Performance degrades when an if-else check is applied on a huge map

[Navgeet]

Consider this schema and these two payloads:
test.schema.json

{
  "type": "object",
  "properties": {
    "foo": {
      "type": "string"
    }
  },
  "if": {
    "properties": {
      "foo": {
        "const": "foo"
      }
    }
  },
  "then": {
    "properties": {
      "bar": {
        "const": "bar"
      }
    }
  }
}

test.json

{
  "foo": "foo",
  "bar": "bar"
}

test-big.json

{
  "foo": "foo",
  "bar": "bar",
  "foo1": "foo",
  "foo2": "foo",
...
  "foo99": "foo",
  "foo100": "foo"
}

Running justify over these payloads (code.zip) 1M times yields these results:

test.json: 14835 ms
test-big.json: 55362 ms

Now this could be due to parsing cost of bigger map, so let's remove the if-then keywords and try again.

test.json: 13994 ms
test-big.json: 38727ms

Hence it is clear that something else is going on.
After running the code through debugger, I believe this is because ConditionalEvaluator's evaluate method is called for every key in the payload. Is this something that can be optimized?

[Navgeet]

For comparison, the same test in Ajv takes same time for both payloads. In this case, I passed parsed json object .

[leadpony]

@Navgeet
Justify is a streaming validator, that means it does not require a whole JSON document to validate in memory and checks it event by event produced by the underlying parser. One of the cons in streaming validator is that we must check all branches of if/then/else at the same time, because we do not know whether if is evaluated as true or false beforehand. In the example you have shown, we may be able to improve the implementation for faster evaluation of if because foo is luckily the first property in the object, but if foo property sits in the last of the object, we must keep evaluating both of if and then until the end of the object.