realizing LLM code is inadequate

src/Djot.hs

@@ -11,12 +11,10 @@ module Djot
 where
 
 import Control.Applicative (many, optional, some, (<|>))
-import Control.Monad.ST (runST)
 import Data.Foldable (for_)
 import Data.Functor (void, (<$>))
 import Data.List (elemIndex)
 import qualified Data.List.NonEmpty as NE
-import Data.STRef (modifySTRef', newSTRef, readSTRef, writeSTRef)
 import Data.Text (Text)
 import qualified Data.Text as T
 import qualified Djot.Lists
@@ -100,105 +98,15 @@ blockQuote attrs = do
 
 listBlock :: (Logger m, Characters s) => Attrs -> Parser s m Element
 listBlock attrs = do
-  list <- Djot.Lists.djotList
-  error ""
+  error "todo"
 
 taskListBlock :: (Logger m, Characters s) => Attrs -> Parser s m Element
 taskListBlock attrs = do
-  error "this is also probably fundamentally wrong"
-  startOffset <- getOffset
-  bullet
-  first_item <- task_item
-  rem_items <- manyTill (listMarker *> task_item) $ lookAhead blockSeparator
-  pure $ TaskList (TL {items = first_item : rem_items}) attrs
-  where
-    task_item = error "todo"
+  error "todo"
 
-lowerAlphabet :: [Char]
-lowerAlphabet = ['a' .. 'z']
-
-upperAlphabet :: [Char]
-upperAlphabet = ['A' .. 'Z']
-
--- inner parsing of roman numerals done via Gemini due to lazy
-romanValue :: Char -> Int
-romanValue 'I' = 1
-romanValue 'V' = 5
-romanValue 'X' = 10
-romanValue 'L' = 50
-romanValue 'C' = 100
-romanValue 'D' = 500
-romanValue 'M' = 1000
-romanValue _ = 0
-
-romanToInt :: Text -> Int
-romanToInt s = runST $ do
-  total <- newSTRef 0
-  prevVal <- newSTRef 0
-
-  for_ (T.unpack s) $ \char -> do
-    let curr = romanValue char
-    p <- readSTRef prevVal
-
-    -- If current > previous (e.g., IV), we subtract the previous twice
-    if curr > p
-      then modifySTRef' total (\t -> t + curr - 2 * p)
-      else modifySTRef' total (+ curr)
-
-    writeSTRef prevVal curr
-
-  readSTRef total
-
-upperRoman :: [Char]
-upperRoman = "IVXLCDM"
-
-lowerRoman :: [Char]
-lowerRoman = T.unpack $ T.toLower $ T.pack upperRoman
-
--- consumes whitespace as well for convenience in element parser
-listMarker :: (Logger m, Characters s) => Parser s m ListType
-listMarker = choice [bullet, decimal, try lower_letter, try upper_letter, try lower_roman_numeral, try upper_roman_numeral] <* space
-  where
-    decimal = do
-      num <- choice $ map ($ some numberChar) [try . surroundParen, try . rightParen, point]
-      let start_number = Just $ read num
-      pure $ Ordered {start_number, style = Just "1"}
-    lower_letter :: (Logger m, Characters s) => Parser s m ListType
-    lower_letter = do
-      letter <- choice $ map ($ lowerChar) [try . surroundParen, try . rightParen, point]
-      let start_number = elemIndex letter lowerAlphabet
-      pure $ Ordered {start_number, style = Just "a"}
-    upper_letter :: (Logger m, Characters s) => Parser s m ListType
-    upper_letter = do
-      letter <- choice $ map ($ upperChar) [try . surroundParen, try . rightParen, point]
-      let start_number = elemIndex letter upperAlphabet
-      pure $ Ordered {start_number, style = Just "A"}
-    lower_roman_numeral = do
-      numeral <- choice $ map ($ some (satisfy (`elem` lowerRoman))) [try . surroundParen, try . rightParen, point]
-      let start_number = Just $ romanToInt $ T.toUpper $ toText numeral
-      pure $ Ordered {start_number, style = Just "i"}
-    upper_roman_numeral = do
-      numeral <- choice $ map ($ some (satisfy (`elem` upperRoman))) [try . surroundParen, try . rightParen, point]
-      let start_number = Just $ romanToInt $ T.toUpper $ toText numeral
-      pure $ Ordered {start_number, style = Just "I"}
-
-    surroundParen :: (Logger m, Characters s) => Parser s m a -> Parser s m a
-    surroundParen parser = do
-      char '('
-      ret <- parser
-      char ')'
-      space
-      pure ret
-    rightParen :: (Logger m, Characters s) => Parser s m a -> Parser s m a
-    rightParen parser = parser <* char ')' <* space
-    point :: (Logger m, Characters s) => Parser s m a -> Parser s m a
-    point parser = parser <* char '.' <* space
-
-bullet :: (Logger m, Characters s) => Parser s m ListType
-bullet = error "todo"
-
-taskMarker :: (Logger m, Characters s) => Parser s m Char
-taskMarker = char '[' *> choice (map char " xX") <* char ']' <* space
+descriptionListBlock :: (Logger m, Characters s) => Attrs -> Parser s m Element
+descriptionListBlock attrs = do
+  error "todo"
 
 codeFence :: (Logger m, Characters s) => Parser s m ()
 codeFence = void $ string "```"

src/Djot/Lists.hs

@@ -2,13 +2,21 @@
 
 -- This is written by Gemini barring minor adjustments made manually to fit
 -- into the Parsing utilities and logger stuff
+--
+-- the Djot root module just does
+-- type translation currently
 
-module Djot.Lists (List, djotList, TaskList, djotTaskList) where
+module Djot.Lists (List (List), djotList, TaskList (TaskList), djotTaskList) where
 
+import Control.Monad.ST (runST)
+import Data.Char (isUpper, ord)
+import Data.Foldable (for_)
 import Data.Functor (void)
+import Data.STRef (modifySTRef', newSTRef, readSTRef, writeSTRef)
 import Data.Text (Text)
 import qualified Data.Text as T
 import Data.Void
+import qualified IR
 import Logger (Logger (logCallStack, logDebug, logError))
 import Text.Megaparsec
 import Text.Megaparsec.Char
@@ -22,13 +30,17 @@ data Delimiter = Period | RightParen | Enclosed -- e.g., '1.', '1)', '(1)'
 -- | The style of the list marker.
 data MarkerStyle
   = Bullet Char -- '*', '-', '+'
-  | Decimal Delimiter -- '1'
-  | LowerAlpha Delimiter -- 'a'
-  | UpperAlpha Delimiter -- 'A'
-  | LowerRoman Delimiter -- 'i'
-  | UpperRoman Delimiter -- 'I'
+  | Decimal Int Delimiter -- '1'
+  | LowerAlpha Int Delimiter -- 'a'
+  | UpperAlpha Int Delimiter -- 'A'
+  | LowerRoman Int Delimiter -- 'i'
+  | UpperRoman Int Delimiter -- 'I'
   deriving (Show, Eq)
 
+intoIRType :: MarkerStyle -> IR.ListType
+intoIRType (Bullet _) = IR.Unordered {style = Nothing}
+intoIRType (Decimal start _) = IR.Ordered {style = Just "1", start_number = Just start}
+
 -- | A single item in a list.
 data ListItem = ListItem
   { itemStyle :: MarkerStyle,
@@ -38,7 +50,7 @@ data ListItem = ListItem
   deriving (Show, Eq)
 
 -- | A full list is a collection of items sharing the exact same MarkerStyle.
-data List = List MarkerStyle [ListItem]
+data List = List MarkerStyle [Djot.Lists.ListItem]
   deriving (Show, Eq)
 
 -- | Parses whitespace but does not consume newlines.
@@ -59,15 +71,21 @@ bulletMarker = Bullet <$> oneOf ['*', '-', '+']
 
 -- | Decimals: '1.', '1)', '(1)'
 decimalMarker :: (Logger m, Characters s) => Parser s m MarkerStyle
-decimalMarker = Decimal . snd <$> orderedDelimiter L.decimal
+decimalMarker = do
+  (val, delim) <- orderedDelimiter L.decimal
+  return $ Decimal val delim
 
 -- | Lower/Upper Alpha: 'a.', 'A)', etc.
 -- (Note: In a full implementation, you must ensure these are single letters)
 lowerAlphaMarker :: (Logger m, Characters s) => Parser s m MarkerStyle
-lowerAlphaMarker = LowerAlpha . snd <$> orderedDelimiter lowerChar
+lowerAlphaMarker = do
+  (c, delim) <- orderedDelimiter lowerChar
+  return $ LowerAlpha (alphaToOffset c) delim
 
 upperAlphaMarker :: (Logger m, Characters s) => Parser s m MarkerStyle
-upperAlphaMarker = UpperAlpha . snd <$> orderedDelimiter upperChar
+upperAlphaMarker = do
+  (c, delim) <- orderedDelimiter upperChar
+  return $ UpperAlpha (alphaToOffset c) delim
 
 -- | Lower/Upper Roman
 -- Uses a simplified roman numeral regex/parser for illustration.
@@ -78,10 +96,16 @@ romanUpper :: (Logger m, Characters s) => Parser s m Text
 romanUpper = toText <$> takeWhile1P (Just "upper roman") (`elem` ['I', 'V', 'X', 'L', 'C', 'D', 'M'])
 
 lowerRomanMarker :: (Logger m, Characters s) => Parser s m MarkerStyle
-lowerRomanMarker = LowerRoman . snd <$> orderedDelimiter romanLower
+lowerRomanMarker = do
+  (numeral, delim) <- orderedDelimiter romanLower
+  let num = romanToInt numeral
+  pure $ LowerRoman num delim
 
 upperRomanMarker :: (Logger m, Characters s) => Parser s m MarkerStyle
-upperRomanMarker = UpperRoman . snd <$> orderedDelimiter romanUpper
+upperRomanMarker = do
+  (numeral, delim) <- orderedDelimiter romanUpper
+  let num = romanToInt numeral
+  pure $ UpperRoman num delim
 
 -- | Attempts to parse any valid Djot list marker.
 listMarker :: (Logger m, Characters s) => Parser s m MarkerStyle
@@ -96,7 +120,7 @@ listMarker =
     ]
 
 -- | Parses a single list item.
-listItem :: (Logger m, Characters s) => Parser s m ListItem
+listItem :: (Logger m, Characters s) => Parser s m Djot.Lists.ListItem
 listItem = do
   -- 1. Grab the current indentation level before the item
   startPos <- L.indentLevel
@@ -219,3 +243,43 @@ taskItemWithStyle expectedStyle = do
   if taskStyle item == expectedStyle
     then pure item
     else fail "Task list marker style mismatch"
+
+-- | Converts a single alpha char to its 1-indexed position (a=1, b=2...)
+alphaToOffset :: Char -> Int
+alphaToOffset c =
+  let base = if isUpper c then ord 'A' else ord 'a'
+   in ord c - base + 1
+
+-- | A real Roman numeral to Int converter (simplified for brevity)
+-- In a production parser, use a library or a more robust fold.
+romanToOffset :: Text -> Int
+romanToOffset = romanToInt -- Implementation of Roman numeral logic
+
+-- inner parsing of roman numerals done via Gemini due to lazy
+romanValue :: Char -> Int
+romanValue 'I' = 1
+romanValue 'V' = 5
+romanValue 'X' = 10
+romanValue 'L' = 50
+romanValue 'C' = 100
+romanValue 'D' = 500
+romanValue 'M' = 1000
+romanValue _ = 0
+
+romanToInt :: Text -> Int
+romanToInt s = runST $ do
+  total <- newSTRef 0
+  prevVal <- newSTRef 0
+
+  for_ (T.unpack s) $ \char -> do
+    let curr = romanValue char
+    p <- readSTRef prevVal
+
+    -- If current > previous (e.g., IV), we subtract the previous twice
+    if curr > p
+      then modifySTRef' total (\t -> t + curr - 2 * p)
+      else modifySTRef' total (+ curr)
+
+    writeSTRef prevVal curr
+
+  readSTRef total