1
$\begingroup$

Currently I'm working on my thesis. I'm writing about subgraph isomorphism problem and various algorithms, solving this problem. One of my task is to integrate that problem in a testing framework, developed by my mentor. It's some kind of universal framework, where you define a problem, implement some algorithms, solving the problem, perform some tests, analyze and compare different algorithms...

I've searched the web about similar frameworks, but I haven't found any framework, similar to the one described above. Is there really no such thing as universal framework to define some problem, run algorithms and analyse the results?

$\endgroup$
  • 1
    $\begingroup$ Why is JUnit not an answer? $\endgroup$ – Raphael Dec 18 '14 at 19:46
  • 1
    $\begingroup$ You can certainly make JUnit do this or -- depending on your API -- quickly rig a script together. I'm sure there are plenty of tools/frameworks that offer runtime measurements out of the box (GNU/Linux has time, for instance). You need to give more detail. Which features not available in standard tools do you need? (Also, I'm not sure if this is ontopic. You seem to be looking for programmer's tools, not an artifact of computer science.) $\endgroup$ – Raphael Dec 18 '14 at 20:38
  • 1
    $\begingroup$ This question is not well formulated in a form that is a good fit for this site. 1. What is the task you are trying to achieve? ("Integrate into my advisor's testing framework" is a how, but focus first on what you want to achieve.) 2. What are you asking for a universal testing framework, when what you really want to do is compare different subgraph isomorphism algorithms? You've put the cart before the horse. Actually, it's not clear why you need a framework at all. $\endgroup$ – D.W. Dec 18 '14 at 23:29
  • 1
    $\begingroup$ 3. What have you tried? Try implementing the algorithms and evaluating them on some workload. This seems like a straightforward of programming/engineering/tools, not a question about the concepts or science of CS; if so, it's off-topic for this site. So, what specifically is the question? You might need to spend a few weeks trying it before you can formulate a well-posed conceptual question. $\endgroup$ – D.W. Dec 18 '14 at 23:30
  • 1
    $\begingroup$ I am interested in this discussion as I feel (very subjectively) there is too little work in objective experimental comparison of competing or similar algorithms. I don't mean complexity, but just "informal" efficiency, including constant factors or even terms. Having a testing framework helps that purpose, but provided it is used in a systematic way. Universality is probably exactly what you do not want, but rather a specialized framework (possibly tailored in a universal one), where you emphasize operations that are meaningful in the theory you are looking at. And that is not SE. CC @D.W. $\endgroup$ – babou Dec 19 '14 at 11:59