@mflatt while doing some pycket benchmarking, @cadr noticed that reading a file of flonums is about 3x slower with the new expander (presumably because the reader is now in Racket). Is that something you know about/expect, or is it worth looking into more?

@mflatt In particular, Racket 6.12 and Racket 7.0.0.10 have 3x difference for this benchmark: https://github.com/pycket/pycket-bench/blob/master/CrossBenchmarks/src/sum1.scm

It makes sense, and I probably looked at it some when implementing the reader, but I don’t know or remember how difficult it would be to improve. I certainly wasn’t any happier with the new number parser than the old one.