Incorrect parsing of integer numbers at scientific form

[sad-fat-cat]

When you try to parse numbers like 1e6 as int or long, parser ignore e part, e.g

import zio.json.*

JsonDecoder.int.decodeJson("1e6") // 1
JsonDecoder.long.decodeJson("1e6") // 1

JsonDecoder.double.decodeJson("1e6") // 1000000.0

Also, e part remains in the buffer, so produce very strange error on parse, for example

import zio.json.*

final case class Test(foo: Long)
object Test {
    implicit val decoder: JsonDecoder[Test] = DeriveJsonDecoder.gen
}

JsonDecoder[Test].decodeJson("""  { "foo": 1e6 } """) // Left((expected ',' or '}' got 'e'))

This looks like to be incorrect behavior, I would expect the number to be parsed and then converted to an integer.

[sad-fat-cat]

Ok, you can use a funny workaround like this

implicit val longDecoder: JsonDecoder[Long] = JsonDecoder.double.map(_.toLong)

but that's not something you want to see in production code ¯\(ツ)/¯

[plokhotnyuk]

@sad-fat-cat zio-json doesn't check for remaining input in buffer and reported error message for the invalid character looks quite reasonable when parsing integers

The message can be improved by adding a field name to the JSON path: .foo(expected ',' or '}' got 'e').

Please be aware that your workaround do silent truncation of parsed doubles, and will loose up to 11 lower bits for big long values, like Long.MaxValue.

A correct workaround could be:
implicit val longDecoder: JsonDecoder[Long] = JsonDecoder.bigDecimal.map(_.longValueExact)