Hi all, I've got a question about how best to structure my code and classes.

The example I'm looking at the moment is resource management, ie, loading resources when their needed and unloading them when there not. However this is a question which could equally apply to other classes.

I've made a ResourceDatabase class which has a HashMap mapping resource name strings to instances of a Resource class which stores information about a resource's type, file location, how to load it, etc.

Currently to add a resource to the resource database you call a "addResource" function which takes a file name and other information.

This is obviously quite tedious if you have a lot of resources, hence a function that scans a folder and its subfolders for files and automatically adds any found resources to the database would be very helpful.

However, the ResourceDatabase is currently only tasked with managing a set of resources so adding this as a member function seems like its adding unrelated functionality. Also in the future I may instead want to store all resources in a single package file and add another function to scan it and add the resources.

Since the "scanFolder" function can just as easily be implemented using the ResourceDatabase's public interface is it best practice to add it as a member function or as a standalone function? If wasn't added as a member function where would you put it, just some lose function somewhere or make a class with static functions for scanning folders, package files etc?

I realize this isn't super important and either would work just as well and that responses will be opinion heavy, but I'm interested in what people think is the best way of dividing up different functionality.

A possible solution:

A ResourceLoader is responsible for loading. It has an ability to mount sources. One kind of source is a directory, another is an archive file. When a directory is mounted, then it is scanned for resource files, and each found resource file is added to an internal table of content. Within each entry the directory and file name is stored as from where to load the resource if it is requested. Also, when an archive file is mounted, it is opened and its table of content is added to the one stored with the ResourceLoader, but only for entries that are not already stored therein (this is a policy question and can be handled in another way if needed). Whenever a client requests a resource from the loader, the loader investigates its table of content and reads the resource as is prescribed there.

With the above mechanism one has the possibility to do something like this: The general source of resources is a small couple of archive files. But is is possible to have an overlay in form of a directory, useful for developing (hot swap), updates, and perhaps modding.

A ResourceCache is the instance that is responsible to hold resources in memory. Whenever a resource is requested, the cache returns the stored data if available, or else nil is returned. The cache implements different strategies how and when resources are allowed to be dropped.

A ResourceManager is the API / wrapper for this system. Internally it uses one or more ResourceCache instances and one (perhaps more) ResourceLoader instance. It parametrizes the loader considering the configuration of mounting. Whenever a client requests a resource, the manager first requests it from a cache. If data is returned then we're done. If not, then the loader is requested to load the resource. The returned data is then inserted into the cache and returned to the client.

In the system described above there is the need for creating the tables of content as written into archive files. This is of course the job of the tool chain. The loader has the ability to scan a directory and to investigate detected files for their usefulness. If you have a sub-system that deals with files, the actual scanning is obviously part of it. This may be as simple as a function returning a list.