Encoding `Data.BigInteger`s encodes as string about plutus-apps

[luigy]

@emiflake thanks for the report. I did a test in plutus-playground-client and I'm getting a different behavior with encodeJSON if that's an option. Here is what I tried roughly:

import Wallet.Emulator.Wallet
import Data.BigInteger as BI
import Foreign.Generic (encodeJSON)

let
  w :: Wallet
  w = wrap ({ getWallet : BI.fromInt 1 })
in [ div_ [ text $ encodeJSON $ BI.fromInt 100]
   , div_ [ text $ encodeJSON w ]
   ]

[emiflake]

Yeah, I understand. RFC7159 says "This specification allows implementations to set limits on the range and precision of numbers accepted." So it all comes down to what the specific implementations support. Aeson supports decoding large integers into Integer, and since we can get PureScript to produce those too, it's fine for those purposes. It's unfortunate to stringly-type BigIntegers just out of fear for other implementations not supporting them. I guess in the case of PureScript we seem to have "reasonable" way to choose which encoding we prefer.

[emiflake]

Yeah, it appears that encoding with encodeJSON does produce correct stringified results (similar to my approach). Manually encode-ing and then using JSON.stringify doesn't (this is what for example Affjax does when you pass it a Json)

[silky]

If it helps, I think this is related to a quirk in JSON; for example if you just have a large integer, then tools like jq choose to interpret it as a floating-point:

echo '{"a":9128730192384570193485710934857}' | jq
{
  "a": 9.12873019238457e+30
}

which is frustrating :) So in some sense it really isn't safe to encode BigIntegers as "integers" in JSON at all; they probably need to be wrapped in strings, so maybe that's why some libraries are making this choice.