My current company is looking into our "next generation" platform for when our datasets exceed what we can do in R. R may not be the best language, but it's great for exploratory data science, has the best or the only library out there for some ML purposes, and we've done a lot of things to make it production-worthy (on small- and medium-data). We'll probably need to involve something else, down the road, when our data sets get larger than what fits on one box. The leading candidates are Haskell, Clojure, and Scala (Scala because of Spark). I'll have to evaluate the languages fairly and relative to our needs, but I hope Haskell wins for a number of reasons, including the fact that Chicago + Haskell is an unfilled niche and we'd attract a ton of talent.
For those who've taken Haskell this far into production: have you encountered any negatives? Are there any times when you think it might be better not to have a strong type system?
To me, the biggest drawback of Haskell isn't anything intrinsic to the language, but the amount of stuff it forces a person to learn. For me, that's a fun challenge... but trying to convince 110 programmers to use a language that forces I/O into a set of types (loosely) called a monad seems like an epic task. Clojure has the advantage of being simple and beautiful once you get past the parentheses. Haskell is demanding and frustrating for the first 6 months (and pays off handsomely later on, but this can make it a hard sell).
Also, how does the Any type in Mu (if anyone familiar with it is here) differ from Data.Dynamic?