Universe

Numty

Numeric Typst

A library for performing mathematical operations on n-dimensional matrices, vectors/arrays, and numbers in Typst, with support for broadcasting and handling NaN values. Numty’s broadcasting rules and API are inspired by NumPy.

#import "numty.typ" as nt

// Define vectors and matrices
#let a = (1, 2, 3)
#let b = 2
#let m = ((1, 2), (1, 3))

// Element-wise operations with broadcasting
#nt.mult(a, b)  // Multiply vector 'a' by scalar 'b': (2, 4, 6)
#nt.add(a, a)   // Add two vectors: (2, 4, 6)
#nt.add(2, a)   // Add scalar '2' to each element of vector 'a': (3, 4, 5)
#nt.add(m, 1)   // Add scalar '1' to each element of matrix 'm': ((2, 3), (2, 4))

// Dot product of vectors
#nt.dot(a, a)   // Dot product of vector 'a' with itself: 14

// Handling NaN cases in mathematical functions
#calc.sin((3, 4)) // Fails, as Typst does not support vector operations directly
#nt.sin((3.4))    // Sine of each element in vector: (0.14411, 0.90929)

// Generate equally spaced values and vectorized functions
#let x = nt.linspace(0, 10, 3)  // Generate 3 equally spaced values between 0 and 10: (0, 5, 10)
#let y = nt.sin(x)              // Apply sine function to each element: (0, -0.95, -0.54)

// Logical operations
#nt.eq(a, b)      // Compare each element in 'a' to 'b': (false, true, false)
#nt.any(nt.eq(a, b)) // Check if any element in 'a' equals 'b': true
#nt.all(nt.eq(a, b)) // Check if all elements in 'a' equal 'b': false

// Handling special cases like division by zero
#nt.div((1, 3), (2, 0))  // Element-wise division, with NaN for division by zero: (0.5, float.nan)

// Matrix operations (element-wise)
#nt.add(m, 1)  // Add scalar to matrix elements: ((2, 3), (2, 4))

// matrix
#nt.transpose(m)  // transposition
#nt.matmul(m,m) //  matrix multipliation
#nt.matmul(c(1,2), r(2,3)) //  colum vector times row vector multiplication.
#np.trace(m) // trace
#np.det(m) /2x2 determinant
 
// printing
#nt.print(m, " != " , (1,2))  // long dollar print, see in pdf 
#nt.p(m, " != " , (1,2))  //  short long print print, see in pdf

Since vesion 0.0.4 n-dim matrices are supported as well in most operations.

Supported Features:

Dimensions:

Numty uses standard typst list as a base type, most 1d operations like dot are suported directly for them.

For matrix specific operations we use 2d arrays / nested arrays, that are also the standard typst list, but nested like in: ((1,2), (1,1)).

For convenience you can create column or row vectors with the #nt.c and #nt.r functions.

#import "numty.typ" as nt
#import "numty.typ": c, r

#let a = (1,2,3)
#let b = (3,2,1)
#c(..a) // ((1,),(2,),(3,)) 
#r(..b) // ((3,2,1),)
#nt.matmul(c(..a), r(..b)) // column @ row

Logic Operations:

#let a = (1,2,3)
#let b = 2

#nt.eq(a,b)  // (false, true, false)
#nt.all(nt.eq(a,b))  // false
#nt.any(nt.eq(a,b))  // true

Math operators:

All operators are element-wise, traditional matrix multiplication is not yet supported.

#nt.add((0,1,3), 1)  // (1,2,4)
#nt.mult((1,3),(2,2)) // (2,6)
#nt.div((1,3), (2,0)) // (0.5,float.nan)

Algebra with Nan handling:

#nt.log((0,1,3)) //  (float.nan, 0 , 0.47...)
#nt.sin((1,3)) //  (0.84.. , 0.14...)

Vector operations:

Basic vector operations

#nt.dot((1,2),(2,4))  //  9
#nt.normalize((1,4), l:1) // (1/5,4/5)

Others:

Functions for creating equally spaced indexes in linear and logspace, usefull for log plots, to sample in acordance to the logarithmic space.

#nt.linspace(0,10,3) // (0,5,10)
#nt.logspace(1,3,3)
#nt.geomspace(1,3,3)

Matrix

#nt.matmul(m,m )              // matrix multiplication
#nt.det(((1,3), (3,4)))       // only 2x2 supported for now
#nt.trace(((1,3), (3,4)))     // trace of square matrix
#nt.transpose(((1,3), (3,4))) // matrix transposition

Printing

Numty supports $ printing to the pdf of numerical matrices, both long and short format.

#nt.print((1,2),(4,2)))  // to display in the pdf
#nt.p((1,2),(4,2)), " random string ")     // to display in the pdf