where's that old ngram study

gist.github.com/lynn/bc6517f3b150e4535748a813ad608aa2

er, github.com/DennisMitchell/jellylanguage/issues/91

wait the linked analysis code doesn't even parse strings out what the hell

well, one parsing strings out later it looks like there are still way more hits for bare x than for x@ in that old corpus

so there's probably something obvious i'm missing :P

okay, naively counting how many of those hits have a nilad on the left, it's 37 out of 65

but that's still 25 non-reversed x to 3 x@

without adicity of the links themselves in the corpus this is fallible but maybe i can check how many have a grouped link to the right :P

because there's probably a lot of times filtering-x with the raw input being filtered is shorter with something grouping the right than with the filter logic on the left and @