Identifying the project root in PHP
May 15, 2009
archive
I currently have problems with identifying the project root in a PHP project. The problem applies to PHP, but the discussion is general and applies to the other languages and environments as well.
Problem description
I will use my PHP-based application framework, Wigbi, as a contextual example, so that we have some real-life situations.
Wigbi has many classes that may be used anywhere in the project folder hierarchy,
so identifying the project root is a major issue. Wigbi provides two ways of doing
this - the two static Wigbi::rootPath and Wigbi::clientRootPath properties.
Before I explain why we need two properties for this, let me discuss why absolute and relative paths are not good alternatives to handle this situation.
Absolute and relative paths
Absolute paths use the server root (e.g. /img/pixel.gif) as a starting point. They shouldn’t be used, since the absolute root may differ between production and live environments. If you don’t want to force a solution to have the same structure in all environments, absolute paths are a no go.
Relative paths (e.g. ../img/pixel.gif) are also common, but mustn’t be used by files that may be used in various parts of the project folder hierarchy. For instance, Wigbi may be included by both index.php and music/index.php, which makes relative paths unusable. The same applies to ASP.NET User Controls, which may be added to pages anywhere in the folder hierarchy.
You may think that relative paths are OK for web pages, since pages have a fixed position in the project folder (e.g. news/index.php). This is also not true, since URL rewriting may throw pages around so that /news/latest.php become /latestnews. This will cause relative paths to fail.
In short, absolute and relative paths are problematic. Only use them when you fully understand how they are to be used, and the drawbacks of each approach.
Identifying the project root
ASP.NET has a great way of identifying the project root: the ~ variable. ~ which can be added as a prefix to any path that is to be used by the client. ASP.NET will then convert it to the application root path when the page is parsed by the server. This means that you can always use ~ and be sure that it will resolve to the project root, which should be the only starting point you have to consider.
However, this only applies to client related paths, not for server files. Wigbi
therefore has two properties that can be used to identify the project root:
Wigbi::rootPath and Wigbi::clientRootPath. The first is to be used on the
server and the second in the browser.
Wigbi::rootPath
Wigbi::rootPath identifies the project root for the executing server file. For
instance, when Wigbi is started and loads the current language file, it uses this
property to locate the file, starting from the path to the project root from where
the code executes. As we’ll see, this isn’t always the executing page.
If you write a class that is to work with files, Wigbi::rootPath will work, no
matter where in the folder hierarchy you are.
Wigbi::clientRootPath
The fact that Wigbi::rootpath applies to the executing page causes problems for
the AJAX pipeline, since all pages call wigbi/postBack.php. This means that any
file paths that are sent to the client (image paths, links etc.) will be relative
to this page, if you use Wigbi::rootpath.
Due to this, there is a Wigbi::clientRootPath property that is set whenever an
AJAX operation is executed. It takes the root path used by the page and sends it
so that any path that is returned to the client will be relative to the page. If
there is no set client root path, the property is equal to the root path.
This works, but I would prefer to have a single variable instead of these two. I also I didn’t predict some problems that I discovered later.
URL rewriting = PROBLEM!!!
So, we have these properties:
Wigbi::rootPath- used by PHP files to work with files on the server.Wigbi::clientRootPath- used by any path that it to be used by the client.
This should be sufficient, right? Wrong! As I added URL rewrite rules, I noticed that themes and JavaScript code stopped working.
The problem turned out to be:
- The root path is the path of the executing page (index.php), which is an empty string.
- In PHP, the root path is correct, since the physical file is executed.
- In the browser, the root path is incorrect, since the page thinks it has an address that it hasn’t.
- PHP doesn’t know about the resulting url, only the expected.
- As such, the root path should be
../, but becomes blank.
Any ideas how to handle this?