Simple Math Programming Language (SMPL) for the Web

SMPL is a math-oriented programming language that can be interpreted in the browser. Its primary use is for the mathe:buddy app.

SMPL's syntax is basically a subset of JavaScript, but extended with intrinsic mathematical data types (e.g. matrices) and operator overloading for these types. SMPL uses MathJS (and later on e.g. TensorFlow) for core calculations.

Many concepts (and also parts of the source code) are taken from the Simple E-Learning Language SELL.

Compared to SELL, SMPL is Turing Complete: It allows e.g. loops and conditions.

SMPL is easily extendible. Refer to the Developer Notes below.

Notes related to the "mathe:buddy" project

We will also support writing code for random questions in Python, Octave, Maxima and SageMath.

SMPL has the following advantage:

• No server is required. Code is executed at client-side. There is no need to install anything
• Simple syntax with intrinsic data types for math
• No need to import math libraries explicitly

Example

Copy the following code to https://npm.runkit.com/ and run it!

var SMPL = require('@mathebuddy/mathebuddy-smpl');

let src = `
// create two (3x3)-matrices A and B, with random
// (integral) entries in range [-5,5] without zero.
let A:B = randZ<3,3>(-5,5);
let C = A * B;
`;

let variables = SMPL.interpret(src);

for (let v of variables) {
  console.log(v.id + ' = ' + v.value.toString());
}

Installation

npm install @mathebuddy/mathebuddy-smpl

Language Definition

SMPL is (mostly) an imperative, typed language. Its base syntax is derived from JavaScript.

Programs

Data Types

boolean (BOOL), integer (INT), rational (RATIONAL), real (REAL), term (TERM), matrix (MATRIX), vector (VECTOR), set (SET), complex (COMPLEX).

Declarations

Declarations are initiated with keyword let, followed by an identifier and finally assigned expression by =. The expression in the right-hand side is mandatory to derive the data type.

Example

let x = 5;
let y = 7, z = 9.1011;
let u = rand(5);
let v = zeros<2,3>();
let a : b = rand<3,3>(-2, 2);
let c :/ d :/ e = rand<3,3>(-2, 2);

• Variables x and u are integral. The value for u is randomly chosen from set {0,1,2,3,4,5}.
• Variable z is a real valued.
• Variables x and z are declared together (without any semantical relation). The notation is equivalent to let y=7; let y=9.1011;
• Variables v, a, b, c and d are matrices. v is a zero matrix with two rows and three columns.
• Matrices a and b consist of randomly chosen, integral elements in range [-2,2].
• The colon separator : evaluates the right-hand side to each of the variables: let a = rand<3,3>(-2, 2); let b = rand<3,3>(-2, 2);. Accordingly, elements for a and b are drawn individually.
• Separator :/ guarantees that no pair of matrices c, d and e is equal.

Expressions

List of operators, ordered by increasing precedence:

Base data types are evaluated at compile-time. Properties like e.g. matrix dimensions are evaluated at runtime. Thus, a RuntimeError is thrown if e.g. two matrices with a different number of rows are added.

Example

let x = 1.23 * sin(y) + exp(u + 2*w);
let C = A * B^T;
let d = det(C);

• The examples assumes, that variables y, u, w, A, B, C have been declared before usage.

Conditions

Example

let s = 0;
if (x > 0) {
  s=1;
}
else if (x < 0) {
  s = -1;
}
else {
  x=0;
}

Loops

Example

while (x > 0) {
  // body
}

Example

do {
  // body
} while (x > 0);

Example

for (let i = 0; i < 5; i++) {
  // body
}

Functions

A function consists of a header and a body:

• The header declared the name of the function, its parameter names and types and the return type.
• The body is represented by a list of statements and returns a result that is compatible to the return type.

Example

function f(x: INT, y: INT): INT {
  return x + y;
}

let y = f(3, 4);

Grammar

The following formal grammar (denoted in EBNF) is currently implemented.

program = { statement };
statement = declaration | if | for | do | while | switch | function | return | break | continue | expression EOS;
declaration = "let" ID { ":" ID } "=" expr { "," ID { ":" ID } "=" expr } EOS;
expression = or { ("="|"+="|"-="|"/=") or };
or = and { "||" and };
and = equal { "&&" equal };
equal = relational [ ("=="|"!=") relational ];
relational = add [ ("<="|">="|"<"|">") add ];
add = mul { ("+"|"-") mul };
mul = unary { ("*"|"/") unary };
unary = unaryExpression [ unaryPostfix ];
unaryExpression = INT | IMAG | REAL | "(" expr ")" | ID | "-" unary;
unaryPostfix = "++" | "--" | [ "<" [ unary { "," unary } ] ">" ] "(" [ expr { "," expr } ] ")" | "[" expr "]";
for = "for" "(" expression ";" expression ";" expression ")" block;
if = "if" "(" expression ")" block [ "else" block ];
block = statement | "{" { statement } "}";
do = "do" block "while" "(" expr ")" EOS;
while = "while" "(" expr ")" block;
switch = "switch" "(" expr ")" "{" { "case" INT ":" { statement } } "default" ":" { statement } "}";
function = "function" ID "(" [ ID { "," ID } ] ")" block;
return = "return" [ expr ] EOS;
break = "break" EOS;
continue = "continue" EOS;

Developer Notes

File src/prototypes.ts declares function prototypes. These must be implemented for the interpreter in file src/interpret.ts.

Example: Create support for a function abs(..) for real and complex numbers.

• Insert the following lines into const functions in file src/prototypes.ts:

  abs(x:REAL):REAL -> _absReal;
  abs(x:COMPLEX):REAL -> _absComplex;
  

Since the function is overloaded with two different types (and JavaScript does not support function overloading), each must be mapped onto a different function ID (_absReal and _absComplex).

• Add two private methods to class SMPL_Interpreter in file src/interpret.ts:

  private _absReal(x: number): number {
    return Math.abs(x);
    // direct implementation: return x < 0 ? -x : x;
  }
  
  private _absComplex(x: Complex): number {
    const x_mathjs = x.toMathjs();
    return mathjs.abs(x_mathjs) as number;
    // direct implementation: return Math.sqrt(x.real*x.real + x.imag*x.imag);
  }

The first method uses Vanilla JavaScript function Math.abs(..) to calculate the absolute values of a real number.

The second method uses MathJS to calculate the absolute value of a complex number. Classes Complex, Matrix, ... provide type conversion methods.

Any third party math library written (or compiled to) JavaScript may be included, but increases the download size of the library (currently less than 200 kiB).