NumPy¶

Strategies related to NumPy in the context of Awkward Array.

These strategies are related to the section of Awkward Array User Guide "How to convert to/from NumPy".

Functions:

Name	Description
`numpy_dtypes`	Strategy for dtypes (simple or array) supported by Awkward Array.
`supported_dtype_names`	Strategy for names of NumPy dtypes supported by Awkward Array.
`supported_dtypes`	Strategy for NumPy dtypes supported by Awkward Array.
`from_numpy`	Strategy for Awkward Arrays created from NumPy arrays.
`numpy_arrays`	Strategy for NumPy arrays from which Awkward Arrays can be created.

numpy_dtypes ¶

numpy_dtypes(
    *,
    dtype: dtype | SearchStrategy[dtype] | None = None,
    allow_array: bool = True,
    max_size: int = 5,
) -> st.SearchStrategy[np.dtype]

Strategy for dtypes (simple or array) supported by Awkward Array.

Examples of simple dtypes are dtype('int32'), dtype('float64')

Array dtypes include ([('f0', 'i4'), ('f1', 'f8')]). They are dtypes of structured NumPy arrays.

Parameters:

Name	Type	Description	Default
`dtype`	`dtype \| SearchStrategy[dtype] \| None`	A simple dtype or a strategy for simple dtypes for determining the type of array elements. If `None`, any supported simple dtype is used.	`None`
`allow_array`	`bool`	Generate only simple dtypes if `False`, else array dtypes as well.	`True`
`max_size`	`int`	Maximum number of scalars in a structured dtype.	`5`

Examples:

>>> numpy_dtypes().example()
dtype(...)

supported_dtype_names ¶

supported_dtype_names() -> st.SearchStrategy[str]

Strategy for names of NumPy dtypes supported by Awkward Array.

Examples:

>>> supported_dtype_names().example()
'...'

supported_dtypes ¶

supported_dtypes() -> st.SearchStrategy[np.dtype]

Strategy for NumPy dtypes supported by Awkward Array.

Examples:

>>> supported_dtypes().example()
dtype(...)

from_numpy ¶

from_numpy(
    *,
    dtype: dtype | SearchStrategy[dtype] | None = None,
    allow_structured: bool = True,
    allow_nan: bool = False,
    regulararray: bool | None = None,
    max_size: int = 10,
) -> st.SearchStrategy[ak.Array]

Strategy for Awkward Arrays created from NumPy arrays.

Parameters:

Name	Type	Description	Default
`dtype`	`dtype \| SearchStrategy[dtype] \| None`	A simple dtype or a strategy for simple dtypes for determining the type of array elements. If `None`, any supported simple dtype is used.	`None`
`allow_structured`	`bool`	Generate only from simple NumPy arrays if `False`, else from structured NumPy arrays as well.	`True`
`allow_nan`	`bool`	Generate potentially `NaN` for relevant dtypes if `True`.	`False`
`regulararray`	`bool \| None`	Passed to `ak.from_numpy`. If `None` (default), randomly generates `True` or `False`.	`None`
`max_size`	`int`	Maximum number of scalars in the array. For structured dtypes, each element counts as multiple scalars (one per field).	`10`

Examples:

>>> from_numpy().example()
<Array ... type='...'>

numpy_arrays ¶

numpy_arrays(
    draw: DrawFn,
    *,
    dtype: dtype | SearchStrategy[dtype] | None = None,
    allow_structured: bool = True,
    allow_nan: bool = False,
    min_dims: int = 1,
    max_dims: int | None = None,
    min_size: int = 0,
    max_size: int = 10,
) -> np.ndarray

Strategy for NumPy arrays from which Awkward Arrays can be created.

Parameters:

Name	Type	Description	Default
`dtype`	`dtype \| SearchStrategy[dtype] \| None`	A simple dtype or a strategy for simple dtypes for determining the type of array elements. If `None`, any supported simple dtype is used.	`None`
`allow_structured`	`bool`	Generate only simple arrays if `False`, else structured arrays as well.	`True`
`allow_nan`	`bool`	Generate potentially `NaN` for relevant dtypes if `True`.	`False`
`min_dims`	`int`	Minimum number of dimensions.	`1`
`max_dims`	`int \| None`	Maximum number of dimensions. If `None`, auto-derived from `max_size`.	`None`
`min_size`	`int`	Minimum number of scalars in the array. For structured dtypes, each element counts as multiple scalars (one per field).	`0`
`max_size`	`int`	Maximum number of scalars in the array. For structured dtypes, each element counts as multiple scalars (one per field).	`10`

Examples:

>>> n = numpy_arrays().example()
>>> ak.from_numpy(n)
<Array ... type='...'>