I have some shared projects that are under version control (concretely svn and bazaar, but I'm seeking for a general solution), but the datasets the projects use are not (too big and shared by different projects).
In the source code I need to "store" somewhere the path to the dataset. The path is possibly different for each user, so hardcoding is definitely a bad idea (as always, I guess).
My actual workaround is to hardcode a text file (say "dataPath.txt") where the actual path is stored, and this file is not under version control (each project contributor creates his own file with his customised info). The solution is, however, quite fragile: 1) if some contributor add to versione control the file it is annoying 2) when I export the "executable", I need to move around the file that is supposed to be in the same dir (relative path).
In my concrete case I'm using Java, so I find this question relevant (even if I've never used properties), but I would like to know if there are more general techniques that can be reused with different programming languages.
Write your program so that it accepts the path to the dataset as a command-line argument. Make sure there are a) sensible defaults if the dataset file is not specified, or b) the program exits gracefully if no dataset file is provided. There is no need to hard-code dataset paths in the source. Then you'd invoke the program e.g. like this (of course you can take any other command-line option character you like :-) ):
prog -d dataPath.txtIn general, providing such settings in a config file is a good idea. With Java, properties help (as pointed out in the SO question you linked). In other languages I'd probably use a JSON-formatted settings file -- parsing libraries are available.