Tensor methods
Tensors are rectangular rank-N exact collections. Selectors and axis numbers are one-based.
Method reference
| Full syntax | Result | Meaning |
|---|---|---|
tensor.Shape() |
Tuple |
Return dimension sizes. |
tensor.Rank() |
Integer |
Return the number of axes. |
tensor.Size() |
Integer |
Return total cell count. |
tensor.Get(selectors...) |
any \| TensorView |
Read a cell or selection. A selector tuple is also accepted. |
tensor.Set(selectors..., value) |
Tensor |
Return a copy with selected cells assigned. |
tensor.Set!(selectors..., value) |
Tensor |
Mutate selected cells. |
tensor.Reshape(shapeTuple) |
Tensor |
Change shape without changing total size. |
tensor.Flatten() |
Tensor |
Return a rank-1 tensor in row-major order. |
tensor.Transpose() |
TensorView |
Swap the axes of a rank-2 tensor. |
tensor.Permute(orderTuple) |
TensorView |
Reorder rank-N axes. |
tensor.Map((value, indexTuple, tensor) -> result) |
Tensor |
Transform every cell. |
tensor.Fill!(value) |
Tensor |
Mutate every cell. |
tensor.Sum() |
number-like | Sum non-hole cells. |
tensor.Mean() |
number-like \| null |
Return arithmetic mean, or null for an empty tensor. |
tensor.Dot(other) |
number-like | Dot two equal-length rank-1 tensors. |
tensor.MatMul(other) |
Tensor |
Multiply compatible rank-2 tensors. |
tensor.Reduce((acc, value, indexTuple, tensor) -> next, initial?) |
any |
Fold cells; default accumulator is a hole-filled mutable tensor of the same shape. |
tensor.Iterator() |
Iterator |
Traverse cells in row-major order. |
tensor.CheckTraits() |
1 \| null |
Validate attached semantic traits. |
Shape, selection, and views
t := {:2x2: 1, 2; 3, 4 };
t.Shape().Get(1) ##@ == 2;
t.Rank() ##@ == 2;
t.Size() ##@ == 4;
t.Get(2, 1) ##@ == 3;
t.Get({: 1, 2 }) ##@ == 2;
t.Set(1, 2, 9).Get(1, 2) ##@ == 9;
t.Reshape({: 4 }).Shape().Get(1) ##@ == 4;
t.Flatten().Get(3) ##@ == 3;
t.Transpose().Get(1, 2) ##@ == 3;
t.Permute({: 2, 1 }).Get(2, 1) ##@ == 2;Mutation, arithmetic, callbacks, and traversal
t := {:2x2: 1, 2; 3, 4 }; t.Set!(1, 2, 9); t.Get(1, 2) ##@ == 9;
t := {:2x2: 1, 2; 3, 4 }; t.Fill!(5); t.Sum() ##@ == 20;
{:2x2: 1, 2; 3, 4 }.Sum() ##@ == 10;
{:2x2: 1, 2; 3, 4 }.Mean() ##@ == 5/2;
{:3: 1, 2, 3 }.Dot({:3: 4, 5, 6 }) ##@ == 32;
product := {:2x3: 1, 2, 3; 4, 5, 6 }.MatMul({:3x2: 7, 8; 9, 10; 11, 12 });
product.Get(2, 2) ##@ == 154;
{:2x2: 1, 2; 3, 4 }.Map((value, index) -> value * index.Get(1)).Get(2, 2) ##@ == 8;
reduced := {:2x2: 1, 2; 3, 4 }.Reduce((acc, value, index) -> acc.Set!(index, value * 10));
reduced.Get(2, 1) ##@ == 30;
it := {:2x2: 1, 2; 3, 4 }.Iterator(); it.Next(3) ##@ == 3;
{:1: 1 }.CheckTraits() ##@ == 1;Transpose is currently rank-2 only. Permute accepts one occurrence of every axis number.