See my post in this discussion "You can write <strike>FORTRAN</strike>Prolog in any language". One subset of Prolog is a nice concise declarative notation for

See my post in this discussion "You can write <strike>FORTRAN</strike>Prolog in any language". One subset of Prolog is a nice concise declarative notation for rule matching engines. Back when I worked at Google, one of my colleagues (Jeremy) did some Bayesian inference on "1% experiment" results, where he needed to see which observations matched a complex hypothesis and complex rules for the context. So, Jeremy found it easiest to think about the problem in terms of Prolog, and wrote some Python objects whose constructors defined these rules. He also whipped up a memoizing backtracking engine that effectively lazily generated a DFA over experiment observations. He talked and thought in terms of Prolog, because that's the mental tool that he used, but the programming tool was Python. Interesting guy, that Jeremy. He used to do cacao price forecasting for Mars (Mars Bar/M&Ms). He also enjoyed craft beer and carving pumpkins with chainsaws. We also worked in the same (NYC) office with Fergus Henders, who in grad school wrote Mercury, a statically typed dialect of Prolog with an optimizing native code compiler.