Nothing? I just use the language features and I use libraries for specific things.
I do not use an automagic configuration and dependency injection boondoggle. I read config in main()
, create the objects I need, and do what I need. It is easy to see what my program does; it is easy to see where configuration comes from. It is easy to test any component, because you can clearly see what you need to provide to build the entry point objects.
Spring combines combinations of environment variables, system properties, files, and classpath resources, and handles a variety of patterns (e.g.: aConfigOption could be configured by system property
A_CONFIG_OPTION
,aConfigOption
, or several other possibilities), so tracking down where the configuration came from is not always easy. Sometimes you think you can just set a property, but it turns out another property triggers loading a resource that overrides yours. This would be fine if applications/libraries clearly documented how to configure them, but most say “config via spring, good luck lol”.And good luck if you are trying to use two different components both built on Spring, and they both rely on the dependency injector settings “db.url”. Now you have to start playing games with dependency injector scopes.