feat: Initialize project (row-types version)

[OlaoluwaM]

Unofficial Haskell SDK for the cloudevents spec. Written with row types to allow for arbitrary extension attributes as per the spec requirements. This gives us field deduplication at the type level but I fear I may have gone too far off the deep end just for that.

Anyways, this is a bare bones SDK that only allows for:

• Creating a cloud event
• Modifying a cloud event
• Serializing a cloud event
• Producing a Kafka message from a cloud event

[z0isch]

You already know the feelings I have about the fancy haskell stuff here, but I'll try to reiterate it.

I think a simpler design is going to work out better:

data CloudEvent extension data = CloudEvent {
    extension :: extension,
    data :: data,
    ...
    }

data KafkaExtension = KafkaExtension {
    partitionKey :: Text
    }

newtype KafkaCloudEvent data = KafkaCloudEvent (CloudEvent KafkaExtension data)

The tradeoff of this design is that it does not get compiler errors when an extension's name conflicts with the CloudEvent top level properties. Given that a user of this library will most likely not be using CloudEvent directly I don't see this as too big of a downside. The most frequent usage of CloudEvent will be in library maintainers implementing new bindings but those are governed by a spec and the likelihood of a conflict seems non-existent.

[chris-martin]

If you follow our style guide and set the language to GHC2021, you don't need most of these extensions explicitly enabled.

[chris-martin]

I haven't the slightest idea what's going on here. Does this function do something complicated?

[chris-martin]

With great power of types comes great responsibility to write documentation and usage examples

[OlaoluwaM]

It's kinda like zoom from the curricula project

[OlaoluwaM]

Yeah, I should add more documentation. I kinda wanted to get the team's thoughts on the overall approach first

[chris-martin]

I suggest writing some tests and putting something into README.lhs as a next step - Seeing this code do something and produce some output would help me understand what it's for and whether the complexity is appropriate for what it does.

[mjgpy3]

I'm blocking to second AJ's suggestion.

The complexity of these new types far outweighs any benefits I can perceive. I'd also like to not have to learn a new DSL to construct an event.

[OlaoluwaM]

I get your point @mjgpy3 (and @z0isch), while I do agree somewhat, I think the complexity here is pretty isolated in that a user never has to interface with row-types directly but only the mkCloudEventRec function (which takes a regular Haskell records btw) to construct a CloudEventRec then make use of the functions defined in Kafka.hs to construct a kafka message out of said cloudevent representation. There's really never a need to muck around with row-types.

Furthermore this much complexity isn't completely unheard of in the realm of SDKs given it's all for the sake of DevEx. Nevertheless, let me write up some example as @chris-martin suggests to show you what I mean

[OlaoluwaM]

@chris-martin added some more details and an example to the README.md

[github-actions]

Restyled by prettier-markdown

Suggested change

	is the latest version, SDKs should also support v0.3.
	is the latest version, SDKs should also support v0.3.

[github-actions]

Restyled by prettier-markdown

Suggested change

	optionally prefixed with a minus sign.
	optionally prefixed with a minus sign.

[chris-martin]

Well that readme definitely helps make the project not look scary anymore.