label (str): class name

        Returns:

            np.ndarray: index encoded target array

        """

        return np.array(self.classes[label], dtype=np.int64)

class LabelDecoder:

    """🏷️ label/class-name for index lookup transform.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> LabelDecoder(["A", "B", "C"])(1)

        'B'

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = classes

    def __call__(self, idx: int) -> str:

        """Standard call method.

        Args:

            idx (int): index of the target class

        Returns:

            str: class name

        """

        return self.classes[idx]

class MultiLabelEncoder:

    """🏷️ encoding of multiple labels/class-names.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> MultiLabelEncoder(["A", "B", "C"])(["A", "C"])

        array([1., 0., 1.], dtype=float32)

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = classes

    def __call__(self, labels: list[str]) -> np.ndarray:

        """Standard call method.

        Args:

            labels (list[str]): (potentially empty) list of class_ names.

        Returns:

            np.ndarray: encoded array of targets

        """

        return np.array([int(p in labels) for p in self.classes], dtype=np.float32)

class MultiLabelDecoder:

    """🏷️ decoding of labels/class-names.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> MultiLabelDecoder(["A", "B", "C"])([1.0, 0.0, 1.0])

        array([1., 0., 1.], dtype=float32)

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = classes

    def __call__(self, labels: list[int | float]) -> list[str]:

        """Standard call method.

        Args:

            labels (list[int | float]): encoded array of targets.

                must have the same length as `self.classes`.

        Returns:

            list[str]: (potentially empty) list of class_ names

        """

        return [p for i, p in enumerate(self.classes) if bool(labels[i])]

def noop(input: Any) -> Any:  # noqa: ANN401, A002

    """🤷 no-operation. Passes through the input.

    Args:

        input (Any): any input

    Returns:

        Any: unmodified input

    """

    return input

def set_value_in_dict(dictionary: dict[str, Any], value: Any, key: str) -> dict:  # noqa: ANN401

    """📖 sets `dictionary[key]` to `value` and returns the result.

    Args:

        dictionary (dict): any dictionary

        value (Any): any value

        key (str): key of the value in `dictionary`. Defaults to "target".

    Returns:

        dict: labeled `dictionary`

    """

    dictionary[key] = value

    return dictionary

def operate_on_key(key: str = "input") -> Callable:

    """🗝️ function decorator, that converts `func(input: dict, ...)` to `func(input[key], ...)`.

    The result is put back into `input`, the first argument to `func`.

    This facilitates working with dictionaries as first arguments.

    Examples:

        >>> @operate_on_key(key="x")

            def add(x, n):

                return x + n

        >>> add({"x": 1, "target": None}, 2)

        {'x': 3, 'target': None}

    Args:

        key (str, optional): key for fetching the value from `input`. Defaults to "input".

    Returns:

        Callable: modified function

    """

    def decorator(func):  # noqa: ANN202, ANN001

        @wraps(func)

        def wrapper(input, *args, **kwargs):  # noqa: ANN202, ANN001, A002, ANN002, ANN003

            x = input[key]

            out = func(x, *args, **kwargs)

            input[key] = out

            return input

        return wrapper

    return decorator

def operate_on_index(idx: int = 0) -> Callable:

    """#️⃣ function decorator, that converts `func(input: Sequence, ...)` to `func(input[idx], ...)`.

    The other parts of the input sequence are not passed to the function, but still returned by the decorator.

    This facilitates working with sequences (like tuples) as first arguments.

    Examples:

        >>> @operate_on_index()

            def add(x, n):

                return x + n

        >>> add((1, "target"), 2)

        (3, "target")

    Args:

        idx (int, optional): index for fetching the value from `input`. Defaults to 0 (the first element).

    Returns:

        Callable: modified function

    """

    def decorator(func):  # noqa: ANN202, ANN001

        @wraps(func)

        def wrapper(input, *args, **kwargs):  # noqa: ANN202, ANN001, A002, ANN002, ANN003

            input_type = type(input)

            input = list(input)  # noqa: A001

            x = input[idx]

            out = func(x, *args, **kwargs)

            input[idx] = out

            return input_type(input)

        return wrapper

    return decorator

Functions

noop

def noop(
    input: Any
) -> Any

🤷 no-operation. Passes through the input.

Parameters:

Name	Type	Description	Default
input	Any	any input	None

Returns:

Type	Description
Any	unmodified input

View Source

def noop(input: Any) -> Any:  # noqa: ANN401, A002

    """🤷 no-operation. Passes through the input.

    Args:

        input (Any): any input

    Returns:

        Any: unmodified input

    """

    return input

operate_on_index

def operate_on_index(
    idx: int = 0
) -> collections.abc.Callable

️⃣ function decorator, that converts `func(input: Sequence, ...)` to `func(input[idx], ...)`.

The other parts of the input sequence are not passed to the function, but still returned by the decorator.

This facilitates working with sequences (like tuples) as first arguments.

Parameters:

Name	Type	Description	Default
idx	int	index for fetching the value from `input`. Defaults to 0 (the first element).	0 (the first element)

Returns:

Type	Description
Callable	modified function

View Source

def operate_on_index(idx: int = 0) -> Callable:

    """#️⃣ function decorator, that converts `func(input: Sequence, ...)` to `func(input[idx], ...)`.

    The other parts of the input sequence are not passed to the function, but still returned by the decorator.

    This facilitates working with sequences (like tuples) as first arguments.

    Examples:

        >>> @operate_on_index()

            def add(x, n):

                return x + n

        >>> add((1, "target"), 2)

        (3, "target")

    Args:

        idx (int, optional): index for fetching the value from `input`. Defaults to 0 (the first element).

    Returns:

        Callable: modified function

    """

    def decorator(func):  # noqa: ANN202, ANN001

        @wraps(func)

        def wrapper(input, *args, **kwargs):  # noqa: ANN202, ANN001, A002, ANN002, ANN003

            input_type = type(input)

            input = list(input)  # noqa: A001

            x = input[idx]

            out = func(x, *args, **kwargs)

            input[idx] = out

            return input_type(input)

        return wrapper

    return decorator

operate_on_key

def operate_on_key(
    key: str = 'input'
) -> collections.abc.Callable

🗝️ function decorator, that converts func(input: dict, ...) to func(input[key], ...).

The result is put back into input, the first argument to func.

This facilitates working with dictionaries as first arguments.

Parameters:

Name	Type	Description	Default
key	str	key for fetching the value from `input`. Defaults to "input".	"input"

Returns:

Type	Description
Callable	modified function

View Source

def operate_on_key(key: str = "input") -> Callable:

    """🗝️ function decorator, that converts `func(input: dict, ...)` to `func(input[key], ...)`.

    The result is put back into `input`, the first argument to `func`.

    This facilitates working with dictionaries as first arguments.

    Examples:

        >>> @operate_on_key(key="x")

            def add(x, n):

                return x + n

        >>> add({"x": 1, "target": None}, 2)

        {'x': 3, 'target': None}

    Args:

        key (str, optional): key for fetching the value from `input`. Defaults to "input".

    Returns:

        Callable: modified function

    """

    def decorator(func):  # noqa: ANN202, ANN001

        @wraps(func)

        def wrapper(input, *args, **kwargs):  # noqa: ANN202, ANN001, A002, ANN002, ANN003

            x = input[key]

            out = func(x, *args, **kwargs)

            input[key] = out

            return input

        return wrapper

    return decorator

set_value_in_dict

def set_value_in_dict(
    dictionary: dict[str, typing.Any],
    value: Any,
    key: str
) -> dict

📖 sets dictionary[key] to value and returns the result.

Parameters:

Name	Type	Description	Default
dictionary	dict	any dictionary	None
value	Any	any value	None
key	str	key of the value in `dictionary`. Defaults to "target".	"target"

Returns:

Type	Description
dict	labeled `dictionary`

View Source

def set_value_in_dict(dictionary: dict[str, Any], value: Any, key: str) -> dict:  # noqa: ANN401

    """📖 sets `dictionary[key]` to `value` and returns the result.

    Args:

        dictionary (dict): any dictionary

        value (Any): any value

        key (str): key of the value in `dictionary`. Defaults to "target".

    Returns:

        dict: labeled `dictionary`

    """

    dictionary[key] = value

    return dictionary

Classes

LabelDecoder

class LabelDecoder(
    classes: list[str]
)

🏷️ label/class-name for index lookup transform.

Attributes

Name	Type	Description	Default
classes	List[str]	list of labels/class-names. Order determines the indices.	None

View Source

class LabelDecoder:

    """🏷️ label/class-name for index lookup transform.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> LabelDecoder(["A", "B", "C"])(1)

        'B'

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = classes

    def __call__(self, idx: int) -> str:

        """Standard call method.

        Args:

            idx (int): index of the target class

        Returns:

            str: class name

        """

        return self.classes[idx]

LabelEncoder

class LabelEncoder(
    classes: list[str]
)

🏷️ index of class-name lookup transform.

Attributes

Name	Type	Description	Default
classes	List[str]	list of labels/class-names. Order determines the indices.	None

View Source

class LabelEncoder:

    """🏷️ index of class-name lookup transform.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> LabelEncoder(["A", "B", "C"])("C")

        array(2, dtype=int64)

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = {class_: idx for idx, class_ in enumerate(classes)}

    def __call__(self, label: str) -> np.ndarray:

        """Standard call method.

        Args:

            label (str): class name

        Returns:

            np.ndarray: index encoded target array

        """

        return np.array(self.classes[label], dtype=np.int64)

MultiLabelDecoder

class MultiLabelDecoder(
    classes: list[str]
)

🏷️ decoding of labels/class-names.

Attributes

Name	Type	Description	Default
classes	List[str]	list of labels/class-names. Order determines the indices.	None

View Source

class MultiLabelDecoder:

    """🏷️ decoding of labels/class-names.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> MultiLabelDecoder(["A", "B", "C"])([1.0, 0.0, 1.0])

        array([1., 0., 1.], dtype=float32)

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = classes

    def __call__(self, labels: list[int | float]) -> list[str]:

        """Standard call method.

        Args:

            labels (list[int | float]): encoded array of targets.

                must have the same length as `self.classes`.

        Returns:

            list[str]: (potentially empty) list of class_ names

        """

        return [p for i, p in enumerate(self.classes) if bool(labels[i])]

MultiLabelEncoder

class MultiLabelEncoder(
    classes: list[str]
)

🏷️ encoding of multiple labels/class-names.

Attributes

Name	Type	Description	Default
classes	List[str]	list of labels/class-names. Order determines the indices.	None

View Source

class MultiLabelEncoder:

    """🏷️ encoding of multiple labels/class-names.

    Args:

        classes (List[str]): list of labels/class-names. Order determines the indices.

    Example:

        >>> MultiLabelEncoder(["A", "B", "C"])(["A", "C"])

        array([1., 0., 1.], dtype=float32)

    """

    def __init__(self, classes: list[str]) -> None:  # noqa: D107

        self.classes = classes

    def __call__(self, labels: list[str]) -> np.ndarray:

        """Standard call method.

        Args:

            labels (list[str]): (potentially empty) list of class_ names.

        Returns:

            np.ndarray: encoded array of targets

        """

        return np.array([int(p in labels) for p in self.classes], dtype=np.float32)

Module streamgen.transforms

Functions

noop

operate_on_index

️⃣ function decorator, that converts func(input: Sequence, ...) to func(input[idx], ...).

operate_on_key

set_value_in_dict

Classes

LabelDecoder

Attributes

LabelEncoder

Attributes

MultiLabelDecoder

Attributes

MultiLabelEncoder

Attributes

️⃣ function decorator, that converts `func(input: Sequence, ...)` to `func(input[idx], ...)`.