[crazyirish]

So in your case doing Olymp* is actually not too difficult (thats just stemming) but doing actual regex matching accross the internets would be hard sauce. One of the traditional ways of storeing an index like this is word -> {documents}. Doing a search for a set of words is then not too expensive, however for full regex support they would have to look at every word entry. Thats just sadpanda to the max.