Why write a glob engine at all when you already have a fast regex implementation that can match both exact paths and plausible subtrees? The bulk of the haskel

Why write a glob engine at all when you already have a fast regex implementation that can match both exact paths and plausible subtrees? The bulk of the haskell code to do this: parseGlob :: Char -> Char -> String -> Parser Glob parseGlob escC sepC forbid = many1' (gpart <|> sep <|> glob <|> alt) >>= return . GGroup . V.fromList where gpart = globPart escC (sepC : (forbid ++ "{*")) >>= return . GPart sep = satisfy (== ch2word sepC) >> return GSeparator alt = do _ <- AttoC.char '{' choices <- sepBy' (GEmpty `option` parseGlob escC sepC (",}" ++ forbid)) (char ',') _ <- AttoC.char '}' return $ GAlternate $ V.fromList choices glob = do res <- takeWhile1 (== ch2word '*') if B.length res == 1 then return GSingle else return GDouble wrapParens s = T.concat ["(", s, ")"] globRegex :: Char -> Glob -> T.Text globRegex sep GSingle = T.concat ["([^", T.singleton sep, "]*|\\", T.singleton sep, ")"] globRegex _ GDouble = ".*" globRegex _ GEmpty = "" globRegex sep GSeparator = T.singleton sep globRegex sep (GRepeat a) = T.concat ["(", T.concat (V.toList $ fmap (globRegex sep) a), ")*"] globRegex sep (GGroup a) = T.concat $ V.toList $ fmap (globRegex sep) a globRegex _ (GPart p) = T.concatMap efun base where base = TE.decodeUtf8 p escChars = S.fromList ".[]()\\{}^$*+" efun c = if S.member c escChars then T.concat ["\\", T.singleton c] else T.singleton c globRegex sep (GAlternate a) = if V.null alts then "" else T.concat [altsStr, if hasEmpty then "?" else ""] where hasEmpty = isJust $ V.find (== GEmpty) a alts = fmap (globRegex sep) $ V.filter (/= GEmpty) a altsStr = wrapParens $ T.intercalate "|" $ V.toList alts