Custom rule for decoding JSON input with Jason

If you want precision in decimals, you can use fixed point arithmetic.

Simple idea - represent everything in integers - multiply everything by 10 ⁿ where n is the number of precision digits .

# i want 2 digits precision so i will multiply with 100
0 * 100 = 0
0.5 * 100 = 50
3.1415 * 100 = 314

{
  "fx_rate": 101 # 1.01 * 100
}

This technique is used in finance, gaming, graphics, etc

Have you seen the floats: :decimals option to Jason.decode/2? That might be what you’re looking for.

uses Decimal.new/1 to parse the binary into a Decimal struct with arbitrary precision.

Edit: For Plug, that looks like this:

  plug Plug.Parsers,
    parsers: [:json],
    json_decoder: {Jason, :decode!, [[floats: :decimals]]}

There is the floats: decimal option, but out of curiosity, would it ever really be a problem if you just use Float.to_string.

> Float.to_string(0.1)
"0.1"
> Float.to_string(0.01)
"0.01"

While “The closest representable number to 0.1 is 0.1000000014” it is still converted to 0.1

Thank you guys! I missed the floats: :decimal!

This plug example has now been added to the docs.