Andy Gill & Graham Hutton's Worker/Wrapper in HOLCF, finally.

Finally completed this bit of work, started sometime around December 2007. The Isabelle proofs available here remain a bit rough, but I did manage to find a bug in the original presentation by Andy Gill and Graham Hutton — unwrap must be strict for fusion to be correct. From there it is a short step to a fusion rule that provably works. I've submitted a corrigendum-ish paper to the JFP and we'll see what happens now.

Brian Huffman's soundness proofs for stream fusion

I never post about anything technical I do any more, so instead I'll talk about someone else's stuff. Brian Huffman, HOLCF's last man standing, has a brand new entry in the Archive of Formal Proofs about verifying stream fusion. From a cursory read I thought he had opted for the "correct but GHC can't optimise it" variant floated in the original paper, but it turns out he has modelled the actual implementation. As always his proofs are mystifying in their clarity and succinctness.

Sweet, someone has ported the library to Standard ML.

Incidentally I stumbled upon VeriFun, which looks like it might be interesting, albeit inert. I wonder what their underlying logic is.

peteg's blog 2009-05-12-StreamFusion.autumn

Andy Gill & Graham Hutton's Worker/Wrapper in HOLCF, finally.

Brian Huffman's soundness proofs for stream fusion