So yes - I built something stupid again. Recently I discovered the awesomeness that is @Big Data Batman. This is a twitter account that simply copies tweets with "Big Data" in them and replaces it with "Batman." It works as well as you may think - either lame or incredibly funny. (At least to me.) Here are a few choice samples.

That last one is a bit subtle.

I thought it would be fun to build something similar for serverless, and obviously, I had to name it Serverless Superman.

Alright, so how did I build this? My basic idea was this:

On a schedule, look for tweets about serverless from the last X minutes, X being the same as my schedule, find a random one, replace the word "serverless" with Superman, and tweet.

As with everything complex I've done with OpenWhisk, my solution involved a sequence of actions. I began by plotting out my actions in text form.

seq1:

action1: set search: serverless since: today

twitter/getTweets

action3: remove RTs or older than X minutes

action4: pick one and use Superman

twitter/sendTweet

I named it "seq1" as I thought maybe I'd end up with multiple sequences, but one was enough. Let's break this down action by action.

Action 1: setupsearch #

The purpose of this action was to serve as the input provider for the next one that will perform the Twitter search. Here is the code:

/*
I basically set up the args to pass to twitter/search
*/
function main(args) {

	let now = new Date();
	let datestr = now.getFullYear() + '-'+(now.getMonth()+1)+'-'+now.getDate();

	let result = {
		term:"serverless",
		since:datestr
	}

	return result;

}

The only real complex part here is since. The Twitter API lets you filter by date. Unfortunately you can't filter by time. That's going to be a problem later on but I'll address that in the third action. Notice I'm using two keys related to my Twitter account. I got these by logging into the developer portal with my Serverless Superman account.

Action 2: twitter/getTweets #

This is an action I built as part of a public package. You can find the complete source code here: https://github.com/cfjedimaster/twitter-openwhisk. I'm not going to share the code since I blogged on it a while back, but I did have to update the package action to support the "since" argument.

Action 3: filterresults #

The purpose of this action is multifold. It's main role is to filter the Tweets, but I also flatten the data quite a bit as well. I filter out retweets, replies, and items older than X minutes, where X is 10.

Finally I return an array of results where I just carry over the id, text, created_at, and hashtags value of the tweets.

/*
given an array of tweets, remove ones older than X minutes, and RTs, and replies
also, we remove a shit-ton of stuff from each tweet
*/

//if a tweet is older than this in minutes, kill it
const TOO_OLD = 10;

// http://stackoverflow.com/a/7709819/52160
function diffInMinutes(d1,d2) {
	var diffMs = (d1 - d2);
	var diffDays = Math.floor(diffMs / 86400000); // days
	var diffHrs = Math.floor((diffMs % 86400000) / 3600000); // hours
	var diffMins = Math.round(((diffMs % 86400000) % 3600000) / 60000); // minutes
	return diffMins;
}

function main(args) {

	let now = new Date();

	let result = args.tweets.filter( (tweet) => {
		//no replies
		if(tweet.in_reply_to_status_id) return false;
		//no RTs
		if(tweet.retweeted_status) return false;

		// http://stackoverflow.com/a/2766516/52160
		let date = new Date(
    tweet.created_at.replace(/^\w+ (\w+) (\d+) ([\d:]+) \+0000 (\d+)$/,
        "$1 $2 $4 $3 UTC"));

		let age = diffInMinutes(now, date);
		if(age > TOO_OLD) return false;

		return true;
	});

	//now map it
	result = result.map( (tweet) => {
		return {
			id:tweet.id,
			text:tweet.text,
			created_at:tweet.created_at,
			hashtags:tweet.entities.hashtags
		};
	});

	return { tweets:result };
}

One thing that kind of bugs me is the TOO_OLD value. Right now I have to ensure it matches my cron job (more on that later) and if I forget then I'll have a issue with my data. It's not that too bad of an issue and so I just got over it.

Action 4: makeresult #

Yeah, that's a pretty dumb name. The idea for this action is - given an input of tweets, pick one by random and replace the word "serverless". Here is where things get a bit wonky. Sometimes I found tweets where "serverless" wasn't in the text. When I looked online, I saw them in the hashtags. Ok, so I updated my code in action 3 to include the hashtags. This is where I then discovered that the Twitter API seemed to not include all the hashtags I could see in the Tweet.

So... I shrugged my shoulders and got over it. As you can see, I wrote a note that it would be good to not give up and select another Tweet, but I thought maybe Serverless Superman could just STFU for a bit and wait.

/*
so i have an array of tweets. i pick one by random and replace serverless w/ superman
*/

function main(args) {

	return new Promise( (resolve, reject) => {

		if(args.tweets.length === 0) return reject("No tweets.");

		let chosen = args.tweets[ Math.floor(Math.random() * (args.tweets.length))];
		console.log('i chose '+JSON.stringify(chosen));

		if(chosen.text.toLowerCase().indexOf('serverless') === -1) return reject("No serverless mention");

		//todo - maybe loop to find another one if first item found didn't have the keyword

		let newText = chosen.text.replace(/serverless/ig, "Superman");
		console.log('new text is: '+newText);

		/*
		ok, so the next step it to tweet, for that, i need to pass:
		status
		*/
		resolve({
			status:newText
		});

	});

}

Note that if I don't have any Tweets or if I can't find the word "serverless", I reject the sequence. This is not the right thing to do. OpenWhisk does support conditional sequences but it's a bit... complex right now. There is an open issue to make it a bit simpler and when that happens, I'll consider updating the post then, but for now I dealt with it. It does mean, however, that my action is going to report errors when an error really didn't occur.

Action 5: twitter/sendTweet #

Finally - I send my Tweet. This is a new action in my Twitter package so I'll share the code here.


const Twitter = require('twitter');

/*
I send a tweet. i need:

args.status (the text)

and that's all I'm supported for now! Note, unlike getTweets
which can get by with less access, for this you need user auth
as documented here: https://www.npmjs.com/package/twitter
*/

function main(args) {

	return new Promise( (resolve, reject) => {

		let client = new Twitter({
			consumer_key:args.consumer_key,
			consumer_secret:args.consumer_secret,
			access_token_key:args.access_token_key,
			access_token_secret:args.access_token_secret
		});

		client.post('statuses/update', {status:args.status}, function(err, tweet, response) {
			if(err) return reject(err);
			resolve({tweet:tweet});
		});

	});

}

exports.main = main;

Nothing real complex here, but note I'm only allowing for text based Tweets. The API supports a lot more than that.

Finally, you may have noticed that my sendTweet action requires multiple authentication tokens. How did I pass them? I didn't. I simply used the OpenWhisk "bind" feature and made a copy of my package with all my tokens attached to it. Bam - done.

Putting it Together #

The final bits included actually setting up the scheduled task. The first part required making a CRON based Trigger. Here's the command I used for that:

wsk trigger create serverless_superman_trigger --feed /whisk.system/alarms/alarm -p cron "*/10 * * * * *"

I used https://crontab.guru to help me build the cron value.

Then I made a rule that associated the trigger with the sequence I created of the actions above.

wsk rule create serverless_superman_rule serverless_superman_trigger serverless_superman/dotweet

And honestly, that was it. I opened up the OpenWhisk dashboard on Bluemix and kept watch of it and it just plain worked. (After some help from Carlos in Slack!)

Here is an example:

And my current favorite:

You can find the code this demo (excluding the Twitter actions which have their own repo) here: https://github.com/cfjedimaster/Serverless-Examples/tree/master/serverless_superman