pre-commit

.pre-commit-config.yaml

@@ -39,6 +39,7 @@ repos:
         language: system
         types: [python]
         require_serial: true
+        exclude: ^(tests|api-coverage-tests)/
       # prettier
       - id: prettier
         name: prettier

ndonnx/__init__.py

@@ -4,8 +4,324 @@
 import importlib.metadata
 import warnings
 
+from ._array import Array, array, from_spox_var
+from ._build import (
+    build,
+)
+from ._data_types import (
+    CastError,
+    CoreType,
+    Floating,
+    Integral,
+    Nullable,
+    NullableFloating,
+    NullableIntegral,
+    NullableNumerical,
+    Numerical,
+    from_numpy_dtype,
+    bool,
+    float32,
+    float64,
+    int8,
+    int16,
+    int32,
+    int64,
+    nbool,
+    nfloat32,
+    nfloat64,
+    nint8,
+    nint16,
+    nint32,
+    nint64,
+    nuint8,
+    nuint16,
+    nuint32,
+    nuint64,
+    nutf8,
+    uint8,
+    uint16,
+    uint32,
+    uint64,
+    utf8,
+)
+from ._funcs import (
+    arange,
+    asarray,
+    empty,
+    empty_like,
+    eye,
+    full,
+    full_like,
+    linspace,
+    ones,
+    ones_like,
+    tril,
+    triu,
+    zeros,
+    zeros_like,
+    astype,
+    broadcast_arrays,
+    broadcast_to,
+    can_cast,
+    finfo,
+    iinfo,
+    result_type,
+    abs,
+    acos,
+    acosh,
+    add,
+    asin,
+    asinh,
+    atan,
+    atan2,
+    atanh,
+    bitwise_and,
+    bitwise_left_shift,
+    bitwise_invert,
+    bitwise_or,
+    bitwise_right_shift,
+    bitwise_xor,
+    ceil,
+    cos,
+    cosh,
+    divide,
+    equal,
+    exp,
+    expm1,
+    floor,
+    floor_divide,
+    greater,
+    greater_equal,
+    isfinite,
+    isinf,
+    isnan,
+    less,
+    less_equal,
+    log,
+    log1p,
+    log2,
+    log10,
+    logaddexp,
+    logical_and,
+    logical_not,
+    logical_or,
+    logical_xor,
+    multiply,
+    negative,
+    not_equal,
+    positive,
+    pow,
+    remainder,
+    round,
+    sign,
+    sin,
+    sinh,
+    square,
+    sqrt,
+    subtract,
+    tan,
+    tanh,
+    trunc,
+    matmul,
+    matrix_transpose,
+    concat,
+    expand_dims,
+    flip,
+    permute_dims,
+    reshape,
+    roll,
+    squeeze,
+    stack,
+    argmax,
+    argmin,
+    nonzero,
+    searchsorted,
+    where,
+    unique_all,
+    unique_counts,
+    unique_inverse,
+    unique_values,
+    argsort,
+    sort,
+    cumulative_sum,
+    max,
+    mean,
+    min,
+    prod,
+    clip,
+    std,
+    sum,
+    var,
+    all,
+    any,
+    take,
+)
+from ._constants import (
+    e,
+    inf,
+    nan,
+    pi,
+)
+
 try:
     __version__ = importlib.metadata.version(__name__)
-except importlib.metadata.PackageNotFoundError as e:  # pragma: no cover
-    warnings.warn(f"Could not determine version of {__name__}\n{e!s}", stacklevel=2)
+except importlib.metadata.PackageNotFoundError as err:  # pragma: no cover
+    warnings.warn(f"Could not determine version of {__name__}\n{err!s}", stacklevel=2)
     __version__ = "unknown"
+
+
+__all__ = [
+    "Array",
+    "array",
+    "from_spox_var",
+    "e",
+    "inf",
+    "nan",
+    "pi",
+    "arange",
+    "asarray",
+    "empty",
+    "empty_like",
+    "eye",
+    "full",
+    "full_like",
+    "linspace",
+    "ones",
+    "ones_like",
+    "tril",
+    "triu",
+    "zeros",
+    "zeros_like",
+    "astype",
+    "take",
+    "broadcast_arrays",
+    "broadcast_to",
+    "can_cast",
+    "finfo",
+    "iinfo",
+    "result_type",
+    "abs",
+    "acos",
+    "acosh",
+    "add",
+    "asin",
+    "asinh",
+    "atan",
+    "atan2",
+    "atanh",
+    "bitwise_and",
+    "bitwise_left_shift",
+    "bitwise_invert",
+    "bitwise_or",
+    "bitwise_right_shift",
+    "bitwise_xor",
+    "ceil",
+    "cos",
+    "cosh",
+    "divide",
+    "equal",
+    "exp",
+    "expm1",
+    "floor",
+    "floor_divide",
+    "greater",
+    "greater_equal",
+    "isfinite",
+    "isinf",
+    "isnan",
+    "less",
+    "less_equal",
+    "log",
+    "log1p",
+    "log2",
+    "log10",
+    "logaddexp",
+    "logical_and",
+    "logical_not",
+    "logical_or",
+    "logical_xor",
+    "multiply",
+    "negative",
+    "not_equal",
+    "positive",
+    "pow",
+    "remainder",
+    "round",
+    "sign",
+    "sin",
+    "sinh",
+    "square",
+    "sqrt",
+    "subtract",
+    "tan",
+    "tanh",
+    "trunc",
+    "matmul",
+    "matrix_transpose",
+    "concat",
+    "expand_dims",
+    "flip",
+    "permute_dims",
+    "reshape",
+    "roll",
+    "squeeze",
+    "stack",
+    "argmax",
+    "argmin",
+    "nonzero",
+    "searchsorted",
+    "where",
+    "unique_all",
+    "unique_counts",
+    "unique_inverse",
+    "unique_values",
+    "argsort",
+    "sort",
+    "cumulative_sum",
+    "max",
+    "mean",
+    "min",
+    "prod",
+    "clip",
+    "std",
+    "sum",
+    "var",
+    "all",
+    "any",
+    "build",
+    "bool",
+    "utf8",
+    "float32",
+    "float64",
+    "int8",
+    "int16",
+    "int32",
+    "int64",
+    "uint8",
+    "uint16",
+    "uint32",
+    "uint64",
+    "nbool",
+    "nutf8",
+    "nfloat32",
+    "nfloat64",
+    "nint8",
+    "nint16",
+    "nint32",
+    "nint64",
+    "nuint8",
+    "nuint16",
+    "nuint32",
+    "nuint64",
+    "NullableNumerical",
+    "Numerical",
+    "NullableFloating",
+    "Floating",
+    "NullableIntegral",
+    "Nullable",
+    "Integral",
+    "CoreType",
+    "CastError",
+    "promote_nullable",
+    "from_numpy_dtype",
+]

ndonnx/_array.py

@@ -472,7 +472,7 @@ def __ne__(self, other):
         return ndx.not_equal(self, other)
 
     @property
-    def mT(self) -> ndx.Array:
+    def mT(self) -> ndx.Array:  # noqa: N802
         """Transpose of a matrix (or a stack of matrices).
 
         If an array instance has fewer than two dimensions, an error should be raised.
@@ -496,7 +496,7 @@ def size(self) -> ndx.Array:
         return ndx.prod(self.shape)
 
     @property
-    def T(self) -> ndx.Array:
+    def T(self) -> ndx.Array:  # noqa: N802
         """Transpose of the array.
 
         The array instance must be two-dimensional.

ndonnx/_build.py

@@ -89,7 +89,7 @@ def _get_dtype(dtype: str, version: int) -> dtypes.StructType | dtypes.CoreType:
 
 
 def _extract_output_names(
-    output_dtypes: dict[str, dtypes.StructType | dtypes.CoreType]
+    output_dtypes: dict[str, dtypes.StructType | dtypes.CoreType],
 ) -> dict[str, ndx.CoreType]:
     """Given a dictionary mapping output names to their data types, extract the
     underlying fully qualified names and their CoreTypes.

ndonnx/_core/_impl.py

@@ -24,7 +24,6 @@
 
 
 class CoreOperationsImpl(OperationsBlock):
-
     # elementwise.py
 
     def abs(self, x):
@@ -522,7 +521,6 @@ def where_dtype_agnostic(a: ndx.Array, b: ndx.Array) -> ndx.Array:
 
     # set.py
     def unique_all(self, x):
-
         new_dtype = x.dtype
         if isinstance(x.dtype, dtypes.Integral) or x.dtype in (
             dtypes.bool,

ndonnx/_corearray.py

@@ -81,7 +81,6 @@ def __setitem__(
     def _normalise_index(
         self, index: IndexType
     ) -> _CoreArray | tuple[ScalarIndexType, ...]:
-
         if isinstance(index, _CoreArray):
             return index
         else:

ndonnx/_propagation.py

@@ -124,9 +124,7 @@ def collect_lazy_arguments(obj):
             else:
                 if obj.dtype in (ndx.utf8, ndx.nutf8):
                     # Lazy variant due to onnxruntime bug
-                    lazy = ndx.array(
-                        shape=obj._static_shape, dtype=obj.dtype
-                    )  # type: ignore
+                    lazy = ndx.array(shape=obj._static_shape, dtype=obj.dtype)  # type: ignore
 
                     # disassemble the array into its core_arrays
                     _flatten(obj, lazy, flattened_inference_inputs)