When there is no language-specific support, you can use tags. This is a generic completion mechanism.
Generate a TAGS file, which contains a list of identifiers and where they are defined. Emacs comes with the etags program to do this in many languages, but not Haskell; ghc comes with hasktags.
Load the TAGS file with M-x visit-tags-table.
Tags are not context-dependent, so they'll indiscriminately suggest types, values, constructors, etc everywhere. They also won't provide advanced features such as easily showing the type of a value. The most important tags commands are:
M-TAB (complete-symbol) completes an identifier according to the loaded list of tags.
M-. (find-tag) goes to the place where the identifier at point is defined, opening the containing file if necessary.
M-* (pop-tag-mark) goes back where you were before M-..
M-x tags-apropos shows a list of identifiers matching a regexp.
For more information, look under "Tags" in the Emacs manual.
For an even cruder, but fully automatic mechanism, there is the dynamic abbrev feature. C-M-/ (dabbrev-completion) looks in most open buffers for a completion; this is completely language-independent, so it'll even find words in strings, comments, whatever. M-/ (dabbrev-expand) is similar, but directly completes to the nearest match before point.