This example consists of two parts:
- Scanner.hs: generated with Alex from Scanner.x. Put your own regular
expressions for your scanner there, and see the "haskell blocks" to the
right of each regular expression for which functions to call to
construct the right tokens. The rest is reusable junk code.

- Example.hs: contains a parser for some basic expression language. I
only used BNF-parsers (sequential <*> and alternative <$>) here. There
are abstractions for EBNF-parsers, and actually, there is a huge amount
of even more abstractions, but for those I always have to look inside
the source files what their names are (bad bad bad documentation).

The idea is that if you have a production:

N -> S1 S2 S3 S4

That you write it down as:

pN :: TokenParser type
pN = function <$> pS1 <*> pS2 <*> pS3 <*> pS4

where the function gets values from parsing S1..S4 as parameters and
returns a value of the type you gave as parameter to TokenParser:

pN = (\v1 v2 v3 v4 -> ...) <$> pS1 <*> pS2 <*> pS3 <*> pS4

if you have more than one production for a nonterminal, then you can write:

pN =   parser-alternative-1
   <|> parser-alternative-2
   <|> ...

Your parsers may be recursive, but they are not allowed to be left
recursive. A parser is left recursive when you call itself again without
having parsed a terminal first. For example, this one is recursive, but
only left-recursive if p can parse the empty string (but then it would
also be ambiguous because then you can't differentiate between p and pSucceed
[]):

pMany :: Parser a -> Parser [a]
pMany p
  =    (\x xs -> x : xs) <$> p <*> pMany p
  <|>  pSucceed []

In this case, pSucceed is a special parser which always succeeds and
yields it's parameter as semantic result. Similarly, you have pFail
which always fails (p <|> pFail === p).

Finally, you'll need parsers that can actually do some parsing, i.e.
parsers for terminals. Since your terminals in this case are tokens, you
basically need one for each token type:

For those tokens constructed with function "reserved":
pKey :: String -> TokenParser String

For those tokens constructed with function "valueToken TkVarid":
pVarid :: TokenParser String

and there are some others. If you want also the position of where these
tokens were parsed, then you can use:

pKeyPos :: String -> TokenParser Pos
pVaridPos :: TokenParser (String, Pos)

where data Pos = Pos Int Int String