I’ve been doing a lot of dabbling with Javascript lately; primarily focusing on working with the canvas element. It’s some pretty sweet stuff and really reminds me of my early pixel manipulation tasks in C++ and REALbasic in the late 90s.

I had written a version of Conway’s Game of Life in C++ on my mac at the time (7600/132mhz) and was surprised at how fast it ran. I figured I could build something similar in JS using the canvas element and direct pixel manipulation using getImageData() and putImageData().

My initial tests ran ok in Chrome, but ran pretty terrible in Safari. I’m pretty much limited to a 200px x 200px world plane and I’m trying to come up with ways of speeding it up. I’ve done a bit of multithreaded programming in the past, but never really did much with it since I don’t typically write things that would benefit from it, but this seemed like a perfect situation, assuming that JS can simulate threads with the setTimeout() function and that these “threads” would run across my processors.

My method of multithreading was to create a function that iterates over a section of the world plane. Since the image data is a one-dimensional array of pixels, this is done by simply creating a function that processes a cycle on a specific slice of the pixel data. When each chunk finishes, it updates an array (thread_status) by appending a ‘true’ to the end. If thread_status.length is equal to the number of threads, then it draws the current graphic to the canvas using putImageData(), resets thread_status and starts the render() function over again.

I was quite surprised when these additions offered absolutely zero benefit. I suppose I’ll have to figure out better ways of iterating over the world plane for each cycle of life. I’m sure there are better ways of doing it than I’m doing and if I had access to a vector unit or some matrix math libraries, it might seriously speed things up, too.

single-threaded code can be found at:

• https://github.com/spikegrobstein/js_conway

multithreaded branch:

• https://github.com/spikegrobstein/js_conway/tree/multithread

demo:

• http://spike.cx/tests/js_conway/?threads=8

in the demo, you can change the number of threads by changing the threads GET variable in the URL.

I just made my first release of some code I was working on for the past couple days. It’s a Javascript library whose purpose is to simplify delegation of web requests by leveraging anchor URLs (the data in the trailing # of the URL).

The project has been released under the MIT license.

The advantage of this library is that you don’t need to use CSS selectors to locate your links and attach click actions. Rather, the library will handle any link whose href attribute that starts with a #.

There are 2 classes included in the library:

• RouterCore — base class that handles routing and automatic handling of links
• MappingRouter — a simple class that allows for semi-complex delegation of request handling.

See the project github page for additional information and documentation:

https://github.com/spikegrobstein/js_router

The purpose of this post is not simply to bash PHP, but rather to express what I see as severe shortcomings in the PHP language. Ultimately, I’d like to make an argument for a fork of PHP as a way to resolve these shortcomings and it a better language for everyone. Because it would be impossible to make my changes without a completely new project and the final result wouldn’t look much like the PHP that people know and love/hate today, it would need to be a “new” language.

This should be a multi-part series (at least 3 parts) which I will write whenever I have some time and motivation to crank out these ideas.

Background

Before I begin, I’d like to share my background a little bit. I’ve been writing code off and on since I took my first programming class in 6th grade. This class taught BASIC on PCjr machines and, at the time, I had fun doing it but didn’t fully grasp what I learned until 2 or 3 years later when I discovered QBasic on a machine at my dad’s job and I had an idea for how I could write a program to animate a ball bouncing around the screen. From that point on, I was writing a significant amount of code and tearing through programming languages first with REALBasic (then, called CrossBasic or XBasic), to C, then C++, then PERL, ObjC, PHP, BASH, Python, Ruby…… and on and on.

Much in the same regard that I’m pretty much a programming language slut, I’m also an operating system slut. I love seeing how OS’s do things differently and all this diversity has given me, what I consider to be, a pretty decent perspective on good, better and best ways of doing things.

Jumping between these languages, I’ve seen all the various -isms that each language attempts to steer you into. I’ve seen languages go out of their way to make things easier for the developer or easier for the compiler. This leads me to my first section.

You’re doing it wrong

The first, very wrong thing that jumped out at me early in my PHP learning process was that, off the bat, it encourages developers to mix their business logic with their view code. PHP was designed, from day one, to be embedded in HTML.

I understand that this was done for the express reason of lowering the barrier to entry and enabling users to build dynamic web content with minimal effort. This is a great idea, and I’m not faulting the original PHP developers for it, but I feel that it is a bad move to make that part of the core language. If a developer is creating some code that they want to share between projects, whether that code will be used for web content or a commandline script or whatever, they must wrap their code in start and end tags (<?php //code… ?>).

Since the majority of the code that I write is stuff that runs on the commandline, I find it a little annoying to always need to wrap my code in these tags. When creating a web application, especially when using MVC design concepts, the only code that should use embedding is the view code. The rest of the app should both be non-web-accessible and also be nothing but code.

Of course, when running things on the commandline or using CGI (with nginx, for instance), you can configure it to use the ‘-r’ switch to disable the requirement of said tags, but when using mod_php in apache (as is most often the case), this is not an easy task. Also, if you’re mixing files that contain start/end tags with files that do not, it can get hairy, especially if you forget to use the switch.

PHP should have a built-in method for automatically detecting that a file should be treated as raw sourcecode vs embedded code. The most obvious solution would be the use of different file extensions. Personally, I’d choose ‘phpe’ for files with start/end tags and just ‘php’ for raw php source. I don’t believe that affects the difficulty for beginner users in any way and also hints that there is a time and a place for each type of file.

This concludes Part 1. If you’ve got any corrections, you can post in the comments or contact me. Contact info is available on my website (http://spike.grobste.in).