Class np.AxisConcatenator

Namespace

NumSharp

Assembly

NumSharp.dll

Translates slice expressions to concatenation along an axis — the machinery shared by r_ and c_.

public abstract class np.AxisConcatenator

Inheritance

object

np.AxisConcatenator

Derived

np.CClass

np.RClass

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

Port of NumPy 2.x numpy.lib.index_tricks.AxisConcatenator. See np.RClass for the full usage documentation.

How Python's a[1:5:2] is spelled in C#. C# has no slice literal, so a slice is written as a string: np.r_["1:5:2"]. Strings do double duty here, exactly as in NumPy, and are told apart by a colon — a string containing ':' is a slice expression, a string without one is a NumPy special directive ("r", "c", "-1", "0,2", "1,2,0"). The two grammars are disjoint in NumPy too — a directive never contains a colon because Python's own syntax supplies the slices — so every NumPy expression transcribes verbatim. A slice string may hold several comma-separated slices (np.r_["1:3, 5:8"] ≙ np.r_[1:3, 5:8]), and a Slice or Slice[] object works too (so np.s_[…] composes).

Weak scalars (NEP50). NumPy distinguishes a Python literal (5, weak — adopts the other operand's dtype) from a NumPy scalar (np.int64(5), strong). C# has no such split, so NumSharp maps it by type: bool / the eight integer primitives / float / double / Complex are weak (the C# literal is the Python literal), while char, Half and decimal — which have no Python literal — plus every NDArray (including 0-d) are strong. To force a strong scalar, wrap it: np.r_[int8Array, NDArray.Scalar(1L)] yields int64 where np.r_[int8Array, 1L] yields int8. As in NumPy, a weak integer that does not fit the resolved dtype raises OverflowException rather than wrapping.

Divergence — no bmat branch. NumPy routes a single bare string key into matrixlib.bmat, which resolves the words in it against the CALLER'S Python frame (sys._getframe().f_back) as variable names. C# has no equivalent, and the branch raises NameError for every string literal anyway (np.r_['1 2; 3 4'] → "name '1' is not defined"), so NumSharp gives a lone string the same reading as any other entry.

Fields

_axis

Axis to concatenate along (NumPy's axis).

protected readonly int _axis

Field Value

int

_matrix

Whether the result is coerced to a 2-D "matrix" (NumPy's matrix).

protected readonly bool _matrix

Field Value

bool

_ndmin

Minimum dimensionality forced onto each entry (NumPy's ndmin).

protected readonly int _ndmin

Field Value

int

_trans1d

Where upgraded entries put their original last axis (NumPy's trans1d).

protected readonly int _trans1d

Field Value

int

Properties

Count

NumPy's len, which is 0 for these objects.

public int Count { get; }

Property Value

int

this[object[]]

Expands the index expression and concatenates the pieces.

public NDArray this[params object[] key] { get; }

Parameters

key object[]

Slice-expression strings, special-directive strings, Slice objects, NDArrays, C# scalars, arrays and collections — in any mix.

Property Value

NDArray