In any web app, it seems that there's always something that you want to be different on the live site versus the dev site: a different database name, verbose logging on, performance metrics enabled, etc. I find myself implementing the live vs. dev switch a little differently every time, though. What do other people do? Do you trigger dev behavior via hostname? Via a config file? Via an URL substring? Manually?
In my current Catalyst project, I wrote this little hack. It's inelegant, but simple to understand.
# If we aren't running on my development computer, refuse to load the dev features
if (! -d '/Users/chris') {
# trick Perl into thinking this file is already loaded
$INC{'MyApp/Controller/Dev.pm'} = 1;
}
#!/usr/bin/perl
use MyApp::Config ROOT => '/home/production/myapp';
use MyApp;
MyApp->run;
If I'm deploying on mod_perl, I tend to set this sort of thing up using PerlSetVar in the Apache config rather than create an application-specific config file.
I've done things like that in the past (or sniffing version numbers of various programs/libraries, and guessing environment therefrom), but I suspect that way lies madness, ultimately -- at least, pick one thing and stick with it, or you will inevitably forget what the decider is for any given app.
I tend to prefer the config file method -- it also allows a slightly wider matrix (development vs. production is common, but many leave off staging, or laptop, or home machine). Ideally, there would be two axes: "configuration" (passwords, databases, etc.) and "mode" (development vs. production, etc.)
If you have a "dev -> stage -> prod" setup, where stage is ideally as close as possible to prod, then what you have won't work. You need a more general way to get the configuration, like a config layer that parts of the application can call through (database, etc. goes through this layer).
Dealing with different configurations sucks, blech. Like when you want to sync your dev from prod, you always end up changing something: a server name in a database table, in a config file, etc. I guess ideally you need a way to automate the application of a configuration to each machine. So if you do like a raid disk copy to sync your dev machine from prod, then you have a script that has some parameter, like "./apply-config --to-server=dev.example.com --type=dev" or "./apply-config --to-server=prod.example.com --type=prod", where the script knows that "prod" means a certain configuration but "dev" means another config.
Well, I guess that has less to do with programming than sysadminning, though.