Keep original unigrams in tokens_compound()

[koheiw]

It is possible to keep unigrams in tokens_ngrams() by setting n = 1 but it is not in tokens_compound(). Keeping unigrams makes it easy to apply pattern matching in the downstream. How about adding an option like keep_unigram?

[kbenoit]

OK, good idea. keep_unigrams = FALSE (plural), or

action = c("replace", "add")

but the action would only make sense if there were a possible third option, and I cannot think of any.