I would argue that it only becomes "research" when reading up on existing literature is involved. Basically, if inspiration hits you and you just fire up your IDE to test it, you are not aware if what you are doing is in any way novel. That's not really the nature of research.
Edit: (I should add that even if something is not novel in a scientific sense, it can still be extremely interesting and cool to do - however, it will likely never lead to a paper)
That is not to say that a researcher is never allowed to just follow her/his inspiration. It has certainly happened to me more than a few times that I had a crazy idea and just ran with it. Oftentimes, the quickly led nowhere and I just forgot about it. Sometimes, initial results proved interesting and then I went ahead and did proper research before going further.
Edit 2: I have just noticed that I did not actually answer your question:
Research before surveying the literature - is it ever a good idea?
No, not really. If you are doing "research" (in the sense that you are hoping to publish it, and not doing it out of personal interest, or as a training exercise, etc.), the only senseful way is to start with a literature review. One needs to keep in mind that in most fields, most of the "low-hanging fruits" (i.e., obvious ideas) are long solved. So just going ahead and doing something in the hopes that it may be novel has a significant error rate.