Module shape 

Layout kernels: broadcast, transpose, squeeze, expand, select, take, concat.

Axes are addressed by dimension name. Layout kernels keep the attributes.

Structs

Slice

A Python-style slice: start:stop:step. Negative indices count from the end.

Enums

PadMode

How pad fills the new positions, like the mode of numpy.pad.

Selector

A selection on one axis for isel.

Functions

broadcast_arrays

Broadcast all arrays onto their shared dimensions, see Dimensions::broadcast.

broadcast_to

Broadcast a onto target without a copy where possible.

concat

Join arrays along an existing dimension, like numpy.concatenate.

expand_dims

Insert a new axis of size 1 at position axis, without a copy.

flip

Reverse the order along dim, like numpy.flip.

isel

Select by position on one or more axes, like xarray isel, in one copy.

move_dim

Move dimension name to position to.

pad

Add before and after positions along dim, like numpy.pad.

rename_dims

Rename dimensions without a copy.

repeat

Repeat each element repeats times along dim, like numpy.repeat.

roll

Roll the elements along dim by shift, like numpy.roll.

squeeze

Remove size-1 axes without a copy. None removes every size-1 axis.

stack

Join arrays along a new dimension at position axis, like numpy.stack.

take

Take indices along dim, like numpy.take(a, indices, axis).

tile

Repeat the whole array reps times along dim, like numpy.tile.

transpose

Reorder the axes. order must name every dimension once.

transpose_reversed

Reverse the axis order, like ndarray.T.