Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

How can I easily measure the complexity of a JSON object?

If I want to compare a range of API responses for the complexity of the response (as a proxy for how much effort it likely takes to parse and validate the response) is there any existing tool or library that can do that pretty effectively? or a simple bit of code?

Ideally something that prints out a quick report showing how deep and wide the whole structure is, along with any other metrics that might be useful.

like image 712
Jun-Dai Bates-Kobashigawa Avatar asked Nov 06 '22 05:11

Jun-Dai Bates-Kobashigawa


1 Answers

A heuristic is to simply count the number of {, }, [, and ] characters. Of course this is only a heuristic; under this method a json object like { value: "{[}{}][{{}{}]{}{}{}[}}{}{" } would be considered overly complex even though its structure is very straightforward.

let guessJsonComplexity = (json, chars='{}[]')) => {
  let count = 0;
  for (let char in json) if (chars.includes(char)) count++;
  return count / (json.length || 1);
};

You would go with this answer if speed is very important.

You'll almost certainly need to parse the json if you want a more concise answer!

We could also consider another approach. Consider assigning a "complexity score" for every possible phenomenon that can happen in json. For example:

  • A string s is included; complexity score: Math.log(s.length)
  • A number n is included; complexity score: Math.log(n)
  • A boolean is included; complexity score: 1
  • An array is included; complexity score: average complexity of elements + 1
  • An object is included; complexity score: average complexity of values plus average complexity of keys + 1

We could even pick out distinct relationships, like "an object is included in an array", or "an array is included in an array", etc, if we want to consider some of these as being more "complicated" than others. We could say, for example, that negative numbers are twice as "complicated" as positive numbers, if that's how we feel.

We can also consider a "depth factor", which makes elements count for more the deeper they go.

If we define how to score all these phenomena, we can write a function that processes json and applies such a score:

let isType = (val, Cls) => val != null && val.constructor === Cls;
let getComplexity = (json, d=1.05) => {
  
  // Here `d` is our "depth factor"
  
  return d * (() => {

    // Take the log of the length of a String
    if (isType(json, String)) return Math.log(json.length);

    // Take the log of (the absolute value of) any Number
    if (isType(json, Number)) return Math.log(Math.abs(json));

    // Booleans always have a complexity of 1
    if (isType(json, Boolean)) return 1;

    // Arrays are 1 + (average complexity of their child elements)
    if (isType(json, Array)) {
      let avg = json.reduce((o, v) => o + getComplexity(v, d), 0) / (json.length || 1);
      return avg + 1;
    }

    // Objects are 1 + (average complexity of their keys) + (average complexity of their values)
    if (isType(json, Object)) {
      // `getComplexity` for Arrays will add 1 twice, so subtract 1 to compensate
      return getComplexity(Object.keys(json), d) + getComplexity(Object.values(json), d) - 1;
    }

    throw new Error(`Couldn't get complexity for ${json.constructor.name}`);
    
  })();
  
};

console.log('Simple:', getComplexity([ 'very', 'simple' ]));
console.log('Object:', getComplexity({
  i: 'am',
  some: 'json',
  data: 'for',
  testing: 'purposes'
}));
console.log('Complex:', getComplexity([
  [ 111, 222, 333, 444 ],
  [ 'abc', 'def', 'ghi', 'jkl' ],
  [ [], [], {}, {}, 'abc', true, false ]
]));
console.log('Deep:', getComplexity([[[[[[ 'hi' ]]]]]]));

If you want to know more detailed information on children of a large json object, you can simply call getComplexity on those children as well.

like image 165
Gershom Maes Avatar answered Nov 11 '22 02:11

Gershom Maes