[dependency upgrades] Poison replaced with jason and httpoison upgrade

[izelnakri]

No description provided.

[ayrat555]

Is there any benefit of using Jason over Poison?

[izelnakri]

Yes allegedly its faster and used instead of poison on ecto v3

[InoMurko]

Is Poison used more then Jason? If yes, that would mean most projects that include exthereum as a dependency would now have two json encoders/decoders. The PR should rather chose an approach that would make the project independent of a particular choice.

[hayesgm]

It would be great for us to benchmark. The only concern is that projects that use this are now likely going to pull in a second JSON parsing library (or, on the other hand, go from two to one...)

[InoMurko]

What if we do something like Phoenix does with pluggable decoders?
https://github.com/elixir-plug/plug/blob/master/lib/plug/parsers.ex#L271

[izelnakri]

Yes, I thought about this as well, ecto does something similar: https://github.com/elixir-ecto/ecto/blob/master/lib/ecto/json.ex#L9.

We can do the encoder thing as a seperate potential feature, however since ethereumex is not as widely used as ecto and there are several json libraries out there, do we really want to support all of them? They might have different APIs, options etc.

I dont mind if people have two json dependencies in their projects through subdependency resolution since this library will be mostly used in the application layer of a server. We should just use the fastest if they both have the good error messages,jason will eventually take over, as it seems for now.

[InoMurko]

Did a quick research and found out that Credo has jason as a dependency. Which means most projects that use Credo (like OmiseGO ewallet and OMG Network), will have Jason already there.