Common Mistakes (doh!)
- [x:xs] is not the same as (x:xs)
- [x:xs] :: [[t]] but (x:xs) :: [t]. Somehow it is very easy to write [x:xs] in a pattern when you really mean (x:xs). The compiler is likely to report a (mystifying) type error caused by this mistake at some other location in the program. The slip is probably so easy because the list constructor, ‘:’, puts one in a list frame of mind. It is almost worth doing a search for \[.*:.*\] before compiling a program. It is possible that [x:xs], a list of one element which is itself a non-empty list, is appropriate in some program but not often.
- Numbers
- Haskell has a sophisticated collection of
types - Int, Integer, Float, Double - and classes -
Num, Real, Fractional, Integral, RealFrac, Floating, RealFloat -
describing numbers and related concepts
(H98 section 6.4 p81).
Usually they work as expected but may sometimes trip
one who is familiar with other languages...
The error may be "discovered" when a function, which itself compiles, is used in a new context. The compiler may detect an error that is distant from the true cause.e.g. 1 e.g. 2 e.g. 3 code n = 4
m = ceiling(n/2)n = length [1,2,3,4]
m = ceiling(n/2)n = length [1,2,3,4]
m = ceiling( (fromIntegral n)/2 )result OK, output:
4.0
2Error: No instance for (RealFrac Int) arising from use of ‘ceiling’ at file.hs:5 In the definition of ‘m’: ceiling(n/2)
(- ghc 9/2002)OK, output:
4
2why 4 ~ fromInteger(4) :: Float, hence 4.0 n :: Int
(/) :: (Num a) => a->a->a
but ¬RealFrac IntNB. 4 not 4.0