Ah, I am being a little silly. Tunnel vision. The fusion only matters if you're skipping a bunch of lists. So ... I don't know why your results are worse, really.
On Aug 19, 2014 6:42 PM, "Bertram Felgenhauer" <bertram.felgenhauer@googlemail.com> wrote:
> Interestingly, initsQ looks better than initsT in these benchmarks.
I love the simplicity of initsT, but unlike initsQ (with Joachim Breitner's fusing scanl') it won't fuse properly with with a list producer. I don't know if that can be fixed or not. If so ... that would be great. In any case, it would be very interesting indeed to see how it performs in a library that uses stream fusion instead of fold/build—there's a library in the vector package for that, and it currently uses the broken version of inits; you might want to try initsT out with that one.