add inRepoDetails

planned to use for an optimisation

most things using stagedDetails were not expecting to get dup files in a
conflicted merge and deal with them, so converted them to use
inRepoDetails.

Annex/View.hs

@@ -345,11 +345,11 @@ narrowView = applyView' viewedFileReuse getViewedFileMetaData
 applyView' :: MkViewedFile -> (FilePath -> MetaData) -> View -> Annex Git.Branch
 applyView' mkviewedfile getfilemetadata view = do
 	top <- fromRepo Git.repoPath
-	(l, clean) <- inRepo $ Git.LsFiles.stagedDetails [top]
+	(l, clean) <- inRepo $ Git.LsFiles.inRepoDetails [] [top]
 	liftIO . nukeFile =<< fromRepo gitAnnexViewIndex
 	viewg <- withViewIndex gitRepo
 	withUpdateIndex viewg $ \uh -> do
-		forM_ l $ \(f, sha, mode, _) -> do
+		forM_ l $ \(f, sha, mode) -> do
 			topf <- inRepo (toTopFilePath f)
 			go uh topf sha (toTreeItemType mode) =<< lookupFile f
 		liftIO $ void clean

Git/LsFiles.hs

@@ -8,6 +8,7 @@
 module Git.LsFiles (
 	Options(..),
 	inRepo,
+	inRepoDetails,
 	inRepoOrBranch,
 	notInRepo,
 	notInRepoIncludingEmptyDirectories,
@@ -67,6 +68,9 @@ guardSafeForLsFiles r a
 
 data Options = ErrorUnmatch
 
+opParam :: Options -> CommandParam
+opParam ErrorUnmatch = Param "--error-unmatch"
+
 {- Lists files that are checked into git's index at the specified paths.
  - With no paths, all files are listed.
  -}
@@ -79,9 +83,18 @@ inRepo' ps os l repo = guardSafeForLsFiles repo $ pipeNullSplit' params repo
 	params = 
 		Param "ls-files" :
 		Param "-z" :
-		map op os ++ ps ++
+		map opParam os ++ ps ++
 		(Param "--" : map (File . fromRawFilePath) l)
-	op ErrorUnmatch = Param "--error-unmatch" 
+
+{- Lists the same files inRepo does, but with sha and mode. -}
+inRepoDetails :: [Options] -> [RawFilePath] -> Repo -> IO ([(RawFilePath, Sha, FileMode)], IO Bool)
+inRepoDetails = stagedDetails' parser . map opParam
+  where
+	parser s = case parseStagedDetails s of
+		Just (file, sha, mode, stagenum)
+			| stagenum == usualStageNum || stagenum == mergeConflictHeadStageNum ->
+				Just (file, sha, mode)
+		_ -> Nothing
 
 {- Files that are checked into the index or have been committed to a
  - branch. -}
@@ -158,12 +171,12 @@ mergeConflictHeadStageNum = 2
  - more than once with different stage numbers.
  -}
 stagedDetails :: [RawFilePath] -> Repo -> IO ([StagedDetails], IO Bool)
-stagedDetails = stagedDetails' []
+stagedDetails = stagedDetails' parseStagedDetails []
 
-stagedDetails' :: [CommandParam] -> [RawFilePath] -> Repo -> IO ([StagedDetails], IO Bool)
-stagedDetails' ps l repo = guardSafeForLsFiles repo $ do
+stagedDetails' :: (S.ByteString -> Maybe t) -> [CommandParam] -> [RawFilePath] -> Repo -> IO ([t], IO Bool)
+stagedDetails' parser ps l repo = guardSafeForLsFiles repo $ do
 	(ls, cleanup) <- pipeNullSplit' params repo
-	return (mapMaybe parseStagedDetails ls, cleanup)
+	return (mapMaybe parser ls, cleanup)
   where
 	params = Param "ls-files" : Param "--stage" : Param "-z" : ps ++ 
 		Param "--" : map (File . fromRawFilePath) l

Logs/Web.hs

@@ -94,12 +94,12 @@ knownUrls = do
 	Annex.Branch.commit =<< Annex.Branch.commitMessage
 	Annex.Branch.withIndex $ do
 		top <- fromRepo Git.repoPath
-		(l, cleanup) <- inRepo $ Git.LsFiles.stagedDetails [top]
+		(l, cleanup) <- inRepo $ Git.LsFiles.inRepoDetails [] [top]
 		r <- mapM getkeyurls l
 		void $ liftIO cleanup
 		return $ concat r
   where
-	getkeyurls (f, s, _, _) = case urlLogFileKey f of
+	getkeyurls (f, s, _) = case urlLogFileKey f of
 		Just k -> zip (repeat k) <$> geturls s
 		Nothing -> return []
 	geturls logsha =

Upgrade/V5.hs

@@ -110,11 +110,11 @@ convertDirect = do
 upgradeDirectWorkTree :: Annex ()
 upgradeDirectWorkTree = do
 	top <- fromRepo Git.repoPath
-	(l, clean) <- inRepo $ Git.LsFiles.stagedDetails [top]
+	(l, clean) <- inRepo $ Git.LsFiles.inRepoDetails [] [top]
 	forM_ l go
 	void $ liftIO clean
   where
-	go (f, _sha, mode, _stagenum) | isSymLink mode = do
+	go (f, _sha, mode) | isSymLink mode = do
 		-- Cannot use lookupFile here, as we're in between direct
 		-- mode and v6.
 		mk <- catKeyFile f

doc/todo/precache_logs_for_speed_with_cat-file_--buffer.mdwn

@@ -13,10 +13,7 @@ Probably that extra round trip means the performance improvement will not
 be as good as --all's was, but it could still be significant.
 
 > Actually, the key lookup could use the same --buffer trick! 
-> Although this would need a way to use git ls-files to get the sha of each
-> file, and the only way I can find is --stage, which lists the file
-> repeatedly when there's a merge conflict. If that can be finessed somehow,
-> pass the file sha through cat-file to get key, and then pass the location
-> log for the key through cat-file to precache logs.
+> Use inRepoDetails to list files and shas, pass through cat-file to get keys,
+> and then pass the location log for each key through cat-file to precache logs.
 
 --[[Joey]]