Export routing information

[skial]

I've been experimenting with the tink_*, web libraries and aws lambda and have a rough tink_http container implementation that works targeting the node runtime.

To have aws lambda working via http, you have to use aws api gateway which uses, imo, an odd request/response integration "frontend" to your aws lambda function.

This requires you to specify a path segment, which http method it accepts, the expected mime type and more.

As tink_web has the ability to access the majority of the http info already from your routes, would there be interest in having tink_web export the info to json or some other format, like openapi?

[back2dos]

Oh, the container part is exciting news :)

I do indeed have plans for this kind of stuff. For me reverse routing is the most important bit (so you don't have to assemble URLs by hand), but exporting a full routing table should also be possible. I'll see what I can do ;)

[back2dos]

@skial Say, I wonder is there no way to bypass this restriction? How do other frameworks cope with that? Where might I find documentation for this?

[skial]

@skial Say, I wonder is there no way to bypass this restriction?

As far as I've been able to figure out, no.

How do other frameworks cope with that?

I haven't actually looked to see if any other frameworks actually support deploying to aws lambda, I just started hacking. :grin: I'll take a quick look.

I'm thinking of using https://www.terraform.io/ so I don't have to do too much bootstrapping by myself.

Where might I find documentation for this?

I'll post some links etc in another comment as I've got a load of tabs open, as two aws services, imo, are poorly linked together.

[skial]

More articles than documentation.

I actually forget to say that aws IAM also, ideally, needs to be automated. IAM is used to allow gateway & lambda to talk to each other and to aws cloudwatch, which is used for monitoring. But this is going beyond the scope of the original issue.

• aws lambda
  • NodeJS programming model
  • Using Lambda with Gateway (On-demand over HTTPS)
• aws api gateway
  • Getting started w/ Lambda & Gateway
• articles
  • Build a serverless anagram solver on aws
  • Accessing HTTP headers in Lambda & Gateway
  • How to report errors from Lambda => I need to look at how to deal with errors again since the recent tink_http update, inspect the response object I guess.
  • How to return html from Lambda
  • Build a serverless URL shortner
  • Creating a request object for your lambda function from gateway
• forks
  • tink_http aws NodeContainer.hx

I've also included the request mapping I'm currently using.

Gateway Request Integration mapping

{
  "ip" : "$context.identity.sourceIp",
  "resourcePath" : "$context.resourcePath",
  "body" : $input.body,
  "headers": {
    #foreach($header in $input.params().header.keySet())
    "$header": "$util.escapeJavaScript($input.params().header.get($header))" #if($foreach.hasNext),#end

    #end
  },
  "method": "$context.httpMethod",
  "params": {
    #foreach($param in $input.params().path.keySet())
    "$param": "$util.escapeJavaScript($input.params().path.get($param))" #if($foreach.hasNext),#end

    #end
  },
  "query": {
    #foreach($queryParam in $input.params().querystring.keySet())
    "$queryParam": "$util.escapeJavaScript($input.params().querystring.get($queryParam))" #if($foreach.hasNext),#end

    #end
  }  
}

[skial]

Looks like the serverless framework would solve alot, at first look.

[back2dos]

Phew, that's quite some reading to do :)

[kevinresol]

Maybe AWS has changed the API or something. Now it doesn't need a manual mapping template. Simply "use lambda proxy integration"

Quick links:

• Data format doc
• Container implementation

A few known limitations:

• duplicated query params not supported (because it is represented as key-value object)
• duplicated headers not supported (because it is represented as key-value object)
• streaming not supported (because the whole request/response is encoded into a json)
• need to distinguish the response type between plain-text or binary. Because if it is binary it has to be base64 encoded

[skial]

Thats great @kevinresol :smile: