Strict comparison?

[koosvanderkolk]

How to do type-safe comparison in this (excellent) library?

math.evaluate('1==true'); // true

I expect this one to return 'false', as 1 !== true.

[josdejong]

Thanks for your question Koos. Mathjs uses type-coercion when comparing values, resulting in expressions like 1==true and "2"==2 to return true. (Similar to JavaScripts ==, and unlike JavaScripts ===).

There are no functions implemented to do strict equality checks. If you need this, you can either create and import a function like strictEqual yourself, or override the built-in equal function with strict behavior.

[koosvanderkolk]

Thanks for the swift response Jos!

Is it perchange on the roadmap to allow for type-safe comparision in expressions? I mean that "1===true" is also allowed as an expression?

I indeed thought of writing a function myself, but... I do not think it can be done in such a way that

math.evaluate('1==true');

will use my custom 'equal' function right? So it should be done something like this:

math.evaluate('strictEqual(1, true)');

[josdejong]

Is it perchange on the roadmap to allow for type-safe comparision in expressions? I mean that "1===true" is also allowed as an expression?

This is not on the roadmap, but if there is enough need for it we can consider it and make a plan.

Side note: I think the official name for this is subject "strict equality comparison" (vs. type-coersion), it does not directly have to do with type safety.

I indeed thought of writing a function myself, but... I do not think it can be done in such a way that

math.evaluate('1==true');

will use my custom 'equal' function right? So it should be done something like this:

math.evaluate('strictEqual(1, true)');

When you create a mathjs instance, you can pass your own factory function for equal (and unequal) instead of using the build-in functions, something like:

import { create, all, factory } from 'mathjs'

function strictEqual(a, b) {
  return a === b
}

const math = create({
  ...all,
  
  // pass our own implementation for equal 
  createEqual: factory('equal', [], () => {
    return strictEqual
  })
})

function testEvaluate (expr) {
  console.log(expr, '=', math.evaluate(expr))
}

testEvaluate('"2"==2') // false

// deepEqual uses equal internally
testEvaluate('deepEqual("2", 2)') // false

[koosvanderkolk]

Thanks!

Just asking for a friend: suppose one is bound to ES5 (so no 'import' etc) how would that work?

This does not work:

function strictEqual(a, b) {
  return a === b;
  };

var my_math = window.math.create({
  // pass our own implementation for equal
  createEqual: window.math.factory('equal', [], () => {
    return strictEqual;
  })
});

console.log(my_math.evaluate('"2"==2'));

[koosvanderkolk]

In the meantime I did find a solution, namely using math.import, with the second argument being {override: true}

window.math.import({
          "smaller": function (a, b) {
            return (a < b);
          }}, {override:true});

[josdejong]

In the meantime I did find a solution, namely using math.import, with the second argument being {override: true}

window.math.import({
          "smaller": function (a, b) {
            return (a < b);
          }}, {override:true});

hm, yeah, that works for overriding smaller or equal, but it will not update the functions which have a dependency on equal, like deepEqual. It looks like there are some limitations trying to do this from the bundled version of mathjs.

[koosvanderkolk]

Mmm... for the time being I think we are fine, but this might indeed pose a problem in the feature.

I am also glad that it seems to be a limitation of the bundled version (and not a limitation of my JS skills ;-), spend hours on it because I thought I was overseeing something obvious...

P.S. Great testimony on your personal home page! <><

[josdejong]

Thanks 😄

I'll give this some more thought. Ideally, when overriding an existing function on import, all functions that use the dependency should be re-created again. I thought this was the case but clearly it's not. Have to look into this.

[josdejong]

The problem of functions not being recreated is being addressed in #1975. Will close this issue now.