Author Archive

More Speed: CSS Sprites!

January 21, 2010 at 6:57 pm · Filed under performance, tips ·Tagged , , ,

Hey folks! This ones for those speed freaks who are never satisfied with their page load times. This is for good reason. Page load time could mean the difference between a conversion and a lost visitor.

Why?

There are many, many things you can do to improve site performance. Today I’ll explain CSS sprites and why they can help with performance. The goal of using CSS sprites is to minimize the number of HTTP requests. Each HTTP requests has overhead – the browser has to send a request to your web server and then the web server has to respond with an answer (which could include an image, css file, etc). So minimize the amount of HTTP requests and you speed up your page.

What’s a CSS Sprite?

What the heck is this CSS sprite thing that I’m talking about? Well you know all those little background images that you put on your webpage via CSS? They may be round corners, gradients, icons, etc. Well each of those comes with the price of HTTP requests. Using CSS sprites is essentially a technique of combining some of those background images into one image (called the sprite) to minimize the number of HTTP requests.

How?

Luckily there is a tool that makes generating this CSS sprite super easy: SpriteMe. It was created by Steve Sounders, a performance guru. Once you create your CSS Sprite, use it in your CSS by utilizing background-position. Using background-position, you can adjust which part of the CSS Sprite shows up, allowing you to reuse the CSS Sprite throughout your page. On a recent project I was involved with, we reduced the number of HTTP requests by 13!
example of css sprite

Interested in more performance tips? Feel free to check out the Yahoo Developer Network, they have some great advice on this subject.

Leave a Comment

Minify! And New Site! Ohnoes!

August 4, 2009 at 3:11 am · Filed under PHP, performance

This one will be quick.

Just got done configuring minify onto my local dev environment and then onto the live environment. It rocks! It was a little fussy about the setup, but just pay close attention to the config.php file and you’ll be fine. It works as advertised and definitely helps on load time.

Oh – and I did this on my new website which I “soft-launched” yesterday. It’s a site that allows you to search for and buy shoes via a super easy to use website. The content is intelligently cached from Amazon’s web services. Keep an eye out for a post going into further detail on the implementation.

Update: CakePHP published my article about Got2BShoes in their Bakery.

Leave a Comment

Google Analytics Anomaly: More Visits than Page Views?

June 4, 2009 at 7:16 pm · Filed under analytics

wtfBy definition, page views on any given site should be always be higher than the number of visits. However, myself and some other people have came across an anomaly in Google Analytics where visits are the one that is higher.

I don’t have a solution to offer yet, however I do have an explanation…

This is related to sub profiles and events. If you have a parent profile with events and make sub profiles from it, you will most likely run into this issue (until it is solved by Google). I’ll first explain events and sub profiles. If you’re already familiar, feel free to skip the next two sections.

Events

Events are not associated with page views at all. That was actually a perk of using an event to keep track of non page view type information. In the past, we had to use trackPageview() to reports that didn’t occur ‘naturally’. The drawback of that is that it created fake page views which inflated our page view numbers. Using events, that information is kept separate from page views, allowing us to gather useful data and still trust our page view figures.

Sub Profiles

In Google Analytics, you have a unique ID for every parent profile you have a unique Web Property ID and a parent profile for that web property. You can then create other profiles for that web property and filter out various data. The filters usually take out traffic to certain URL’s, certain hostnames, etc etc. This is useful when you want to view only a subset of data or want to see an abstraction of the data.

Ohnoes!

the event visitsThe anomaly occurs because filters in sub profiles won’t also filter events. So what? The events are associated with visits. Since events aren’t filtered, the visits come with them. If your parent profile has 562k visits with events (random number) then your sub profile will automatically have 562k visits even if you filtered out ALL traffic data. That’s exactly what happened in the screenshot from the first paragraph.

 

 

Hopefully that saves some head scratching out there. I’ll post again if I find a solution or if I notice that Google takes action on it.

Comments (7)

Sphinx Search: An Appetizer

April 18, 2009 at 8:38 pm · Filed under PHP

Recently I’ve been playing with Sphinx which promises to be a great site search engine solution at my place of work. This post isn’t meant to be a comprehensive tutorial but a brief overview meant to wet your appetite.

What Is Sphinx?

Sphinx is a standalone search engine that can be used to power search capability in many applications. It’s extremely quick, relevant, scalable, and highly configurable.

If you’re trying to create search functionality and using MySQL to do ‘LIKE’ searches, I highly recommend you at least look into using sphinx.

Get Sphinx

Download the source here at this link. My install experience on a Linux machine went very smoothly. It’s simply a matter of unpacking and then doing
./configure, make, make install. I decided to do ./configure --prefix=/usr/local/sphinx as that is the prefix used in the Sphinx documentation.

Here’s a quick rundown on the contents of your installation:

  • [sphinx dir]/etc/ – sphinx configuration file goes here
  • [sphinx dir]/var/data/ – index files
  • [sphinx dir]/bin/ – useful command line tools and the search daemon

Simple enough so far, right?

Setting Up Sphinx

Before you can start searching, you need to edit and create a sphinx.conf file. Go into the [sphinx dir]/etc/ folder and copy the example sphinx configuration file. Go through it and edit it to your hearts desire. Make sure to become good friends with the documentation as it’ll walk you through each and every available option.

The heart of config file is as follows:

  • Define sources. Each source includes an SQL query. This query is the information you want to be searchable. You can even include fields in this query which you declare as attributes. Afterwards, you’ll be able to sort and/or filter by these attributes.
  • Define indices. Each index points to a source and includes various additional options for how the information is searched. There can be multiple indices pointing to the same source. When searching, you have the ability to search one or more specific indices.

But again, make sure to check out Sphinx’s own documentation. After setting up your config file, run the [sphinx bin]/bin/indexer tool to collect the data and make your indices.

Use Sphinx!

To search directly on the index without going through the daemon, use [sphinx dir]/bin/search. Doing this after running indexer is a good idea as it bypasses api’s and daemons, which can be a source of bugs or confusion. You might also want to play with [sphinx dir]/bin/indextool which will give you some information about the indices you just created and along with [sphinx dir]/bin/search, can prove to be great debugging tools.  

If the search looks to be working well, go ahead and turn on the [sphinx dir]/bin/searchd daemon and try the API’s. Currently Sphinx provides API’s for PHP, Python, and Java. My experience was using the PHP version. These API’s have very useful options related to sorting, field weighting, filtering, etc.

Ohnoes – API/daemon trouble!

This caused me hours of head scratching so I’m hoping I can save some people a bit of frustration here. Remember to restart the searchd daemon after you change configuration options! I had issues when I was getting garbage search results from the PHP API after changing sphinx.conf.

Now go and give your users an awesome search experience!

Comments (1)

code deployment with lftp and md5deep

August 30, 2008 at 7:29 pm · Filed under linux ·Tagged , , ,

After a long time of developing on a live server, the startup that I work for now has a development server as a result of getting funded. This is great but also necessitated the creation of a process to deploy code from test -> live.

I tackled this deployment issue by creating a few scripts which make special use of a couple nifty tools – lftp and md5deep.

lftp
Lftp is FTP on steroids. This tool, among it’s features, allows scripting and mirroring which proved very valuable here.

md5deep
This tool is md5sum on steroids. The built in md5sum is great for verifying file integrity which is a great thing to do after moving code from one server to another. However, I just couldn’t figure out how to make it recursive and automaitcally do sub-directories. This is when I found md5deep. md5deep CAN work recursively and it does so quite well.

The process goes a little something like the following:
The test server runs a script which . . .

  • backs up the code (tar + gzip) and slaps that backup into a backup directory
  • runs md5deep on the directory and produces a file
  • uses lftp to mirror that directory (including md5deep’s file) to a deployment directory on the live site

The live server has a script which . . .

  • does a similar backup but of the live directory
  • moves code from the deployment directory to the live directory
  • runs md5deep to verify the integrity of all the files using the file created on the development server

If md5deep gives the okay, we’re good to go. Otherwise we investigate any files that seem to vary from development to live.

For more information on these two tools:
lftp
md5deep

Leave a Comment

Abstract Classes Vs. Interfaces

August 30, 2008 at 6:21 pm · Filed under Uncategorized

These two are new to PHP5 and are part of a new set of features that greatly improves OOP in PHP. At first glance abstract classes and interfaces look very similar but they have key differences. I’ll briefly introduce both and then lay out the significant differences.

Abstract Classes
An abstract class is created as a parent class which can be used when extended. Instead of saying:

class messages
{
}

you would do this instead… 

abstract class messages
{
}

In an abstract class, you can define properties and methods as you would normally with any class. The kicker is that you can add ‘abstract’ in front of your methods to just define that they are required in the child class and then the child class expands on those definitions. Kind of like a header file in C. Here’s an example:

abstract class messages
{
    var $msg;

    abstract public function showMessage();
    public function cleanMessage()

    public function setMessage($msg)
    {
        $this->msg = $msg;
    }
}

Interfaces on the other hand, also allow you to define a “header” of sorts so that you define what properties and methods should be in any class that uses it BUT it doesn’t allow you to expand on it. Here’s an example:

interface messages
{
    var $msg;

    public function showMessage();
    public function cleanMessage()
    public function setMessage($msg);
}

From there, you can then start building a class that “implements” that interface and build out the actual guts of what the interface defines:

class myMessage implements messages
{
    var $msg = 'Default Message';

    public function showMessage()
    {
        // ....
    }
    public function cleanMessage()
    {
        // ....
    }
    public function setMessage($msg);
    {
        // ....
    }
}

Interfaces are great when you want to provide a kind of API that other classes must follow. Also, a class can implement multiple interfaces whereas you can only extend one abstract class. You can do this by separating each interface with a comma . . . 

class myMessage implements messages, communication
{
    //.....
}

This is where the creative parts of a programmers brain must kick in . . . when do you use an abstract class and when does an interface make sense? This all depends on what the needs of your application are. Consider the differences between abstract classes and interfaces. And if that wasn’t enough to get your brain going . . . when is it best to just use a simple parent class vs. an abstract class.

I’d say keep these things in mind:

  • use a simple parent class when the child objects will share common properties and methods but then have a vast variety of it’s own properties & methods
  • use an abstract class when the child objects will both share common properties and methods but also have similar functionality that they perform in different ways
  • use interfaces when you simply want to define a rigid ‘API’ that the classes must follow

But all applications vary so you must do what will work best for your application.

And lastly . . . don’t spend a large chunk of your development time debating between these little intricacies. Speed is a competitive advantage. Make your choice and follow through with it.

Leave a Comment

One Salt Per Hashed Password

July 1, 2008 at 3:13 am · Filed under PHP, security ·Tagged , , ,

Intro
Hashing passwords are a great idea (Note: some people use the terms hashed and encrypted interchangeably). The basic premise is easy: make a hash of a password that only goes one way . . . password –> hashed string. You can’t go from hashed string to password. Then the process of logging in goes like this:

  • take a user-supplied username/password combination
  • run the same hash process on that password
  • take the result and and compare it to what you have in the database

Salt!
What could be even better? Adding a salt, of course! A salt is an additional string/characters that you add to the password before you make a hash of it. This increases the ‘randomness’ of the resulting hash. No longer is it made by just (for example) a weak password like ‘password’, but an extra salt is added to increase the strength.

Doing this, the process to login a user would be as follows:

  • take a user-supplied username/password combination
  • running the same hash process on that password, including the same salt
  • take the result and and compare it to what you have in the database

Even Better
Now, lets say hypothetically that an attacker gains access to your database full of hashed passwords. If they manage to figure out the hash and the salt, they have just gained access to all those passwords. An additional hurdle you can throw at a would-be attacker is making sure you use a different salt on each password. How? Well there’s many ways to do this and you can get creative. One common method is using something such as a username in the salt.

Now, the process would go like this:

  • take a user-supplied username/password combination
  • generate your “dynamic” salt using the person’s username. then use that salt and make a hash of the password
  • take the result and and compare it to what you have in the database

But What If . . .
What happens if the user changes their username? Wouldn’t that make it impossible to then generate the same salt originally used to hash the password? Well, yes. In this case, ask the user to also resubmit their password. Upon changing the username, regenerate the hash as well and store it in the database.

Interesting debate/discussion on this @ devnetwork.

Leave a Comment

Test Driven Development

June 30, 2008 at 6:41 am · Filed under PHP, tips ·Tagged , ,

What’s the best way to sleep at night and what’s the best way to be fairly sure that you’re code is as bullet proof as possible? Learn about test driven development and learn to love tests! There have been numerous articles that go in depth about this subject so I’ll {try} be brief about it…

No real purist will distill TDD into just a few guidelines but I will anyways:

  • Think about what functionality you want to build. Design and code a test that will test that functionality (before you code the actual functionality!).
  • Run the test and make sure it fails as expected.
  • Code your functionality.
  • Run your test again. Did it pass? Congrats, move on to the next task. Didn’t pass? You have some bug fixing to do.

The trick here is to put more emphasis on the design of the functionality than you’re used to. Think it through and design it so that it’s testable. If your test fails after coding the functionality and you feel compelled to fix the test rather than the functionality, then one of two things happened: 1) you did not give enough thought to the design of the functionality and make it easily testable or 2) you coded the test wrong. In either case, the design part of it all needed work.

Rinse, repeat, rinse, repeat.

The beauty in having a suite of tests is piece of mind. No, not just a warm fuzzy feeling. I’m talking about time-saving, bug-identifying, save-my-ass, look-ma-no-bugs piece of mind. Did you just add a new feature or refactor some code? Run your suite of tests. You’ll be able to much more quickly spot the bugs. Perhaps you’ll spot something that might have otherwise gone live.

I personally use SimpleTest to do my testing. PHPUnit is supposed to be pretty good as well.

Some links if you want to read more:
http://www.onpk.net/talks/fosdem2005/introduction_simpletest.html
http://www.developerspot.com/print/php/test-driven-development/

Comments (1)

Practice: Round 2

June 19, 2008 at 6:52 am · Filed under PHP ·Tagged , , ,

Passed again. Tempted to setup the real thing for this saturday.
2nd practice test

Leave a Comment

Not Bad . . .

June 17, 2008 at 3:24 am · Filed under PHP ·Tagged , , ,

After some on and off studying, I decided to take the ZCE practice test this evening. If the practice test is any indication than I’m looking forward to taking the real thing!

I passed 10 of the 12 areas and 5 of those with ‘excellent’. Funny thing was I failed “basic language.” It’s nice taking this in advance so I have an idea of the structure and difficulty of the real thing. It also gives me a chance to zero-in on the areas that I apparently have some trouble in. I’ll probably be scheduling the real exam pretty soon. Maybe take it within a month.
zce practice

Leave a Comment

Older Posts »