Mathew Davies

Upgrading Kohana 3.0 to 3.1

2011-01-16T00:00:00+00:00

Last updated on Sunday the 16th of January 2011

Common things to look out for when upgrading from Kohana 3.0 to 3.1

Auth

Hashing is now done with hash_hmac. This will break all your current logins. To upgrade your user passwords, use the old function to check their passwords. If they match, generate a new hash.
The ORM driver has been moved into the ORM module.

Controller

The constructor takes 2 arguments (Request $request, Response $response) over just (Request $request)

Core

Kohana::exception_handler($e) has been moved to Kohana_Exception::handler($e)

ORM

save() and check() now have a Validation argument, make sure you update any child classes to ensure you don’t break strict compatibility.
Rules, filters and labels are now defined in public functions (rules, filters and labels) which return an array.
A DB group is now specified with $_db_group over $_db

Request

$this->request->controller, action, directory are now methods: $this->request->controller(), action(), directory()
A response is set with: $this->response->body('content') over $this->request->response = 'content';
A new set of exceptions has been created to deal with http errors. They range from Http_Exception_400 to Http_Exception_505

Validation

Validate has been split into two classes: Valid and Validation. The former houses all the rules and the latter has the logic for checking, messages, etc …
filter() no longer exists. Either filter manually or use Arr::map if you need to filter on all values.

<?php
// Filtering post values using a lambda function.
Arr::map(function($value)
{
	return trim($value);
},
$_POST);

callback() has gone. Instead, rule()'s 2nd argument can take any valid PHP callback. The 3rd argument is an array containing the arguments to pass to the callback.

<?php $validation->rule('email', array($this, 'email_available'), array(':validation', ':field'));

Callback syntax has changed from array('function' => $params) to array('function', $params).

Changelog

Sun 16 Jan 2011 - Initial version

Nginx HTTP Server Book Review

2010-08-28T00:00:00+01:00

Introduction

A little bit about myself before I jump into the review. My name is Mathew Davies, I’m a PHP developer from a small village in England called Barton-Le-Clay and I’m also roller coaster enthusiast. I run a fairly big website dedicated to my hobby (nolimits-exchange) so I’m always on the lookout to improve performance while keeping the costs down. That’s how I came to understand Nginx with its low memory and CPU requirements, perfect for my small VPS.

I was kindly asked to review the book and I couldn’t miss out reviewing the first Nginx book. Special thanks go to PacktPub for giving me the opportunity to do so.

The book is about Nginx (aka Engine x) which is gaining a lot of popularity throughout the server world. This is down to its asynchronous architecture and low memory and CPU requirements; Perfect for squeezing more performance out of your current hardware. The book aims to convert over Apache users and anyone with an interest in optimizing their infrastructure.

The Review

I was confused about the first chapter Preparing your Work Environment. It explains such basics as: How to use a terminal emulator (Putty), the Linux directory layout and file permissions. The blurb gave me the impression that you should be able to administer a Linux server to a basic level, so to find this information was a bit odd. I don’t expect the first chapter to be useful to the target audience. Putting this into consideration, I would remove the first chapter and make it up by going into Nginx module development; A subject which could benefit.

I found one mistake in the first chapter. On page 44, the octal permissions are wrong, here is what is in the book:

1 = Read
2 = Write
4 = Execute

and what the values should be:

1 = Execute
2 = Write
4 = Read

The next chapter Downloading and Installing Nginx is basic and teaches the reader how to compile Nginx along with its dependencies. It would have been educating to show the reader how to install the packages with apt-get or aptitude. I suggest this because the first chapter assumes the reader has little to zero knowledge of Linux.

Chapters three till five are about the Nginx configuration directives and modules. A lot of the content here can be found online at websites like the Nginx Wiki, but it goes into more detail with real life examples. They serve as a great reference and will probably be the most used after the initial learning is over.

We then get into the realm of setting up a PHP-FPM/Python combo. Just a note about the PHP-FPM guide: FPM is now part of PHP from version 5.3.3 onwards, you no longer need to apply the FPM patch. It’s recommenced you avoid the patch as it’s no longer maintained. I did learn something new from this chapter: Setting up PHP as an upstream server in Nginx. Normally I’d just use fastcgi_pass directly to the server, but the upstream block makes it easier to scale out your application servers.

If you don’t want to move to Nginx right away, it might be a good idea to keep Apache as the backend and use Nginx as the front-end. This means you can free Apache of processing your static files (css and js) which gives it a bit more memory to play with; Your static content will get served faster too! Apache and Nginx Together goes into how you would set this up and describes common problems you might encounter (forwarding client IP addresses, SSL issues, etc …)

So, you’ve decided to move to Nginx after hearing how great it is, now you’re thinking “Where do I start?”. From Apache to Nginx (Chapter 8) will be able to help you there. Personal experience of being on the Nginx mailing list tells me this chapter is going to be helpful to those who are strugging to port over their rewrite rules and server configurations. Direct comparisons have been between both software configuration styles so it’s easy to see how Apache directives translate. Rewrite rules for popular web applications such as Wordpress have been included for convenience and even more can be found online at the Nginx Wiki.

Conclusion

The book was an enjoyable read and I took away some methods which will help scale my website in the future. I’d certainly recommend the book to anyone who doesn’t use Nginx (just to see what they’re missing out on) or a basic Nginx user who wants to further their knowledge on the subject. It covers most of what you’ll need and makes improvements where the Nginx Wiki is lacking.

One suggestion: Page 105 Upgrading Nginx Gracefully should include how to switch back to your old Nginx process if the new one failed to work. Here is how that would be done:

kill -HUP 123 - this will restart the workers under the old master process
kill -QUIT 321 - shuts down the new master process workers
kill -TERM 321 - shuts down the new master process

Source: Software Projects Inc - Upgrade nginx binary with no down-time.

I also spotted a minor syntax error (missing “;”) and a bit of code that was copied and pasted, but not changed. That said, it’s a great book and despite its mistakes I recommend purchasing it.

A free chapter is available from PacktPub, check it out: Basic Nginx Configuration

More Nginx Resources

The Nginx Wiki: Great for documentation and tips. These two pages are invaluable; Read and understand them: Nginx Pitfalls and If Is Evil.
The Nginx Mailing list

Biography

I specialize in PHP development and work with the Kohana framework. I have several open source modules available and they can be found on my github page: thepixeldeveloper. I’m available for hire and can be contacted using the email address in the website footer.

What I've learned in the past year while running a website.

2010-08-16T00:00:00+01:00

It’s been one year since my little hobby project Nolimits-Exchange went online. I’ve learned a lot over that year and I’d like to share some of what I’ve learned with you.

Advertising

Adsense was very poorly suited to the content of my website. I had a lot of Public Service Ads which although nice, don’t make any money. It was not worth annoying my users for the sake of 30 dollars a year.

We’re currently experimenting with another company. I’d love to say who, but that would be a breach of contract.

Backups

I’ve always backed up data with Amazon S3 via Duplicity; It’s cheap enough and works well, but I came across something better. I switched to an alternative called tarsnap.

The reasons:

Write only encryption keys: I can be sure my backups wont be destroyed when the server is compromised.
Pay as you go: This avoids any nasty surprises that could occur in my Amazon invoice.
It has a similarity to tar, hence the name. It makes writing backup scripts easier.

It’s slightly more expensive at $0.30/GB-month versus Amazon’s $0.15/GB-month.

Community

I’ve had to ban a few members and one thing to look out for is suicide threats. I made the the assumption it was an empty threat and didn’t do anything about it. Don’t make that mistake and report the incident to the local police department.

Source: How do you handle suicide threats.
You should handle it just like any other threat to someone’s life. A person threatening suicide, or to kill someone else, is still threatening to kill someone. The only sensible thing to do is try and report it.

Beyond that it’s been a lot of fun meeting up with the community in real life.

I had a lot of help from Community Spark when starting. It features a lot of advice about choosing your staff, kick-starting your community, banning members, etc …

Development

The amount of improvement I’ve seen in my programming is a reminder that I’m not progressing as slowly as I’d thought. Take this example:

Original code: The XML is a string ($this->output) that was concatenated at various points.

<?php defined('SYSPATH') OR die('No direct access allowed.');

class Sitemap_Driver
{
	// XML header
	const header = '<?xml version="1.0" encoding="UTF-8"?>';

	// XML output
	protected $output = '';

	// Change frequency
	protected $changefreq = array('always', 'hourly', 'daily', 'weekly', 'monthly', 'yearly', 'never');

	public function __construct()
	{
		$this->output = self::header.PHP_EOL;
	}
}

1 year later: I’m using the DOMDocument class which offers formatting, syntax checking and an easier API to work with.

<?php defined('SYSPATH') or die('No direct script access.');

class Kohana_Sitemap
{
	/**
	 * @var DOMDocument
	 */
	protected $_xml;

	/**
	 * @var DOMElement
	 */
	protected $_root;

	/**
	 * Setup the XML document
	 */
	public function __construct()
	{
		// XML document
		$this->_xml = new DOMDocument('1.0', Kohana::$charset);

		// Attributes
		$this->_xml->formatOutput = TRUE;

		// Root element
		$this->_root = $this->_xml->createElement('urlset');

		// Append to XML document
		$this->_xml->appendChild($this->_root);
	}

So take a look at your old code and be proud that you’re improving.

Deployment

When you have a busy website, you realise how important it is to keep the downtime to a minimum when pushing an update. The horror of FTP’ing a tarball to a server, then unpacking it only to find it didn’t work is a crap feeling (at best). This is why I learnt about version control, proper website deployment and database migrations.

I now use:

Git for code version control,
Capistrano for deploying the websites &
LiquiBase for database migrations.

Downtime is no longer than 10 seconds and if I ever need to, can revert to an earlier state. I know very few developers that know all three, how are you surviving without migrations?

If you’d like to learn more about Capistrano, then my article PHP Deployment might be helpful to you.

Nginx Server Book Preview

2010-08-15T00:00:00+01:00

I will be reviewing Nginx HTTP Server by Clement Nedelcu very shortly. So to you get excited, here is a preview of what the book covers:

Get started with Nginx to serve websites faster and safer
Learn to configure your servers and virtual hosts efficiently
Set up Nginx to work with PHP and other applications via FastCGI
Explore possible interactions between Nginx and Apache to get the best of both worlds
A step-by-step guide to switching from Apache to Nginx
Complete configuration directive and module reference

Source: PacktPub

As a Nginx fanatic, I am glad to see it get some face time.

If you can’t wait for the review and would like to purchase it, then visit the PacktPub store by clicking on the book image.

Nginx Logwatch Configuration

2010-06-22T00:00:00+01:00

My installation of logwatch did not detect I was running Nginx and thus did not create the configuration files for me. Here is how I fixed that:

This will only work if you use the combined access log format. Create the following file:

/usr/share/logwatch/dist.conf/logfiles/http.conf

and insert the data below:

LogFile = /usr/local/nginx/logs/access.log
Archive = /home/website/logs/old/access.*.log.*.gz

# Expand the repeats (actually just removes them now)
*ExpandRepeats

# Keep only the lines in the proper date range...
*ApplyhttpDate

Hopefully that helped someone out.

Uninstalling Lubuntu

2010-06-19T00:00:00+01:00

If you tried lubuntu-desktop and decided you didn’t like it, you’d probably be right in thinking a quick removal of the lubuntu-desktop package would do the job. Well it didn’t (for me). Here’s how I fully removed it:

Open a terminal and run this command:

sudo aptitude purge abiword aqualung cheese epdfview galculator gnome-mplayer
gnumeric gpicview leafpad lubuntu-artwork lubuntu-default-settings lubuntu-desktop 
lubuntu-plymouth-theme lxappearance lxdm lxinput lxlauncher lxrandr lxsession-edit
lxshortcut lxterminal mtpaint ntp obconf osmo parcellite pidgin powernowd 
pyneighborhood simple-scan sylpheed transmission wvdial xarchiver xfburn 
xfce4-taskmanager xserver-xorg-input-evtouch

In order to fully remove it.

Free Admin Template

2010-03-13T00:00:00+00:00

A professional admin template ready to use in your personal and commerical projects free of charge.

Dashboard

Inspired by the BBC homepage. The dashboard features moveable blocks allowing the user to customise the website to their liking.

Provides a quick overview of system information.
Links in the header can be used to dynamically load more data or jump to a section.

View the dashboard

Forms

Clean, smooth and user friendly.

View the form

Table

View the table

Donations

If this template helped to speed up your day to day work then consider a donation. My paypal email address is in the website footer.

Modifications

Robert Jan Dreu has made a jQuery port. Download it.

Download

The template source code is open-source and can downloaded at github.

Smush.it PHP Class

2010-01-01T00:00:00+00:00

This small class makes use of the Smush.it service provided by Yahoo to compress images.

So far only web accessible images can be compressed, but I’m working on a solution so local images can be compressed too.

<?php
/**
 * Description: Compresses images using Smush.it
 * 
 * @license MIT
 * @author  Mathew Davies <thepixeldeveloper@googlemail.com>
 */
class smushit 
{
    const user_agent = 'Smush.it PHP Class (+http://mathew-davies.co.uk)';

    private $curl = NULL;

    /**
     * Smush.it request URL
     */
    const url = 'http://www.smushit.com/ysmush.it/ws.php';

    /**
     * Make sure any prerequisite are installed.
     */
    public function __construct() {
        if ( ! extension_loaded( 'json' ) ) {
            throw new RuntimeException('The json extension was not found');
        }
        if ( ! extension_loaded( 'curl' ) ) {
            throw new RuntimeException('The cURL extension was not found.');
        }

        // cURL handler
        $this->curl = curl_init();

        // Return HTTP response
        curl_setopt( $this->curl, CURLOPT_RETURNTRANSFER, TRUE );
    }
    
    /**
     * Compress image using smush.it. Image must be available online
     *
     * @param  string url to image.
     * @throws Smush_exception
     * @return object
     * 
     *  src       = source location of input image
     *  src_size  = size of the source image in bytes
     *  dest      = temporary location of the compressed image
     *  dest_size = size of compressed image in bytes
     *  percent   = how much smaller the compressed image is
     */
    public function compress($image) {
        // Set appropriate URL.
        curl_setopt( $this->curl, CURLOPT_URL, self::url.'?'.http_build_query( array('img' => $image ) ) );

        // Set user agent
        curl_setopt( $this->curl, CURLOPT_USERAGENT, self::user_agent );

        // Execute the HTTP request
        $request = curl_exec($this->curl);

        // JSON response
        $result = json_decode($request);

        if ( isset ( $result->error ) ) {
            throw new Smush_exception($result->error, $image);
        }
        
        $result->dest = urldecode($result->dest);
        
        // Return response data
        return $result;
    }
}

/**
 * Custom exception handler
 */
class Smush_exception extends Exception {

    /**
     * Path to image
     * @var string $image
     */
    private $image = '';

    /**
     * Overload the exception construct so we can provide an image name
     * @param string $message
     * @param string $image
     */
    public function  __construct($message, $image) {
        $this->image = $image;
        parent::__construct($message);
    }

    // Return image path.
    final function getImage() {
        return $this->image;
    }
}

On average I save 17% per image. Not bad when you have thousands like I do.

Happy New Year

2009-12-31T00:00:00+00:00

This year wasn’t bad!

I launched a new roller coaster website called Nolimits-Exchange in mid August. It had over 300,000 page views in the space of 5 months so I’m quite impressed with that, I’ll be aiming for 2 million in 2010. I also created a little PHP class which may be of use to some of you.

I’d like to thank you for the comments left on my deployment article. Hopefully the world of PHP is a better place.

PHP Deployment

2009-10-28T00:00:00+00:00

Learn how to deploy PHP applications the proper way using Capistrano.

Introduction

Picture this scenario: You’ve finished your work for the week, it’s time to relax, but you get an urgrent call from your client, “Matt!, There’s a bug on the site. It needs fixing now!”. You quickly scramble to get it fixed as fast as possible, so you login to the server, edit the file from the terminal and the day is saved “hooray”.

There are a few flaws with this method of fixing bugs, lets see what they are:

Since the file was edited directly there’s no change log, so you’ll have to keep a personal record. Wouldn’t it be easier if you could add notes (or messages) when you commit changes to code?
The patch didn’t work and now you have to roll back (sigh). You should have tested locally first, but you were so frantic you forgot (it happens). How would you roll back quickly and easily? If it takes more than a second then it’s too long.

It’s a slow and error prone way of dealing with an emergency. I wonder how we can make this better …

Using version control

I’ll be using bazaar for my deployment, but there’s a huge choice out there. Here are a few you might be interested in: Git, Subversion, CVS or Mercurial

sudo aptitude install bzr

Set up Capistrano

Capistrano is client side software which means we are done touching the server (for now). Prerequisites for deployment are:

capistrano
capistrano-ext (used for its multi-stage plugin)
railsless-deploy (overrides rails actions that don’t apply to PHP)

Writing the recipe.

I’ll assume you have a local code repository setup and populated with content.

From a directory outside of the code repo (unless you want to version your recipe files) run:

capify .

Here is the output of that command:

[add] writing './Capfile'
[add] making directory './config'
[add] writing './config/deploy.rb'
[done] capified!

Since I’ll be using the multi-stage plugin (you don’t have to) I need to modify the “Capfile” that was generated.

load 'deploy' if respond_to?(:namespace) # cap2 differentiator

require 'rubygems'
require 'railsless-deploy'
require 'capistrano/ext/multistage'

set :stages, %w(staging production)
set :default_stage, "staging"

Here we have included the extensions, defined our stages and setup a default stage. All self explanatory. Recipes follow the following structure:

config/deploy/:stage

You can find an example recipe on my gist page.

From here on out it’s normal capistrano deployment procedure so I wont bother repeating what’s on the internet already. I’m happy to answer any questions you may have though.