Identifier.hs

{-# LANGUAGE OverloadedStrings #-}

-- | Turning column names into valid Haskell identifiers for generated code.
module DataFrame.TH.Identifier (
    sanitize,
    isHaskellIdentifier,
    isReservedId,
    isVarId,
) where

import qualified Data.Char as Char
import qualified Data.Text as T

-- | Haskell Report 2.4 reserved words.
isReservedId :: T.Text -> Bool
isReservedId t = case t of
    "case" -> True
    "class" -> True
    "data" -> True
    "default" -> True
    "deriving" -> True
    "do" -> True
    "else" -> True
    "foreign" -> True
    "if" -> True
    "import" -> True
    "in" -> True
    "infix" -> True
    "infixl" -> True
    "infixr" -> True
    "instance" -> True
    "let" -> True
    "module" -> True
    "newtype" -> True
    "of" -> True
    "then" -> True
    "type" -> True
    "where" -> True
    _ -> False

isVarId :: T.Text -> Bool
isVarId t = case T.uncons t of
    -- '_' is skipped on purpose so "Name" and "_Name_" cannot collide.
    Just (c, _) -> Char.isLower c && Char.isAlpha c
    Nothing -> False

isHaskellIdentifier :: T.Text -> Bool
isHaskellIdentifier t = Prelude.not (isVarId t) || isReservedId t

sanitize :: T.Text -> T.Text
sanitize t
    | isValid = t
    | isHaskellIdentifier t' = "_" <> t' <> "_"
    | otherwise = t'
  where
    isValid =
        Prelude.not (isHaskellIdentifier t)
            && isVarId t
            && T.all Char.isAlphaNum t
    t' = T.map replaceInvalidCharacters . T.filter (Prelude.not . parentheses) $ t
    replaceInvalidCharacters c
        | Char.isUpper c = Char.toLower c
        | Char.isSpace c = '_'
        | Char.isPunctuation c = '_' -- '-' will also become a '_'
        | Char.isSymbol c = '_'
        | Char.isAlphaNum c = c -- Blanket condition
        | otherwise = '_' -- If we're unsure we'll default to an underscore
    parentheses c = case c of
        '(' -> True
        ')' -> True
        '{' -> True
        '}' -> True
        '[' -> True
        ']' -> True
        _ -> False