[Ixiaus]

This is how I feel about it. Mercury is an elegant language but I think the applicable domain is restricted to a very narrow subset of the industry. As a [primarily] web programmer I will generally only become more "cultured" by studying esoteric languages (Scheme, Haskell, Mercury, Curry) - that acculturation does diffuse into my day-to-day practice, though. I find myself structuring code in a novel manner (often times it's more clean, cogent, and effective too) that I would not normally have done with my narrower knowledge space, pre-saturation. I also see myself creating logic and using data structures that weren't historically typical of me.

I think every programmer should study esoteric languages. It's akin to the contrast of arithmetic and mathematical logic - you have the opportunity to appreciate and experience the ideal rather than the mundane. Not that beauty can't be found in building a web application (or any other application for that matter) using a framework and one's own set of libraries; I do everyday because I pour the expression of who I am into my practice - the thing is, it's [esoteric languages] the same sort of joy one derives from contemplating deep philosophical questions or from deliberately cultivating character through introspection.

I felt like whipping out the PhilosoRaptor tonight I guess...

[swannodette]

The fact that people write tediously comprehensive imperative unit tests to verify their programs should be a warning sign that the use cases for logic programming may be more much general than you think. A hunch, nothing more. I'm just starting to dig into this stuff.

[silentbicycle]

This!

People are itching for declarative testing systems. Writing testing code by hand is tedious and error-prone. Static typing works for a lot of things, but isn't completely satisfactory. Randomized testing is also useful (give a function that tests a property, give random_int -> value generators for its arguments, let it find edge cases), but those are only two facets of a much bigger picture.

I'm really interested in more modern takes on Prolog, but from what I've read about Mercury, it seems to throw out my favorite aspect: computing with partial information. Constraint logic programming seems like a more valuable direction, though I see the appeal in a Prolog-esque language informed by modern type theory, and I'm particularly interested in what it does with linear types.