Every abstract array in Julia has a shape that determines how that array behaves with respect to indexing and certain array operations.
The shape of an array is fully specified by a tuple of lengths called its dimensions, which tell you how long the array is in each direction.
You can see the dimensions of an array using size:
size([1 2 3; 4 5 6]) → (2, 3)size(1:4) → (4,)Reshaping an array means changing its dimensions. For example, here's a simple four-element range:
1:4 → 1:4The range is shaped like a one-dimensional column. We can reshape it to a two-dimensional square with reshape:
reshape(1:4, (2, 2)) → [1 3; 2 4]We can reshape to a row vector the same way, keeping in mind that row vectors in Julia are represented by 1-by-n arrays:
reshape(1:4, (1, 4)) → [1 2 3 4]Reshaping an array preserves the product of its dimensions — the number of elements stays unchanged.
As a convenience you can pass : as one of the dimensions to ask reshape to compute the value of that dimension for you:
reshape(1:8, (2, :)) → [1 3 5 7; 2 4 6 8]In addition to reshape there are several convenience functions for reshaping arrays in particular ways.
The function vec reshapes into a one-dimensional column vector:
vec([1 3; 2 4]) → [1, 2, 3, 4]The function dropdims removes dimensions of length 1, known as singleton dimensions. For example, you can turn a 1x2 array into a vector of length 2 by dropping the first dimension:
dropdims([1 2 3 4], dims=1) → [1, 2, 3, 4]reshape itself has a short form — you can omit the () around the dimensions and pass them individually:
reshape(1:4, 2, 2) → [1 3; 2 4]You can exchange or otherwise permute the dimensions of a matrix using permutedims:
permutedims([1 2; 3 4]) → [1 3; 2 4]You can rotate the contents of a matrix by 90° or 180° degrees:
rotl90([1 2; 3 4]) → [2 4; 1 3]rotr90([1 2; 3 4]) → [3 1; 4 2]rot180([1 2; 3 4]) → [4 3; 2 1]now() → 2018-09-15T06:16:00.983