Class np.IndexExpression

Namespace

NumSharp

Assembly

NumSharp.dll

A nicer way to build up index tuples for arrays — the type behind s_ and index_exp. Use those two instances rather than constructing this directly.

public sealed class np.IndexExpression

Inheritance

object

np.IndexExpression

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

Port of NumPy 2.x numpy.lib.index_tricks.IndexExpression. For any index combination, a[indices] equals a[np.index_exp[indices]], but np.s_[…] can be stored in a variable, passed around and reused.

The indexer surface mirrors NDArray's. NumPy's s_ passes ANY index object straight through — np.s_[[1, 2]] is the list, np.s_[..., mask] the tuple — so anything writable as arr[…] must be writable as np.s_[…] too. The three overloads below cover the same vocabulary NDArray's three public indexers do, and each returns the array type that indexer consumes: basic expressions (slices, slice strings, integers) come back as Slice[], while anything advanced — a fancy index array, a boolean mask, or a basic/advanced mix — comes back as object[]. Either can be handed straight to arr[…].

C# divergence — s_ and index_exp return the same thing. NumPy's two instances differ only in maketuple: np.s_[2::2] yields the bare slice(2, None, 2) while np.index_exp[2::2] wraps it as (slice(2, None, 2),). C# has no such distinction at the index site — NDArray's indexers are this[params Slice[]] / this[params object[]], so a lone Slice and a one-element Slice[] are literally the same call. The pair is kept so NumPy code ports across verbatim.

https://numpy.org/doc/stable/reference/generated/numpy.s_.html

Properties

this[Slice[]]

Builds an index tuple out of already-constructed Slice objects — the programmatic spelling of the string form. Integer and string literals convert implicitly, so np.s_[0] and np.s_["1:3", "::2"] land here too.

public Slice[] this[params Slice[] item] { get; }

Parameters

item Slice[]

Property Value

Slice[]

this[object[]]

The ADVANCED-indexing form: captures any index expression NDArray's this[params object[]] accepts, so everything writable as arr[…] is also writable as np.s_[…] and stored for later.

public object[] this[params object[] item] { get; }

Parameters

item object[]

Any mix of Slice, slice-notation string, integer, NDArray (integer fancy index or boolean mask), int[] / long[], bool[] (and rectangular bool[,]), IList sequences, and ITuples.

Property Value

object[]

Remarks

Overload resolution keeps the basic forms on the Slice[] overload — a lone Slice, an int (which converts implicitly to Slice) and a slice string all still return Slice[]. This overload takes over the moment an entry cannot be a Slice: a fancy index array, a mask, or a basic/advanced mix such as np.s_[Slice.All, np.array(new[] {0, 2})]. Both return types feed arr[…] directly, including a NDArray<TDType> (its own indexers do not hide the inherited this[params object[]]).

this[string]

Parses Python index notation — comma-separated slices, indices, ... and newaxis — into the index tuple NumSharp's array indexer consumes.

public Slice[] this[string item] { get; }

Parameters

item string

e.g. "2::2", ":, 0", "..., ::-1", "np.newaxis, 3:".

Property Value

Slice[]

Exceptions

ArgumentException

The notation is malformed.

maketuple

True for index_exp (NumPy always returns a tuple), false for s_. Kept to mirror NumPy's constructor; it has no observable effect in C# because a Slice[] is already how NumSharp spells an index tuple.

public bool maketuple { get; }

Property Value

bool