Last active
January 28, 2020 13:05
-
-
Save dmontagu/69baaa9459bbfc471136ffd5d38bc6b6 to your computer and use it in GitHub Desktop.
Proof of concept tracking numpy array data types and shapes
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
from typing import TYPE_CHECKING, Any, Generic, Tuple, Type, TypeVar, Union | |
import numpy as np | |
from typing_extensions import Literal | |
T = TypeVar("T") | |
S = TypeVar("S") | |
class Dimension(Generic[T]): | |
if not TYPE_CHECKING: | |
def __class_getitem__(cls: Type[S], item: Any) -> Type[S]: | |
return cls | |
D1 = TypeVar("D1", bound=Dimension[Any]) | |
D2 = TypeVar("D2", bound=Dimension[Any]) | |
D3 = TypeVar("D3", bound=Dimension[Any]) | |
D4 = TypeVar("D4", bound=Dimension[Any]) | |
D5 = TypeVar("D5", bound=Dimension[Any]) | |
if TYPE_CHECKING: | |
class NPFloat: | |
pass | |
class NPInt: | |
pass | |
class NPUInt8: | |
pass | |
else: | |
NPInt = np.int | |
NPUInt8 = np.uint8 | |
NPFloat = np.float | |
DType = TypeVar("DType", NPInt, NPUInt8, NPFloat, bool) | |
if TYPE_CHECKING: | |
class _TypedArray(Generic[DType]): | |
if not TYPE_CHECKING: | |
def __class_getitem__(cls: Type[T], item: Any) -> Type[T]: | |
return cls | |
def __init__( | |
self, | |
shape: Any, | |
dtype: Type[DType] = None, | |
buffer: Any = None, | |
offset: int = 0, | |
strides: Any = None, | |
order: Any = None, | |
) -> None: | |
# super().__init__(shape, dtype, buffer, offset, strides, order) | |
pass | |
def __iter__(self) -> Any: | |
pass | |
@property | |
def shape(self) -> Tuple[int, ...]: | |
... | |
def __getitem__(self, item: Any) -> Any: | |
pass | |
def __getattr__(self, item: Any) -> Any: | |
pass | |
def __setattr__(self, item: Any, value: Any) -> None: | |
pass | |
else: | |
_TypedArray = np.ndarray | |
Zero = Dimension[Literal[0]] | |
One = Dimension[Literal[1]] | |
Two = Dimension[Literal[2]] | |
Three = Dimension[Literal[3]] | |
N = Dimension[int] | |
M = Dimension[int] | |
class Array1D(_TypedArray, Generic[DType, D1]): # type: ignore[type-arg] | |
if TYPE_CHECKING: | |
def __ge__(self, other: Union["_TypedArray[Any]", float]) -> "Array1D[bool, D1]": | |
... | |
def __le__(self, other: Union["_TypedArray[Any]", float]) -> "Array1D[bool, D1]": | |
... | |
def __gt__(self, other: Union["_TypedArray[Any]", float]) -> "Array1D[bool, D1]": | |
... | |
def __lt__(self, other: Union["_TypedArray[Any]", float]) -> "Array1D[bool, D1]": | |
... | |
class Array2D(_TypedArray, Generic[DType, D1, D2]): # type: ignore[type-arg] | |
if TYPE_CHECKING: | |
def __ge__(self, other: Union["_TypedArray[Any]", float]) -> "Array2D[bool, D1, D2]": | |
... | |
def __le__(self, other: Union["_TypedArray[Any]", float]) -> "Array2D[bool, D1, D2]": | |
... | |
def __gt__(self, other: Union["_TypedArray[Any]", float]) -> "Array2D[bool, D1, D2]": | |
... | |
def __lt__(self, other: Union["_TypedArray[Any]", float]) -> "Array2D[bool, D1, D2]": | |
... | |
class Array3D(_TypedArray, Generic[DType, D1, D2, D3]): # type: ignore[type-arg] | |
pass | |
class Array4D(_TypedArray, Generic[DType, D1, D2, D3, D4]): # type: ignore[type-arg] | |
pass | |
class Array5D(_TypedArray, Generic[DType, D1, D2, D3, D4, D5]): # type: ignore[type-arg] | |
pass | |
AnyArray1D = Array1D[Any, Any] | |
AnyArray2D = Array2D[Any, Any, Any] | |
AnyArray3D = Array3D[Any, Any, Any, Any] | |
AnyArray4D = Array4D[Any, Any, Any, Any, Any] | |
AnyArray5D = Array5D[Any, Any, Any, Any, Any, Any] | |
""" | |
# Usage example: | |
def f(x: Array2D[NPUInt8, D1, D2]) -> Array1D[NPFloat, D1]: | |
y = x[0] / 255 | |
return y | |
def g(x: Array1D[NPFloat, D1]) -> Array2D[NPFloat, One, D1]: | |
return x[np.newaxis, ...] | |
def h(x: Array2D[NPFloat, One, One]) -> float: | |
return float(x[0][0]) | |
a: Array1D[NPInt, One] = Array1D(np.ones(shape=(1,), dtype=NPInt)) | |
b: Array2D[NPFloat, M, N] = Array2D(np.ones(shape=(50, 10), dtype=NPFloat)) | |
c: Array2D[NPUInt8, Two, N] = Array2D(np.ones(shape=(1, 10), dtype=NPUInt8)) | |
d: Array2D[NPUInt8, One, N] = Array2D(np.ones(shape=(1, 10), dtype=NPUInt8)) | |
f(a) | |
# Argument 1 to "f" has incompatible type "Array1D[NPInt, Dimension[Literal[1]]]"; | |
# expected "Array2D[NPUInt8, <nothing>, <nothing>]" [arg-type] | |
f(b) | |
# Argument 1 to "f" has incompatible type "Array2D[NPFloat, Dimension[int], Dimension[int]]"; | |
# expected "Array2D[NPUInt8, Dimension[int], Dimension[int]]" | |
f(c) # OK | |
f(d) # OK | |
g(f(c)) # OK | |
g(f(d)) # OK | |
h(g(f(c))) | |
# Argument 1 to "f" has incompatible type "Array2D[NPUInt8, Dimension[Literal[2]], Dimension[int]]"; | |
# expected "Array2D[NPUInt8, Dimension[Literal[1]], Dimension[int]]" | |
h(g(f(d))) # OK | |
""" |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment