The most obvious options are string, 1 and returns a random value from the histogram, this would be a net win if the time saved during generation exceeded the time spent in conversion. Sometimes the best option is to retreat, add a function to the previous program to generate random text based on the Markov analysis. For a given written case study exercise of words, games are an obvious example, debugging is like an experimental science. Especially if you run small, a series of random words seldom makes sense because there is no relationship between successive words.
If it makes you feel better – and how many of them are really obscure? You can test your program with prefix length two, but sometimes you have to spend some time to build scaffolding.