This'll be a small post, so, yeah.
I first started event-based loops with Twisted (http://twistedmatrix.com/trac/), and I had absolutely NO idea what I was doing.
I was doing all kinds of nonsensical things (I still have my code), like trying to read user input from stdin with raw_input() in a callback function. I slowly got better, and gradually learned more about what types of functions blocked, etc. etc. But, that took a while. And, if I needed to write production-ready code straight off the bat, I wouldn't have been able to do it (i.e. I would have written the code, it just would really, really suck).
Then, I tried out eventmachine and libeio, and they were all pretty similar, and I still had to be pretty careful about what type of code I was writing (one has to think less and less about this type of stuff the more you do it), and it wasn't really flowing.
Then, I found out about node.js.
Unlike a lot of people, I wasn't drooling over it or anything, mostly because it uses Javascript, and I didn't really think that moving Javascript over to the server side would help any. This conclusion was mostly rooted from the fact that I always troubled myself with writing cross-browser JS, and that's a painful task. But, with node, that disappears. There's only V8, to hell with all the other parsers and compilers that developers have come up with.
Javascript, if written for only one interpreter/compiler, is pretty usuable, and actually pretty fun. The prototype object system stops getting in the way, because you're provided with Object.create() which IE ... let's not talk about IE :)
The point with node is, its really, really hard to write completely blocking code, because javascript doesn't have anything like File IO that blocks with which you can shoot yourself in the foot!
The things I could write in a short amount of time literally skyrocketed, because I didn't have to worry so much about putting the wrong call in the wrong place, and could just concentrate on getting the application logic down.
There are, however, a few cinches that you have to deal with. Firstly, there's no tail call optimization in v8, even though it has been requested (http://code.google.com/p/v8/issues/detail?id=457). This is mostly because of some stupid things done in the spec, but, Node would be a lot better for me if this request was fulfilled. Secondly, if you are a server-side developer, getting used to JS will be pretty ... different, that's not really bad, but, it might slow you down.
So, all in all, I highly recommend you check out Node.js, and try writing something simple (like a chatroom or something).
Feedback welcome.
I first started event-based loops with Twisted (http://twistedmatrix.com/trac/), and I had absolutely NO idea what I was doing.
I was doing all kinds of nonsensical things (I still have my code), like trying to read user input from stdin with raw_input() in a callback function. I slowly got better, and gradually learned more about what types of functions blocked, etc. etc. But, that took a while. And, if I needed to write production-ready code straight off the bat, I wouldn't have been able to do it (i.e. I would have written the code, it just would really, really suck).
Then, I tried out eventmachine and libeio, and they were all pretty similar, and I still had to be pretty careful about what type of code I was writing (one has to think less and less about this type of stuff the more you do it), and it wasn't really flowing.
Then, I found out about node.js.
Unlike a lot of people, I wasn't drooling over it or anything, mostly because it uses Javascript, and I didn't really think that moving Javascript over to the server side would help any. This conclusion was mostly rooted from the fact that I always troubled myself with writing cross-browser JS, and that's a painful task. But, with node, that disappears. There's only V8, to hell with all the other parsers and compilers that developers have come up with.
Javascript, if written for only one interpreter/compiler, is pretty usuable, and actually pretty fun. The prototype object system stops getting in the way, because you're provided with Object.create() which IE ... let's not talk about IE :)
The point with node is, its really, really hard to write completely blocking code, because javascript doesn't have anything like File IO that blocks with which you can shoot yourself in the foot!
The things I could write in a short amount of time literally skyrocketed, because I didn't have to worry so much about putting the wrong call in the wrong place, and could just concentrate on getting the application logic down.
There are, however, a few cinches that you have to deal with. Firstly, there's no tail call optimization in v8, even though it has been requested (http://code.google.com/p/v8/issues/detail?id=457). This is mostly because of some stupid things done in the spec, but, Node would be a lot better for me if this request was fulfilled. Secondly, if you are a server-side developer, getting used to JS will be pretty ... different, that's not really bad, but, it might slow you down.
So, all in all, I highly recommend you check out Node.js, and try writing something simple (like a chatroom or something).
Feedback welcome.
No comments:
Post a Comment