We did the whole new on substring trick on a sparse csv. Turned out we had over a hundred mb of empty strings. Lots and lots of duplicates. So empty string is s

We did the whole new on substring trick on a sparse csv. Turned out we had over a hundred mb of empty strings. Lots and lots of duplicates. So empty string is special cased to return a constant