
Ryan Trinkle
Why not just use a datastructure with 6 Bools? E.g.:
data SixDots = SixDots { dot1 :: Bool , dot2 :: Bool , dot3 :: Bool , dot4 :: Bool , dot5 :: Bool , dot6 :: Bool }
Because it is convenient to have an Enum instance, and because braille dot patterns are used as sort of constants in the code. So if I had a data structure like you described above, I'd still need 64 functions with convenient names that return such a data structure with the appropriate bits set. -- eeek! dot123 :: SixDots' dot123 = SixDots' True True True False False False It seems *a lot* easier to just use a sum type and enumerate all the possibilities. Besides, it seems overly inefficient to use 6 Bools for something that actually just needs 6 bits. Also, the resulting enumeration actually matches the Unicode standard. To illustrate, I have a function to convert Braille dot patterns to Char: toChar :: SixDots -> Char toChar = toEnum . (+ 0x2800) . fromEnum That said, I am not really looking for advice on the example at hand, I am more interested in using that example as a motivation to learn TH. [...]
On Wed, Sep 28, 2016 at 9:06 AM, Mario Lang
wrote: Hi.
As a long-term Lisp fan, and someone who always admired compile-time code-generation, I wanted to at least do something simple with Template Haskell once.
In a small project of mine, I have this basically auto-generated data type:
-- Braille music code only uses the old 6-dot system. We enumerate all -- possible dot patterns to use the type system to avoid accidentally -- specifying invalid dot patterns in the source code. -- -- gen :: String -- gen = -- "data Braille = " ++ intercalate " | " ctors ++ " deriving (Enum, Eq)" where -- ctors = "NoDots" : map ctorName [1..63] where -- ctorName :: Int -> String -- ctorName = (++) "Dot" . concatMap (show . succ) . flip filter [0..5] . testBit
data SixDots = NoDots | Dot1 | Dot2 | Dot12 | Dot3 | Dot13 | Dot23 | Dot123 | Dot4 | Dot14 | Dot24 | Dot124 | Dot34 | Dot134 | Dot234 | Dot1234 | Dot5 | Dot15 | Dot25 | Dot125 | Dot35 | Dot135 | Dot235 | Dot1235 | Dot45 | Dot145 | Dot245 | Dot1245 | Dot345 | Dot1345 | Dot2345 | Dot12345 | Dot6 | Dot16 | Dot26 | Dot126 | Dot36 | Dot136 | Dot236 | Dot1236 | Dot46 | Dot146 | Dot246 | Dot1246 | Dot346 | Dot1346 | Dot2346 | Dot12346 | Dot56 | Dot156 | Dot256 | Dot1256 | Dot356 | Dot1356 | Dot2356 | Dot12356 | Dot456 | Dot1456 | Dot2456 | Dot12456 | Dot3456 | Dot13456 | Dot23456 | Dot123456 deriving (Bounded, Enum, Eq, Read, Show)
So, while actually quite simple, this looks like an opportunity to use Template Haskell for me. In other words, I want to figure out what is necessary to generate this data type with TH, instead of the gen function that basically generates a piece of plain Haskell code.
I have been reading "A practical Template Haskell Tutorial"[1] but I find it a little bit too terse to actually solve this very little riddle on my own.
For one, I find it confusing that some TH functions return "Q Dec" while others just return Dec. I am aware that this is some sort of Monad for the TH system, but I have never seen it explained anywhere.
Also, all the examples I can find seem to be mostly focused in generating Q Exp or similar, but I didn't really find an example for Q Dec.
I realize this should be simple to figure out on my own, but it apparently is not. I have tried to wrap my head around this on my own at least three times now, but always stopped after an hour or two due to frustration. Is there some comprehensive TH documentation I haven't seen yet? Could you please give me enough of a head-start that I actually manage to write something which can generate this simple data type above?
[1] https://wiki.haskell.org/A_practical_Template_Haskell_Tutorial
-- CYa, ⡍⠁⠗⠊⠕