TDD, configuration management, configuration patterns (PHP)

Question

I'm writing a PHP application, following test driven development principles.

I want two different configurations: one for development, and one for production.

I will be continually expanding the application's features, so one problem is I don't want the production credentials/configuration to be overwritten each time I upload an upgrade. This is just one heuristic. And it could be done 100 different ways. I can't see the wood for the trees, I need help!

Example 1: the application interracts with Ebay's API. For the application to authenticate itself with Ebay, it needs to submit authentication keys. But I'll need two sets of keys: one for development in Ebay's sandbox, and another for production. Eventually these will be stored a database or similar, since they expire and need renewing, but pretend that's not the case - assume they are static.

Example 2: Database host, name and credentials which would be different in testing and production

I'm having trouble articulating my question. In short, want to follow TDD best practice.

My instinct would be to have a configuration file which is read by some class. That gives two options:

1. Either there would be two files, one for tests and one for production, or
2. just one config file, who's contents would be altered for production.

But say we choose option 1, we're back to square one, because somewhere there needs to be a setting saying whether this the test or production environment.

And say we choose option 2, the file's location would still need(?) to be hard coded somewhere.

And then there's the scenario where objects need to be swapped depending on whether it's production or development. AFAIK the factory pattern solves this, but again, somewhere, the code needs to be told what environment we're in, and/or which of the two classes to use.

In books I've read, the principle I think is to treat the application as an onion, feeding in configuration information to the application so the whole application is context-independent. So for a command-line tool, configuration could be given by command line arguments. But how might this be done for a Java Swing application, or a PHP web app?

Answer 1

One solution I've used in the past is to have the application use the hostname of the machine it's running on to locate the configuration file, so if it's running on a machine called www7, it'll load config.www7.inc, whereas if it's running on minerva it'll load config.minerva.inc. It takes a moment to set up a new machine, but after that it's pretty easy.

Answer 2

You don't need a setting that says if you are in development or production or not. You just need settings that tell your application how to handle certain things. API keys, database connection strings,... these are all basic 'read this value and use it' things and this is the kind of stuff that configuration files are good at.

It's a little different if your implementation differs. For example, let's say that you want your development environment to write e-mails to a logging directory and the production environment to actually mail them to the recipient. You write all your code against an interface ('IMailer') and write two implementations: SMTPMailer and LogMailer. Your config file contains an entry that says what implementation IMailer should use. For development/testing environment, you simply have that entry point to the LogMailer. Look into IoC and IoC frameworks for that.

Answer 3

A very clean way is using Git:

Git is a free and open source distributed version control system designed to handle everything from small to very large projects with speed and efficiency.

For more information and downloads: https://git-scm.com

With Git you use branches, such as develop, master, feature/, etc. Read more here: https://www.atlassian.com/git/tutorials/comparing-workflows

Store production config in the master branch, the development config in develop branch, etc.

Answer 4

One solution that sounds vaguely reasonable stems from this 12 factor methodology page on configuration.

The most focused guidance I have found so far is here in the 12 factor methodology. They suggest that such configurations should be stored in environment variables. But how would that be done in PHP? Here it is suggested to use a .htaccess file.

Answer 5

Without seeing the actual code or tests it's hard to say, but why do you need to actually use the service in the test? You'd be better off mocking the ebay API and not worrying about live/production differences.