time-to-botec/squiggle/node_modules/@stdlib/ndarray/base/broadcast-shapes/README.md
NunoSempere b6addc7f05 feat: add the node modules
Necessary in order to clearly see the squiggle hotwiring.
2022-12-03 12:44:49 +00:00

10 KiB

broadcastShapes

Broadcast array shapes to a single shape.

Usage

var broadcastShapes = require( '@stdlib/ndarray/base/broadcast-shapes' );

broadcastShapes( shapes )

Broadcasts array shapes to a single shape.

var sh1 = [ 8, 1, 6, 1 ];
var sh2 = [ 7, 1, 5 ];

var sh = broadcastShapes( [ sh1, sh2 ] );
// returns [ 8, 7, 6, 5 ]

Notes

  • When operating on two arrays, the function compares their shapes element-wise, beginning with the trailing (i.e., rightmost) dimension. The following are examples of compatible shapes and their corresponding broadcasted shape:

    A      (4d array):  8 x 1 x 6 x 1
    B      (3d array):      7 x 1 x 5
    ---------------------------------
    Result (4d array):  8 x 7 x 6 x 5
    
    A      (2d array):  5 x 4
    B      (1d array):      1
    -------------------------
    Result (2d array):  5 x 4
    
    A      (2d array):  5 x 4
    B      (1d array):      4
    -------------------------
    Result (2d array):  5 x 4
    
    A      (3d array):  15 x 3 x 5
    B      (3d array):  15 x 1 x 5
    ------------------------------
    Result (3d array):  15 x 3 x 5
    
    A      (3d array):  15 x 3 x 5
    B      (2d array):       3 x 5
    ------------------------------
    Result (3d array):  15 x 3 x 5
    
    A      (3d array):  15 x 3 x 5
    B      (2d array):       3 x 1
    ------------------------------
    Result (3d array):  15 x 3 x 5
    
    A      (5d array):  8 x 1 x 1 x 6 x 1
    B      (4d array):      1 x 7 x 1 x 5
    C      (5d array):  8 x 4 x 1 x 6 x 5
    -------------------------------------
    Result (5d array):  8 x 4 x 7 x 6 x 5
    
    A      (5d array):  8 x 1 x 1 x 6 x 1
    B      (1d array):                  0
    -------------------------------------
    Result (5d array):  8 x 1 x 1 x 6 x 0
    
    A      (5d array):  8 x 0 x 1 x 6 x 1
    B      (2d array):              6 x 5
    -------------------------------------
    Result (5d array):  8 x 0 x 1 x 6 x 5
    
    A      (5d array):  8 x 1 x 1 x 6 x 1
    B      (5d array):  8 x 0 x 1 x 6 x 1
    -------------------------------------
    Result (5d array):  8 x 0 x 1 x 6 x 1
    
    A      (3d array):  3 x 2 x 1
    B      (0d array):
    -----------------------------
    Result (3d array):  3 x 2 x 1
    
    A      (0d array):
    B      (3d array):  3 x 2 x 1
    -----------------------------
    Result (3d array):  3 x 2 x 1
    

    As demonstrated above, arrays are not required to have the same number of dimensions in order to be broadcast compatible. Array shapes with fewer dimensions are implicitly prepended with singleton dimensions (i.e., dimensions equal to 1). Accordingly, the following example

    A      (2d array):  5 x 4
    B      (1d array):      4
    -------------------------
    Result (2d array):  5 x 4
    

    is equivalent to

    A      (2d array):  5 x 4
    B      (2d array):  1 x 4
    -------------------------
    Result (2d array):  5 x 4
    

    Similarly, a zero-dimensional array is expanded (by prepending singleton dimensions) from

    A      (3d array):  3 x 2 x 1
    B      (0d array):
    -----------------------------
    Result (3d array):  3 x 2 x 1
    

    to

    A      (3d array):  3 x 2 x 1
    B      (3d array):  1 x 1 x 1
    -----------------------------
    Result (3d array):  3 x 2 x 1
    

    Stated otherwise, every array has implicit leading dimensions of size 1. During broadcasting, a 3 x 4 matrix is the same as a 3 x 4 x 1 x 1 x 1 multidimensional array.

  • Two respective dimensions in two shape arrays are compatible if

    1. the dimensions are equal.
    2. one dimension is 1.

    The two aforementioned rules apply to empty arrays or arrays that have a dimension of size 0. For unequal dimensions, the size of the dimension which is not 1 determines the size of the output shape dimension.

    Accordingly, dimensions of size 0 must be paired with a dimension of size 0 or 1. In such cases, by the rules above, the size of the corresponding output shape dimension is 0.

  • The function returns null if provided incompatible shapes (i.e., shapes which cannot be broadcast with one another).

    var sh1 = [ 3, 2 ];
    var sh2 = [ 2, 3 ];
    
    var sh = broadcastShapes( [ sh1, sh2 ] );
    // returns null
    

    The following are examples of array shapes which are not compatible and do not broadcast:

    A      (1d array):  3
    B      (1d array):  4                   # dimension does not match
    
    A      (2d array):      2 x 1
    B      (3d array):  8 x 4 x 3           # second dimension does not match
    
    A      (3d array):  15 x 3 x 5
    B      (2d array):  15 x 3              # singleton dimensions can only be prepended, not appended
    
    A      (5d array):  8 x 8 x 1 x 6 x 1
    B      (5d array):  8 x 0 x 1 x 6 x 1   # second dimension does not match
    

Examples

var lpad = require( '@stdlib/string/left-pad' );
var broadcastShapes = require( '@stdlib/ndarray/base/broadcast-shapes' );

var shapes;
var out;
var sh;
var i;
var j;

function shape2string( shape ) {
    return lpad( shape.join( ' x ' ), 20, ' ' );
}

shapes = [
    [ [ 1, 2 ], [ 2 ] ],
    [ [ 1, 1 ], [ 3, 4 ] ],
    [ [ 6, 7 ], [ 5, 6, 1 ], [ 7 ], [ 5, 1, 7 ] ],
    [ [ 1, 3 ], [ 3, 1 ] ],
    [ [ 1 ], [ 3 ] ],
    [ [ 2 ], [ 3, 2 ] ],
    [ [ 2, 3 ], [ 2, 3 ], [ 2, 3 ], [ 2, 3 ] ],
    [ [ 1, 2 ], [ 1, 2 ] ]
];

for ( i = 0; i < shapes.length; i++ ) {
    sh = shapes[ i ];
    for ( j = 0; j < sh.length; j++ ) {
        console.log( shape2string( sh[ j ] ) );
    }
    console.log( lpad( '', 20, '-' ) );

    out = broadcastShapes( sh );
    console.log( shape2string( out )+'\n' );
}

C APIs

Usage

#include "stdlib/ndarray/base/broadcast_shapes.h"

stdlib_ndarray_broadcast_shapes( M, **shapes, *ndims, *out )

Broadcasts array shapes to a single shape.

#include "stdlib/ndarray/base/broadcast_shapes.h"
#include <stdint.h>

int64_t N1 = 4;
int64_t sh1[] = { 8, 1, 6, 1 };

int64_t N2 = 3;
int64_t sh2[] = { 7, 1, 5 };

int64_t ndims[] = { N1, N2 };
int64_t *shapes[] = { sh1, sh2 };

int64_t out[] = { 0, 0, 0, 0 };
int8_t status = stdlib_ndarray_broadcast_shapes( 2, shapes, ndims, out );
if ( status != 0 ) {
    // Handle error...
}

The function accepts the following arguments:

  • M: [in] int64_t number of shape arrays.
  • shapes: [in] int64_t** array of shape arrays (dimensions).
  • ndims: [in] int64_t* number of dimensions for each respective shape array.
  • out: [out] int64_t* output shape array.
int8_t stdlib_ndarray_broadcast_shapes( int64_t M, int64_t *shapes[], int64_t ndims[], int64_t *out );

If successful, the function returns 0; otherwise, the function returns -1 (e.g., due to incompatible shapes).

Notes

  • Even if the function is unsuccessful, the function may still overwrite elements in the output array before returning. In other words, do not assume that providing incompatible shapes is a no-op with regard to the output array.

Examples

#include "stdlib/ndarray/base/broadcast_shapes.h"
#include <stdint.h>
#include <stdio.h>
#include <inttypes.h>

int main() {
    int64_t N1 = 4;
    int64_t sh1[] = { 8, 1, 6, 1 };

    int64_t N2 = 3;
    int64_t sh2[] = { 7, 1, 5 };

    int64_t ndims[] = { N1, N2 };
    int64_t *shapes[] = { sh1, sh2 };

    int64_t out[] = { 0, 0, 0, 0 };
    int8_t status = stdlib_ndarray_broadcast_shapes( 2, shapes, ndims, out );
    if ( status != 0 ) {
        printf( "incompatible shapes\n" );
        return 1;
    }
    int64_t i;
    printf( "shape = ( " );
    for ( i = 0; i < N1; i++ ) {
        printf( "%"PRId64"", out[ i ] );
        if ( i < N1-1 ) {
            printf( ", " );
        }
    }
    printf( " )\n" );
    return 0;
}