Hi,
I am building an application currently in PHP and I am trying to decide on whether to use a pre-existing framework like codeigniter or build my own framework. The application needs to be really scalable and I want to be completely in control of it which makes me think I should build my own but at the same time I dont want to reinvent the wheel if I dont have to.
Any advice greatly appreciated.
Thanks
The short answer is symfony IHMO.
The reasons are:
Of course the previous points are valid for a very small set of professionally made frameworks! My favorite is symfony.
Make a list of requirements for your framework (ORM, PHP 5.3, PDO, etc). Then iterate over the existing frameworks and narrow them down to find the ones matching with your requirements. Then look at the codebase, documentation, community, project activity – does it feel something you would like to work with? Also be realistic about the time needed to to implement all your requirements by yourself – do you want to focus to building an application or a framework?
The decision is very plain and simple.
If you want to learn and have full control - go for your own
If you want just to make money fast - go for ready made one.
While I've build my own CMS framework in the past, and used custom (in-house) general php frameworks, I'd find an active framework that fits your development style, and use it.
Unless your main product/application is the framework. But it doesn't seem that way.
Your concerns about control and scalability should be applied to the host of frameworks out there - giving you a short list of options that fit your requirements.
Certainly, it's not a question of 'in-house' vrs 'public', then once you've made that call just pick any old framework.
To answer the question behind the question, for a framework that gives you complete control and should be able to scale reasonable well (I'm not sure how you need the framework to scale), I would suggest the Zend Framework. You can use single parts of it, rewrite what you want, and it's far more than just a MVC implementation.
Update: A quick example of customization with Zend. If you don't want to use their MVC stack, but need something to route requests, you can just use Zend's router library. If you like the MVC stack, but hate the way the router works, you can just implement the interface and write your own router.
This applies outside the MVC stack as well. Zend has a ton of libraries for mail, rss feeds, caching, auth, db, etc. Use what you want, and ignore the rest. Extend what you want, most of the framework is tiered with interfaces/abstract/generalized classes that you can build on if the standard functionality doesn't suite your needs.
I work in a company that initially wrote its own framework, built by a guy that worked here. It was only used on one project. The reason for this is that we soon realized that although it was clever and very good, there was no documentation for it. So if we employ another developer or a freelance guy, they'd have to learn it.
We run with CakePHP for a while, which is popular, but it does seem a mess. We eventually settled on KohanaPHP. Easy to extend, some good documentation (probably not up there with some others though), nicely formatted code (meaning if you can't find documentation you can quickly work out whats going on). The way the framework is written makes it pretty easy for a developer to pick up and follow what is going on. Whereas we always had trouble doing this with CakePHP.
I think the only argument for rolling your own framework is that you may want some highly customised stuff. But Kohana is so easy to extend, you can just throw it in there. You don't have to use their packed libraries if you really don't want to.
That said, some project I don't bother with a framework at all, just some sort of routing solution, like GluePHP. Since it would be overkill to use a full stack framework.
Use an existing framework.
First of all creating a framework from scratch represents a massive investment in time and effort. The process involves a lot of trial and error, because you're designing something that needs to be both simple and powerful. For every design decision you'll have to ask yourself how it will affect every single future project that will be built on your framework.
You'd think that you could take each design decision and weigh it against the requirements like you would do for any other software project, but the thing is you don't know your requirements. You can't know them, because a framework is supposed to be able to do almost anything (or have the ability to be extended to do almost anything) within its domain. Future project a will need to be able to do x. Can your framework allow that without turning it into spaghetti code? And what if project b needs to do y? What if project c needs to do z?
Have you predicted everything?
Now the normal response to this is that if something doesn't work, you'll just change it in the future. It's software after all. A framework however isn't like a simple application. It's supposed to have an interface and once you expose that to the software that will be using it, you can't change it. You can extend it, but not change it. So now you have to think about deprecating methods, api versions and version compatibility. It's a whole new set of problems to deal with along with normal framework maintenance and new application writing.
Then there's documentation. You need an API, tutorials, example code. Once you build your own framework you have to deal with this as well. You could ignore it, but I assure you that eventually you yourself will need to find out what that method you wrote 6 months ago does. What does it return? What if special case x happens? Have you written all that down, or do you need to step through the code again? And I wont even mention how easy it will be for a new team member to get started on a custom framework whose documentation lies completely or at least mostly in your head.
You also have to acknowledge that unless you're working with the very best and brightest (and have a budget to match) you'll never have the extensive set of libraries that existing frameworks boast. Can you analyze, design, code, test and debug faster than an open-source community?
Finally you should ask yourself if you are proficient enough to be writing a framework. Have you dived deep into the code of a modern OO PHP5 framework to find out what makes it tick? And most importantly do you know why it does things that particular way? Keep in mind that any mistake you make in your design can blow up in your face months from now and you can end up paying for them over and over again.
To sum things up I'd advice you to go with an existing framework; it doesn't mean however that you have to pick one and like it. Take the time you'd otherwise devote to developing a new framework and devote it to learning an existing one. Then you can extend it to fit your needs. Also remember that there could be things you wont be able to do. But I assure you there would be things you wouldn't be able to do with your own framework either, so it doesn't matter all that much. A framework imposes a few limitations. It's the price you pay for being able to develop applications faster.
If you were just trying to practice, I'd suggest writing your own. Otherwise, though, definitely go with using an existing one. Symfony ftw.
I'm building my own and I'm very glad I did in terms of learning, which is important to me. Also I LOVE to be in total control of my code. It comes with lots of negatives though, the big one being I'm the only one that knows how to use it. Also, a lot of development time is spent on improving my framework rather than delivering products to my clients. But I can't stress enough that I really, really enjoy building and using it.
If you want to learn (A LOT) build your own. If you just want to get work done use an existing one. (Before starting my own I almost went with CodeIgniter)