Very nice, thanks. I'm leaning toward windows as well. nctx is shorter, but it's not clear I use it often enough to justify the terseness.
Those single quote issues on the last line are def bugs, thanks.
partition-words isn't a tokenizer, it partitions the string along word boundaries. Hence the tokens containing spaces and punctuation. You can easily extract tokens with a single keep, but I found it useful to be able to easily recombine the partitions to reconstruct the original string.
