Compare commits

..

No commits in common. "efbd7bbcd5f6e94dfaeacbd9d44ec37d2f734cf2" and "d897ccce099e2f0a292caa6c4df1573cb214c8d7" have entirely different histories.

View file

@ -1,15 +1,14 @@
# Kripke-style untyped NbE in racket
[![status-badge](https://woodpecker.electriclam.com/api/badges/4/status.svg)](https://woodpecker.electriclam.com/repos/4)
An implementation of normalization by evaluation loosely based on [A
Denotational Account of Untyped Normalization by
Evaluation](https://www.brics.dk/RS/03/40/BRICS-RS-03-40.pdf) and
[Abel's thesis on NbE](https://www.cse.chalmers.se/~abela/habil.pdf).
Since the implementation is untyped, the `normalize` function only
gives you β-normal forms. However, you can get a little bit of
βη-equivalence by invoking Coquand's algorithm (`η-eq?`) on
β-normal forms.
gives you $\beta$-normal forms. However, you can get a little bit of
$\beta\eta$-equivalence by invoking Coquand's algorithm (`η-eq?`) on
$\beta$-normal forms.
I'm hoping to expand the repository into a benchmark suite where I can
visualize the performance trade-off of type-directed NbE compared to