Event Handlers

contract-created.ts

//filepath: src/handlers/messages/contract-created.ts
import {vaidationMiddleware} from '@helloextend/api-utils'

//Handle a single Cloud Event. 


function contractCreatedHandler(ev: CloudEvent, ctx: Context) {
  const messageBody = JSON.parse(ev.message.body) //Or something. Whatever the cloudEvent specification says

  const isRetry = ctx.get<number>('event.metadata.retries') > 0
  const lastRetry = ctx.get<date>('event.metadata.lastTriedAt')
  // Do Something
  return new CloudEventResponse.ok()
}

//Add valdiation middleware 
const mid = validationMiddleware({
  req: {
    schema: V1Schema
  }
})
    
const messageHandler = smerf.cloudEvents().use(mid).handler(contractCreatedHandler)
const retryHandler = messageHandler
//const dlqHandler = some other handler with the same 

export default {
    handler: 
    smerf.use(mid).handler(contractCreatedHandler)

// Retry Logic would be based on the status code returned. The backoff timing would be customized in the smerf.config.ts file, same with dlq settings.
    
    // in smerf.config.ts
    ```{
       events: {
         'contracts-created': {
           backoff: [5, 10, 20, 30]
           dlq: true,
           schedule: '*/5 * * * *'
          }
        }
    ```}

contracts.ts

/////////////////////////////////////////WIP//////////////////////////////////////////////////////////

//filepath: src/handlers/topics/contracts.ts

// For Handling an entire topic
//Filename /topics/contract.ts

function batchHandler(cloudEvents, context) {
  foreach(cloudEvent in cloudEvents) {
  switch(cloudEvent.type) {
    case "ContractsCreated":
      //Do Something 
    case "ContractsUpdated":
      //Do Something
      
      context.retrySyncEmitter.emit(cloudEvent) // I think providing these emmitters in context would be nice. 
  
  }
  }
  return BatchResponse.ok() // or BatchResponse.failed()
},)
  
  export const createRetryHandler<ContractsCreated | ContractsUpdated>((cloudEvent, context) => {
  switch(cloudEvent.type) {
    case "ContractsCreated":
      //Do Something 
    case "ContractsUpdated":
      //Do Something
  
  }
  return CloudEventResponse.ok() // Or CloudEventResponse.failed()
  })
  
    export const createDLQHandler<ContractsCreated | ContractsUpdated>((cloudEvent, context) => {
  switch(cloudEvent.type) {
    case "ContractsCreated":
      //Do Something 
    case "ContractsUpdated":
      //Do Something
  
  }
  return CloudEventResponse.ok() 
  })

// Some configuration can happen in the config
    
    // in smerf.config.ts
    ```{
       topics: {
         'contracts': {
           backoff: [5, 10, 20, 30],
           dlq: true,
            // etc.
          }
        }
    ```}
  

I would avoid implementing split batch in kafka

return new TopicResponse(successfulEvents, failedEvents)

I like the folder pathing strategy separating topics from events.

So, the intention behind the TopicResponse.ok(successfulEvents, failedEvents) would be that it would return a successful response to adapter so it can continue processing, and the adapter would determine what to do with failed events. This is just a high level interface, a lot of the complexity would be the adapters responsibility, but we allow devs to specify which events were successful vs failed, write retry and dlq handlers to support those events, but leave the adapter the responsibility for determining the workflow pattern and integration with infrastructure components. In the case of kafka, we could implement whatever pattern you think makes the most sense and covers the most use cases.

code would be something like...

function contractCreatedHandler(
  ev: CloudEvent<"contracts-created", V1Schema>,
  ctx: Context
) {
  const messageBody = ev.data;

  const isRetry = ev.failure?.attemptCount > 0;
  const lastRetry = ev.failure?.timestamp;
  // Do Something
  return new CloudEventResponse.ok();
}

what is the retryEmitter doing? Is it retry sending OR send the event for retry of processing?

context.retrySyncEmitter.emit(cloudEvent)

should cloudEvents be an async iterable OR array is another debate... async iterable would allow for back pressure modeling