haskell / haskell/binary

Alternative instance for Get does not respect identity law in error situations

Open
#203 0 comments 1 reaction 0 assignees View on GitHub
Dominant language
Haskell
Stars
120
Forks
70
PR merge metrics
No merged PRs in 30d

Description

Hello,

While trying to combine `Get` parsers using `asum`, I found out that `instance Alternative Get` is not lawful: `x <|> empty` is not the same as `x` with respect to failure. `x <|> empty` errors with `"Data.Binary.Get(Alternative).empty"`, overriding the error message from `x`.

For me, an unfortunate consequence is that the implementation of `asum` leaks through: `asum [x, y] = x <|> y <|> empty` provides less helpful parse errors than `x <|> y`.

I have attached a minimal working example, also hosted on this Gist: https://gist.github.com/sugarbleat/4f30751feedf8d3e06911deae7ef4a5a, which can be run with `cabal run Main.hs`. For reference, a similar test with Parsec shows no violation of identity.

Thanks for the great work on the library, by the way!

```haskell
#!/usr/bin/env cabal
{- cabal:
build-depends: base ^>= 4.15
, bytestring ^>= 0.10.12
, binary == 0.8.9.1
-}
module Main (main) where

import Control.Applicative (Alternative (..))
import Data.Binary (Get, Word8, get)
import Data.Binary.Get (runGetOrFail)
import qualified Data.ByteString.Lazy as B
import Data.Foldable (asum)

testGet :: Show a => Get a -> B.ByteString -> IO ()
testGet p s = do
putStrLn "p"
print $ runGetOrFail p s

putStrLn "p <|> empty"
print $ runGetOrFail (p <|> empty) s

putStrLn "asum [p]"
print $ runGetOrFail (asum [p]) s

main :: IO ()
main = testGet (get :: Get Word8) B.empty

{-
Output:
p
Left ("",0,"not enough bytes")
p <|> empty
Left ("",0,"Data.Binary.Get(Alternative).empty")
asum [p]
Left ("",0,"Data.Binary.Get(Alternative).empty")
-}
```

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.