Use cryptohash rather than SHA for hashing.

This is a massive win on OSX, which doesn't have a sha256sum normally.

Only use external hash commands when the file is > 1 mb,
since cryptohash is quite close to them in speed.

SHA is still used to calculate HMACs. I don't quite understand
cryptohash's API for those.

Used the following benchmark to arrive at the 1 mb number.

1 mb file:

benchmarking sha256/internal
mean: 13.86696 ms, lb 13.83010 ms, ub 13.93453 ms, ci 0.950
std dev: 249.3235 us, lb 162.0448 us, ub 458.1744 us, ci 0.950
found 5 outliers among 100 samples (5.0%)
  4 (4.0%) high mild
  1 (1.0%) high severe
variance introduced by outliers: 10.415%
variance is moderately inflated by outliers

benchmarking sha256/external
mean: 14.20670 ms, lb 14.17237 ms, ub 14.27004 ms, ci 0.950
std dev: 230.5448 us, lb 150.7310 us, ub 427.6068 us, ci 0.950
found 3 outliers among 100 samples (3.0%)
  2 (2.0%) high mild
  1 (1.0%) high severe

2 mb file:

benchmarking sha256/internal
mean: 26.44270 ms, lb 26.23701 ms, ub 26.63414 ms, ci 0.950
std dev: 1.012303 ms, lb 925.8921 us, ub 1.122267 ms, ci 0.950
variance introduced by outliers: 35.540%
variance is moderately inflated by outliers

benchmarking sha256/external
mean: 26.84521 ms, lb 26.77644 ms, ub 26.91433 ms, ci 0.950
std dev: 347.7867 us, lb 210.6283 us, ub 571.3351 us, ci 0.950
found 6 outliers among 100 samples (6.0%)

import Crypto.Hash
import Data.ByteString.Lazy as L
import Criterion.Main
import Common

testfile :: FilePath
testfile = "/run/shm/data" -- on ram disk

main = defaultMain
        [ bgroup "sha256"
                [ bench "internal" $ whnfIO internal
                , bench "external" $ whnfIO external
                ]
        ]

sha256 :: L.ByteString -> Digest SHA256
sha256 = hashlazy

internal :: IO String
internal = show . sha256 <$> L.readFile testfile

external :: IO String
external = do
	s <- readProcess "sha256sum" [testfile]
        return $ fst $ separate (== ' ') s

Backend/SHA.hs

@@ -1,6 +1,6 @@
 {- git-annex SHA backends
  -
- - Copyright 2011,2012 Joey Hess <joey@kitenet.net>
+ - Copyright 2011-2013 Joey Hess <joey@kitenet.net>
  -
  - Licensed under the GNU GPL version 3 or higher.
  -}
@@ -12,10 +12,10 @@ import qualified Annex
 import Types.Backend
 import Types.Key
 import Types.KeySource
+import Utility.Hash
 import Utility.ExternalSHA
 
 import qualified Build.SysConfig as SysConfig
-import Data.Digest.Pure.SHA
 import qualified Data.ByteString.Lazy as L
 import Data.Char
 
@@ -70,12 +70,14 @@ shaCommand shasize filesize
 	| shasize == 512 = use SysConfig.sha512 sha512
 	| otherwise = error $ "bad sha size " ++ show shasize
   where
-	use Nothing sha = Left $ showDigest . sha
+	use Nothing hasher = Left $ show . hasher
-	use (Just c) sha
+	use (Just c) hasher
-		{- use builtin, but slower sha for small files
+		{- Use builtin, but slightly slower hashing for
-		 - benchmarking indicates it's faster up to
+		 - smallish files. Cryptohash benchmarks 90 to 101%
-		 - and slightly beyond 50 kb files -}
+		 - faster than external hashers, depending on the hash
-		| filesize < 51200 = use Nothing sha
+		 - and system. So there is no point forking an external
+		 - process unless the file is large. -}
+		| filesize < 1048576 = use Nothing hasher
 		| otherwise = Right c
 
 {- A key is a checksum of its contents. -}

Git/CatFile.hs

@@ -18,7 +18,6 @@ module Git.CatFile (
 import System.IO
 import qualified Data.ByteString as S
 import qualified Data.ByteString.Lazy as L
-import Data.Digest.Pure.SHA
 import Data.Char
 import System.Process (std_out, std_err)
 import Numeric
@@ -31,6 +30,7 @@ import Git.Command
 import Git.Types
 import Git.FilePath
 import qualified Utility.CoProcess as CoProcess
+import Utility.Hash
 
 data CatFileHandle = CatFileHandle CoProcess.CoProcessHandle Repo
 
@@ -103,7 +103,7 @@ catObjectDetails (CatFileHandle hdl repo) object = CoProcess.query hdl send rece
 				}
 		fileEncoding h
 		content <- L.hGetContents h
-		let sha = (\s -> length s `seq` s) (showDigest $ sha1 content)
+		let sha = (\s -> length s `seq` s) (show $ sha1 content)
 		ok <- checkSuccessProcess pid
 		return $ if ok
 			then Just (content, Ref sha)

Remote/Bup.hs

@@ -10,6 +10,7 @@ module Remote.Bup (remote) where
 import qualified Data.ByteString.Lazy as L
 import qualified Data.Map as M
 import System.Process
+import Data.ByteString.Lazy.UTF8 (fromString)
 
 import Common.Annex
 import Types.Remote
@@ -25,8 +26,7 @@ import Remote.Helper.Ssh
 import Remote.Helper.Special
 import Remote.Helper.Encryptable
 import Crypto
-import Data.ByteString.Lazy.UTF8 (fromString)
+import Utility.Hash
-import Data.Digest.Pure.SHA
 import Utility.UserInfo
 import Annex.Content
 import Annex.UUID
@@ -277,7 +277,7 @@ bup2GitRemote r
 bupRef :: Key -> String
 bupRef k
 	| Git.Ref.legal True shown = shown
-	| otherwise = "git-annex-" ++ showDigest (sha256 (fromString shown))
+	| otherwise = "git-annex-" ++ show (sha256 (fromString shown))
   where
 	shown = key2file k
 

Utility/ExternalSHA.hs

@@ -1,6 +1,7 @@
 {- Calculating a SHA checksum with an external command.
  -
- - This is often faster than using Haskell libraries.
+ - This is typically a bit faster than using Haskell libraries,
+ - by around 1% to 10%. Worth it for really big files.
  -
  - Copyright 2011-2013 Joey Hess <joey@kitenet.net>
  -

Utility/Hash.hs

@@ -0,0 +1,29 @@
+{- Convenience wrapper around cryptohash.
+ -
+ - The resulting Digests can be shown to get a canonical hash encoding. -}
+
+module Utility.Hash where
+
+import Crypto.Hash
+import qualified Data.ByteString.Lazy as L
+
+sha1 :: L.ByteString -> Digest SHA1
+sha1 = hashlazy
+
+sha224 :: L.ByteString -> Digest SHA224
+sha224 = hashlazy
+
+sha256 :: L.ByteString -> Digest SHA256
+sha256 = hashlazy
+
+sha384 :: L.ByteString -> Digest SHA384
+sha384 = hashlazy
+
+sha512 :: L.ByteString -> Digest SHA512
+sha512 = hashlazy
+
+-- sha3 is not yet fully standardized
+--sha3 :: L.ByteString -> Digest SHA3
+--sha3 = hashlazy
+
+

Utility/WebApp.hs

@@ -12,6 +12,7 @@ module Utility.WebApp where
 import Common
 import Utility.Tmp
 import Utility.FileMode
+import Utility.Hash
 
 import qualified Yesod
 import qualified Network.Wai as Wai
@@ -24,7 +25,6 @@ import qualified Data.CaseInsensitive as CI
 import Network.Socket
 import Control.Exception
 import Crypto.Random
-import Data.Digest.Pure.SHA
 import qualified Web.ClientSession as CS
 import qualified Data.ByteString.Lazy as L
 import qualified Data.ByteString.Lazy.UTF8 as L8
@@ -214,7 +214,7 @@ genRandomToken = do
 	return $
 		case genBytes 512 g of
 			Left e -> error $ "failed to generate secret token: " ++ show e
-			Right (s, _) -> showDigest $ sha512 $ L.fromChunks [s]
+			Right (s, _) -> show $ sha512 $ L.fromChunks [s]
 
 {- A Yesod isAuthorized method, which checks the auth cgi parameter
  - against a token extracted from the Yesod application.

debian/changelog

@@ -1,3 +1,11 @@
+git-annex (4.20130921) UNRELEASED; urgency=low
+
+  * Use cryptohash rather than SHA for hashing when no external hash program
+    is available. This is a significant speedup for SHA256 on OSX, for
+    example.
+
+ -- Joey Hess <joeyh@debian.org>  Sun, 22 Sep 2013 19:42:29 -0400
+
 git-annex (4.20130920) unstable; urgency=low
 
   * webapp: Initial support for setting up encrypted removable drives.

git-annex.cabal

@@ -76,7 +76,7 @@ Executable git-annex
   Build-Depends: MissingH, hslogger, directory, filepath,
    containers, utf8-string, network (>= 2.0), mtl (>= 2),
    bytestring, old-locale, time, HTTP,
-   extensible-exceptions, dataenc, SHA, process, json,
+   extensible-exceptions, dataenc, SHA, cryptohash, process, json,
    base (>= 4.5 && < 4.8), monad-control, MonadCatchIO-transformers,
    IfElse, text, QuickCheck >= 2.1, bloomfilter, edit-distance, process,
    SafeSemaphore, uuid, random, dlist, unix-compat