So, this app I hadn't been looking at in a few months did not work at all. I traced the bug down to a method that itself called a parent method. The parent only contains __call and __callStatic methods, and for some reason __callStatic was called, although the class it was called from was an object instance.

After using some sample code from the PHP manual, they spotted how things truly worked - the call is always sent to __callStatic when called from the extending class but not when called directly. Apparently they aren't the only ones that noticed the bug but noted that it was removed in PHP 5.3.4 to fix the issue.

One of the many things you don't often do with PHP (actually, I'm not sure you do this much with any language) is daemonize a process. A daemon is program that runs in the background (read more here). On Unix systems, processes are usually created by forking the init process and then manipulating the process to your liking. To create a daemon though, you need to get the init process to adopt your process. To do that, as soon as you fork the parent process, you kill the parent. Since you child process is parent-less, the init process generally adopts it. Once that happens, your process has been daemonized.

He uses the pcntl_fork function to spawn off the child process, detach it from a terminal window, create a ".pid" file so the system knows about it and then, of course, have the child script do something.

I'm in the process of creating models for my CI project, and realized that certain functions within the models were getting repetitive. Using CI's ability to create my own custom libraries, I was able to create my own custom 'Model' which extends from the core 'Model' object. How this simple architecture has cleaned up my code is simply remarkable.

Rather than overwriting the main Model.php file with some of your own changes, he suggests creating a new library, a "parent model" that can be extended instead to provide some common functions that all of your application's models might need. For something a bit more complex, he also points to this library that extends the models to give it CakePHP-like functionality.

The Doctrine library comes with a feature called nested set, which makes saving trees in a database easy. However, it's quite easy to accidentally cause a lot of extra unneeded queries if not being careful. Here are some pointers to keep in mind while working with the nested set, and some example queries to make understanding it easier.

He gives an example, showing how to get rows from the database - parent and child - and some optimization tips to keep things light. There's also some pros and cons included for doing it either way (the standard fetching or using the more optimized versions).

If the software calls for only 2 levels of categorization (Parent and Child only), a simple nested for loop will suffice. However, software requirements change and you'll soon find yourself up shit creek without a paddle if you need to support 3 or 4 levels of nesting. [...] To those who's training is less formal (most web developers I meet have practical training, not formal), I'll help you out: Tree Traversals (or if you are completely lost, Recursion).

He creates a recursive function that, when passed in a category set with different types in it, can handle each of them and then calls itself again with the new child data. His sample code creates url out of a set of categories.

This weekend I wrapped up a few small tests and sent the patch in and it was subsequently pushed to php 5.3 and php 6.0. Now, this is not at all the way I wanted things to work, in all honesty I think the patch is pretty hokey but unfortunately nobody really spoke up in support of the changes I wanted to make to parent:: in regards to LSB.

His example shows how to override a static method and push that new method's execution to the parent class (in two ways - safe using forward_static_call and the not so safe calling itself with a parent:: override).

PHP has a function serialize, which can create a hash from any type of variable, scalars, array, but objects as well. Using the unserialize function, PHP can recreate the variable from the serialized hashed. If we look at how an object is serialized, we see only the properties and the class name are stored.

His method allows for class manipulation via changes to the serialized class information (like changing the value of the name parameter). His "casttoclass()" function makes changing this value simple.

I won't rehash all of the arguments as you can quite easily find out my full thoughts by previous posts and on that mailing list thread. To put it simply I feel that somehow there needs to be a way to call methods in a parent class without losing the ability to reference back to the original called static.

He gives an example of what he's talking about (code) and, to help further his cause he's created three patches to make it work how he'd like.

One of the nice things about using a scripting language is that it automates garbage collection for you. You don't have to worry about releasing memory when you're done with your variables. [...] But there is at least one circumstance in which PHP will not free memory for you when you call unset(). Cf. http://bugs.php.net/bug.php?id=33595.

The problem is that, when trying to unset the parent object in a parent/child relationship, the memory space for the parent reference in the child isn't removed. The problem, left unchecked can cause a script to go over the memory limit pretty easily (especially with larger objects).

He suggests two solutions - one that the user can control (calling a destructor before unsetting the object) or a PHP internals method that involves fixing issues with the reference counts created with the objects.

The good news is late static binding has been introduced into head and looks like it will be merged into 5.3 before it is released. The horrible news is I really don't think the patch went as far as it needs to.

He talks about the original intention of the functionality (flexible inheritance for static methods/properties/constants) and how it was implemented, but with one small issue - that "static will ALWAYS return the 'resolved' name of the class used to call the current function". He illustrates with a code example showing an extended class returning a static property.

He also mentions two suggestions to help fix this issue:

• setting the behavior of parent:: such that it forwards the calling class through the next function call.
• introducing another scope [...] using a new keyword so parent:: could remain the same