I am following this guide on writing a scheme interpreter. In trying to left-factor the grammar for DottedList/List, I came up with this:
E -> (H T)
H -> E H'
H' -> <space> H
H' -> <term>
T -> <term>
T -> <space> . <space> E
--
spaces :: Parser ()
spaces = skipMany1 (space <?> "spaces")
parseExpr :: Parser LispVal
parseExpr = (... omitted ...) <|>
do char '('
h <- sepBy parseExpr spaces
t <- optionMaybe ((spaces' >> char '.' >> spaces' >> parseExpr) <?> "parseDotExpr failed")
z <- if isJust t then return $ DottedSuffix $ fromJust t else return Tail
z' <- case z of Tail -> return $ List x
DottedSuffix s -> return $ DottedList x s
char ')'
return z'
Unfortunately this doesn't handle the basic dottedlists:
test/Spec.hs:23:
1) test eval 1 evals DottedList
expected: "(1 2 . 1)"
but got: "Parse error at \"lisp\" (line 1, column 7):\nunexpected \".\"\nexpecting spaces' or parseExpr!"
test/Spec.hs:26:
2) test eval 1 evals DottedList (quoted)
expected: "((1 2) . 1)"
but got: "Parse error at \"lisp\" (line 1, column 15):\nunexpected \".\"\nexpecting spaces' or parseExpr!"
test/Spec.hs:29:
3) test eval 1 evals DottedList (sugared)
expected: "((1 2) . 1)"
but got: "Parse error at \"lisp\" (line 1, column 9):\nunexpected \".\"\nexpecting spaces' or parseExpr!"
Update: From @pat's response, I got my tests to pass using:
parseExpr :: Parser LispVal
parseExpr = {- omitted -}
<|> do char '('
x <- many1 (do e <- parseExpr; spaces'; return e)
{- omitted -}
The sepBy
parser is seeing the space before the dot, and committing to parse another expression, which fails.
You should have lexemes consume and discard trailing spaces (see parsec's lexeme
) and change the sepBy
to just many1
. The optionMaybe
can then commit after seeing a dot, which would otherwise have required a try
.