sooo it turns out the bad data I found from #openaddresses was more pervasive than I thought. Street field is a mess of prefix/suffix only, direction only, street name only and everything in between.

Most of the way done for mitigation of the first combinations I was aware of, but found more before I finished debugging.

Probably going to skip this month's opensupermaps release to work on this. It requires adding a suffix and direction filter which means it'll only work for English currently and probably add a ton of processing time. Also planning to add filters for length>5 characters and having 2 words in sql to run before the complex filter to hopefully slim down what gets run through the complex filter.