core: Improve integerValue

ucd2haskell/exe/UCD2Haskell/Modules/DerivedNumericValues.hs

@@ -53,6 +53,7 @@ genNumericValuesModule moduleName = Fold step mempty done
         , ""
         , "import Data.Ratio ((%))"
         , ""
+        , "{-# NOINLINE numericValue #-}"
         , "numericValue :: Char -> Maybe Rational"
         , "numericValue = \\case" <> mkEntries values
         , "  _ -> Nothing"

unicode-data/Changelog.md

@@ -1,5 +1,9 @@
 # Changelog
 
+## TBD
+
+- Changed `integerValue` from `Char -> Maybe Int` to `(Integral a) => Char -> Maybe a`.
+
 ## 0.4.0.1 (December 2022)
 
 - Fix [Unicode blocks handling on big-endian architectures](https://github.com/composewell/unicode-data/issues/97).

unicode-data/bench/Unicode/Char/NumericBench.hs

@@ -2,7 +2,8 @@ module Unicode.Char.NumericBench
     ( benchmarks
     ) where
 
-import Test.Tasty.Bench (Benchmark)
+import Data.Int (Int64)
+import Test.Tasty.Bench (Benchmark, bgroup)
 
 import Unicode.Char.Bench (
     Bench (..),
@@ -22,7 +23,15 @@ benchmarks r = bgroupWithCharRange "Unicode.Char.Numeric" r $ \chars ->
   , bgroupWithChars "numericValue" chars
     [ Bench "unicode-data" Num.numericValue
     ]
-  , bgroupWithChars "integerValue" chars
-    [ Bench "unicode-data" Num.integerValue
-    ]
+  , bgroup "integerValue"
+        [ bgroupWithChars "Integer" chars
+            [ Bench "unicode-data" (Num.integerValue :: Char -> Maybe Integer)
+            ]
+        , bgroupWithChars "Int64" chars
+            [ Bench "unicode-data" (Num.integerValue :: Char -> Maybe Int64)
+            ]
+        , bgroupWithChars "Int" chars
+            [ Bench "unicode-data" (Num.integerValue :: Char -> Maybe Int)
+            ]
+        ]
   ]

unicode-data/lib/Unicode/Char/Numeric.hs

@@ -26,11 +26,16 @@ module Unicode.Char.Numeric
     ) where
 
 import Data.Char (digitToInt, intToDigit, isDigit, isHexDigit, isOctDigit)
+import Data.Int (Int64)
 import Data.Maybe (isJust)
-import Data.Ratio (numerator, denominator)
+import Data.Ratio (denominator, numerator)
+
 import qualified Unicode.Char.Numeric.Compat as Compat
 import qualified Unicode.Internal.Char.DerivedNumericValues as V
 
+-- $setup
+-- >>> import Data.Int (Int32, Int64)
+
 -- | Selects Unicode numeric characters, including digits from various
 -- scripts, Roman numerals, et cetera.
 --
@@ -86,16 +91,30 @@ numericValue = V.numericValue
 --
 -- This is a special case of 'numericValue'.
 --
--- __Note:__ a character may have a numeric value but return 'False' with
+-- __Warning:__ There is a risk of /integer overflow/ depending of the chosen
+-- concrete return type. As of Unicode 15.0 the results range from 0 to 1e12.
+--
+-- >>> integerValue '\x5146' :: Maybe Int64 -- OK
+-- Just 1000000000000
+-- >>> integerValue '\x5146' :: Maybe Int32 -- Will overflow!
+-- Just (-727379968)
+--
+-- Therefore it is advised to use: @'integerValue' \@'Int64'@.
+--
+-- __Note:__ A character may have a numeric value but return 'False' with
 -- the predicate 'Unicode.Char.Numeric.Compat.isNumber', because
 -- 'Unicode.Char.Numeric.Compat.isNumber' only tests
 -- 'Unicode.Char.General.GeneralCategory': some CJK characters are
 -- 'Unicode.Char.General.OtherLetter' and do have a numeric value.
 --
 -- @since 0.3.1
-integerValue :: Char -> Maybe Int
+{-# INLINE integerValue #-}
+{-# SPECIALIZE integerValue :: Char -> Maybe Integer #-}
+{-# SPECIALIZE integerValue :: Char -> Maybe Int64   #-}
+{-# SPECIALIZE integerValue :: Char -> Maybe Int     #-}
+integerValue :: (Integral a) => Char -> Maybe a
 integerValue c = do
     r <- V.numericValue c
     if denominator r == 1
-        then Just (fromIntegral (numerator r))
+        then Just (fromInteger (numerator r))
         else Nothing

unicode-data/lib/Unicode/Char/Numeric/Compat.hs

@@ -21,16 +21,16 @@ import qualified Unicode.Internal.Char.UnicodeData.GeneralCategory as UC
 -- scripts, Roman numerals, et cetera.
 --
 -- This function returns 'True' if its argument has one of the
--- following 'GeneralCategory's, or 'False' otherwise:
+-- following 'Unicode.Char.General.GeneralCategory's, or 'False' otherwise:
 --
--- * 'DecimalNumber'
--- * 'LetterNumber'
--- * 'OtherNumber'
+-- * 'Unicode.Char.General.DecimalNumber'
+-- * 'Unicode.Char.General.LetterNumber'
+-- * 'Unicode.Char.General.OtherNumber'
 --
 -- __Note:__ a character may have a numeric value (see
--- 'Unicode.Char.Numeric.numericValue') but return
--- 'False', because 'isNumber' only tests 'GeneralCategory':
--- some CJK characters are 'OtherLetter' and do have a numeric value.
+-- 'Unicode.Char.Numeric.numericValue') but return 'False', because 'isNumber'
+-- only tests 'Unicode.Char.General.GeneralCategory': some CJK characters are
+-- 'Unicode.Char.General.OtherLetter' and do have a numeric value.
 -- Use 'Unicode.Char.Numeric.isNumeric' to cover those cases as well.
 --
 -- prop> isNumber c == Data.Char.isNumber c
@@ -48,4 +48,4 @@ isNumber c =
         UC.LetterNumber  -> True
         UC.OtherNumber   -> True
         _                -> False
-    where cp = ord c
+    where !cp = ord c

unicode-data/lib/Unicode/Internal/Char/DerivedNumericValues.hs

@@ -15,6 +15,7 @@ where
 
 import Data.Ratio ((%))
 
+{-# NOINLINE numericValue #-}
 numericValue :: Char -> Maybe Rational
 numericValue = \case
   '0' -> Just 0