add ?~> and ?:= combinators

[kanterov]

Example usage

instance encodeJsonOption :: EncodeJson Option where
  encodeJson (Option o)
    =   "backgroundColor" ?:= o.backgroundColor
    ?~> "color"           ?:= o.color
    ?~> jsonEmptyObject

If backgroundColor is Nothing, it results into

{
  "color": "#ff0000"
}

[jdegoes]

These combinators are not necessary since Maybe already has EncodeJson and DecodeJson instances. Or am I missing something???

[kanterov]

It has different behaviour, for this example it would result in:

{
  "color": "#ff0000",
  "backgroundColor": null
}

[jdegoes]

Ah, I see, so this is undefined. Is the goal to compact the JSON objects as much as possible? Or just an alternate take on what Maybe should be serialized into?

[kanterov]

My goal was to preserve semantic, in Javascript, there is a difference between {a: undefined} and {}. In my case, I was working with javascript library that somehow merges JSON object into its own. Most of the merge implementations that I've seen will append a new value no matter if it's null or undefined.

I've seen similar combinators in original argonaut https://github.com/argonaut-io/argonaut/blob/master/src/main/scala/argonaut/Json.scala#L159.

[jdegoes]

OK, looks good. Can you add another example to README.md on how to use these combinators?

[kanterov]

OK, I will wait for #10

[jdegoes]

@kanterov #10 has been merged, can you updated this so we can merge it in?

[joneshf]

Before this gets merged, I was thinking about these operators the other day. Maybe it would make sense to keep the same combinator names as the scala version, including changing the names of the combinators that are already there. It would probably be less jarring to people coming here from there. They should all be valid in ps.

[jdegoes]

That sounds good to me.

[kanterov]

I'm not quite familiar with haskell infix operator naming conventions, as I got the idea from sources question mark should be placed next to Maybe a, so convention is

(?xx) :: forall a. Maybe a -> a -> a
(xx?) :: forall a. a -> Maybe a -> a
(?xx?) :: forall a. a -> Maybe a -> Maybe a -> a

There is ?>>= that has the same convention:

(?>>=) :: forall a b. Maybe a -> String -> Either String a

What do you think? @jdegoes @joneshf

So changes are (?:=) -> (:=?), (?>>=) (?~>) remains the same

[kanterov]

Probably prefixing with ? makes sence for Scala for proper operator precedence

[kanterov]

@joneshf @jdegoes ping

[jdegoes]

@kanterov Can you resolve conflicts?

[thomashoneyman]

Closing as resolved by purescript-contrib/purescript-argonaut-codecs#38