Startup Clix: Cleanup & Fighting with Pusher Application State

Last night and today I fixed some of the packaging issues with the help of some build util scripts.

node_modules for every Lambda function

I added a package.json for every Lambda function, so I would not have to copy the back-ends base node_modules folder every time I changed something.

Then I use this script to run npm i in every Lambda function directory:

const { readdirSync, statSync } = require("fs");
const { join, resolve } = require("path");
const { execSync } = require("child_process");

const functionsDir = resolve(__dirname + "/../functions/");

const dirs = p =>
  readdirSync(p).filter(f => statSync(join(p, f)).isDirectory());

console.log("Running 'npm i' for Lambda functions:");
dirs(functionsDir)
  .map(d => {
    console.log("  - " + d);
    return d;
  })
  .map(d => join(functionsDir, d))
  .forEach(d => execSync(`cd ${d} && npm i`));

Shared code between the Lambda functions

I added a shared folder that gets copied to every Lambda function. Not very pretty but it works right now.

const { readdirSync, statSync } = require("fs");
const { join, resolve } = require("path");
const { ncp } = require("ncp");

const sharedDir = resolve(__dirname + "/../shared/");
const functionsDir = resolve(__dirname + "/../functions/");

const dirs = p =>
  readdirSync(p).filter(f => statSync(join(p, f)).isDirectory());

console.log("Copy shared code to Lambda Functions:");
dirs(functionsDir)
  .map(d => {
    console.log("  - " + d);
    return d;
  })
  .map(d => join(functionsDir, d, "shared"))
  .forEach(d => ncp(sharedDir, d));

Kappa for async Lambda functions

I only share one helper, called it kappa, at the moment, which solves one of the problems I had, the basic Lambda function interaction.

const lambda = require("../index.js");

exports.handler = (event, context, callback) =>
  lambda(event, context)
    .then(r => {
      if (r && r.body) r.body = JSON.stringify(r.body);
      callback(null, r);
    })
    .catch(callback);

This function becomes the actual Lambda handler and lets you write your Lambda function async. It also stringifies the body of a response if it exists.

exports.handler = (event, context, callback) => {
  callback(null, JSON.stringify({result: event.body + "!"}));
}

Now becomes

module.exports = async (event, context) => ({
  result: event.body + "!"
});

Pusher Application State

Then, after all the maintainence was done I tried to get some game/channel joining logic on the server going.

The idea was, when someone wants to join a channel, they have to request a channel-ID via GET /getgamechannel.

The server will calculate one based on the open channels and how many players are inside of them

This /getgamechannel endpoint, together with the /pusherauth endpoint, should later keep players from joining full or already running games.

I read that I could query the application state of my Pusher app on the server side, but somehow this doesn't work consistently.

pusher.get({
  path: "/channels",
  params: {
    info: ["user_count"]
  }
});

My clients get presence updates from Pusher and I see every client on every client, but when I run this query on the back-end, often get an empty list of channels or something.

Either the application state is eventually consistent (but I don't know when?) or I have some error in my code, which wouldn't be too surprising :D

Conclusion

Anway, a fun project and I already learned much about Pusher and AWS :)