I have been learning plenty of new stuff lately, so I thought I’d write about the things in the works, the concepts grasped, and the ideas acquired in the process.
Moe
A few weeks ago I started contributing to Moe, an ultra-modern Perl interpreter written in Scala.
This has been extremely fun, since it is -Ofun. I fleshed out the AST quite a bit and helped make the class-instance model more functional.
I did not need to know the ins and outs of perl5’s interpreter. As far as I know, Moe is nothing like perl5’s interpreter. Moe is built using ASTs from the ground up. It may not be the most efficient route to design a programming language, but it is the quickest route to a working implementation and is incredibly easy to grasp once you get your feet wet.
By no means am I an expert in Scala either. The Scala School was more than enough to get me started. I took my time, and when I wanted to explore deeper, I jumped into The Tour of Scala and revisited old code to fix things up.
Neptune
The past few weeks I accidentally deflated my “Moe”mentum with a new project: Neptune.
The point of this project isn’t to benefit other people, but to benefit my knowledge of Scala and other tools in general. I have no professional experience in distributed systems beyond deploying code to Linode and Heroku. I want to get my feet wet in having multiple services communicate in a way that scales, and at the same time is not massively over-engineered. This project also houses my first-ever python script!
At this point it publically faces to a TCP port that users connect to (via telnet or some-such), though it does not communicate with the internal universe at all. The universe is managed and scheduled using an Akka scheduler; and will a) receive commands, and b) send updates using Thrift. I haven’t yet decided on a persistent store for the universe data. For now, I am currently importing everything from a flat JSON script and loading everything into Redis. And honestly, I haven’t finally decided on Thrift either. I have heard just as many good things about Google’s protobuf.
In two different execution contexts (one for the “nature” scheduler, and one for redis IO work), I implement the dynamic of the universe through fluid, event-based timing. When I glue the services together with Thrift, I will probably create a third execution context for its own blocking needs.
I am caching the universe data like there is no tomorrow. Nearly everything you can think of is in a set. Mobiles (NPCs) are in a set per location, and reciprocally, mobiles own a “location” property. My aim is to keep property lookup work to a minimum to mitigate unnecessary blocking. Besides, frequent updating is Redis’ specialty! I also hope to abstract away all the extra update work as much as possible to avoid bugs.
Details pending
This is just a preface to all the different topics I hope to share in upcoming posts. Stay tuned!