isMultipleOf() false negatives?

[pateketrueke]

I'm validating this value:

{
  "num": 0.47000000000000003
}

...against the schema:

{
  "type": "object",
  "properties": {
    "num": {
      "type": "number",
      "minimum": 0,
      "maximum": 1,
      "multipleOf": 0.01
    }
  },
  "required": [
    "num"
  ]
}

...but is failing with 'has a remainder' which I think is bit wrong?

Also I tested the same sample/schema with: tv4, z-schema and jayschema and it there's no problem.

I noticed that 0.47 passes isMultipleOf as expected but then the other validators fails...

[LinusU]

...but is failing with 'has a remainder' which I think is bit wrong?

Isn't that how it should work? 🤔

0.47000000000000003 isn't a multiple of 0.01? e.g. 0.47 and 0.48 is though?

[pateketrueke]

Yeah, I think it's correct, but then all other validators are wrong... I mean, I don't know if the others are coercing somehow such long values. That's the question.

I'm want to use is-my-json-valid here too.

But then it happens: such values are generated and only the validation made by this module didn't passed.

TBH probably it's wrong behavior from my side, just I'm investigating. 💣

[LinusU]

Hmm 🤔

A possible solution could be to change value % multipleOf === 0 into Math.abs(value % multipleOf) < Number.EPSILON.

That could potentially let something thru that someone isn't expecting though 🤔

Appreciate the investigations! 🙌

[pateketrueke]

I made a change:

var isMultipleOf = function(name, multipleOf) {
  var res;
  var factor = ((multipleOf | 0) !== multipleOf) ? Math.pow(10, multipleOf.toString().split('.').pop().length) : 1
  if (factor > 1) {
    var factorName = ((name | 0) !== name) ? Math.pow(10, name.toString().split('.').pop().length) : 1
-    if (factorName > factor) res = true
+    if (factorName > factor) res = (factorName / factor) % factor
    else res = Math.round(factor * name) % (factor * multipleOf)
  }
  else res = name % multipleOf;
  return !res;
}

The code I changed was related to if (factorName > factor) res = true which I did not understood properly.

With that change my tests are passing now, without issues.

Also tests from is-my-json-valid seems OK:

1..408
# tests 408
# pass  408

# ok

I wonder if that's fine and could be a PR?

[LinusU]

Hmm, I was trying to figure out why that function was so complicated. Turns out that there is a good reason 😆

[pateketrueke]

So I found a bug? 🤔