git-annex/Command/Unused.hs

304 lines
9.4 KiB
Haskell
Raw Normal View History

{- git-annex command
-
- Copyright 2010-2012 Joey Hess <joey@kitenet.net>
-
- Licensed under the GNU GPL version 3 or higher.
-}
{-# LANGUAGE BangPatterns #-}
module Command.Unused where
import qualified Data.Set as S
import qualified Data.Text.Lazy as L
import qualified Data.Text.Lazy.Encoding as L
import Data.BloomFilter
import Data.BloomFilter.Easy
import Data.BloomFilter.Hash
import Control.Monad.ST
2011-10-05 20:02:51 +00:00
import Common.Annex
import Command
import Logs.Unused
2011-10-04 04:40:47 +00:00
import Annex.Content
import Utility.FileMode
2011-10-15 20:21:08 +00:00
import Logs.Location
2012-03-22 04:23:15 +00:00
import Config
2010-11-15 22:04:19 +00:00
import qualified Annex
import qualified Git
2011-12-14 19:56:11 +00:00
import qualified Git.Command
2011-12-12 22:23:24 +00:00
import qualified Git.Ref
import qualified Git.LsFiles as LsFiles
import qualified Git.LsTree as LsTree
import qualified Backend
2011-04-03 00:59:41 +00:00
import qualified Remote
2011-10-04 04:40:47 +00:00
import qualified Annex.Branch
2012-01-06 14:14:37 +00:00
import qualified Option
2011-10-04 04:40:47 +00:00
import Annex.CatFile
def :: [Command]
def = [withOptions [fromOption] $ command "unused" paramNothing seek
"look for unused file content"]
fromOption :: Option
2012-01-06 14:14:37 +00:00
fromOption = Option.field ['f'] "from" paramRemote "remote to check for unused content"
seek :: [CommandSeek]
2012-02-16 04:41:30 +00:00
seek = [withNothing start]
{- Finds unused content in the annex. -}
start :: CommandStart
start = do
2012-01-06 14:14:37 +00:00
from <- Annex.getField $ Option.name fromOption
let (name, action) = case from of
Nothing -> (".", checkUnused)
Just "." -> (".", checkUnused)
Just "here" -> (".", checkUnused)
Just n -> (n, checkRemoteUnused n)
showStart "unused" name
next action
checkUnused :: CommandPerform
2012-03-12 01:08:48 +00:00
checkUnused = chain 0
[ check "" unusedMsg $ findunused =<< Annex.getState Annex.fast
, check "bad" staleBadMsg $ staleKeysPrune gitAnnexBadDir
, check "tmp" staleTmpMsg $ staleKeysPrune gitAnnexTmpDir
]
where
findunused True = do
showNote "fast mode enabled; only finding stale files"
return []
findunused False = do
showAction "checking for unused data"
2012-03-12 19:21:20 +00:00
excludeReferenced =<< getKeysPresent
2012-03-12 01:08:48 +00:00
chain _ [] = next $ return True
chain v (a:as) = do
v' <- a v
chain v' as
checkRemoteUnused :: String -> CommandPerform
2012-03-12 01:08:48 +00:00
checkRemoteUnused name = go =<< fromJust <$> Remote.byName (Just name)
where
go r = do
showAction "checking for unused data"
_ <- check "" (remoteUnusedMsg r) (remoteunused r) 0
next $ return True
remoteunused r =
excludeReferenced =<< loggedKeysFor (Remote.uuid r)
check :: FilePath -> ([(Int, Key)] -> String) -> Annex [Key] -> Int -> Annex Int
check file msg a c = do
l <- a
let unusedlist = number c l
unless (null l) $ showLongNote $ msg unusedlist
writeUnusedLog file unusedlist
2012-03-12 01:08:48 +00:00
return $ c + length l
2011-05-15 06:49:43 +00:00
2012-03-12 01:08:48 +00:00
number :: Int -> [a] -> [(Int, a)]
number _ [] = []
number n (x:xs) = (n+1, x) : number (n+1) xs
2011-04-03 00:59:41 +00:00
table :: [(Int, Key)] -> [String]
2011-07-15 16:47:14 +00:00
table l = " NUMBER KEY" : map cols l
where
cols (n,k) = " " ++ pad 6 (show n) ++ " " ++ show k
2010-11-15 22:04:19 +00:00
pad n s = s ++ replicate (n - length s) ' '
2011-04-03 00:59:41 +00:00
staleTmpMsg :: [(Int, Key)] -> String
staleTmpMsg t = unlines $
["Some partially transferred data exists in temporary files:"]
++ table t ++ [dropMsg Nothing]
staleBadMsg :: [(Int, Key)] -> String
staleBadMsg t = unlines $
["Some corrupted files have been preserved by fsck, just in case:"]
++ table t ++ [dropMsg Nothing]
2011-04-03 00:59:41 +00:00
unusedMsg :: [(Int, Key)] -> String
unusedMsg u = unusedMsg' u
["Some annexed data is no longer used by any files:"]
[dropMsg Nothing]
2011-04-03 00:59:41 +00:00
unusedMsg' :: [(Int, Key)] -> [String] -> [String] -> String
unusedMsg' u header trailer = unlines $
header ++
table u ++
["(To see where data was previously used, try: git log --stat -S'KEY')"] ++
trailer
2011-12-31 08:11:39 +00:00
remoteUnusedMsg :: Remote -> [(Int, Key)] -> String
remoteUnusedMsg r u = unusedMsg' u
["Some annexed data on " ++ name ++ " is not used by any files:"]
[dropMsg $ Just r]
where
name = Remote.name r
2011-12-31 08:11:39 +00:00
dropMsg :: Maybe Remote -> String
2011-04-03 00:59:41 +00:00
dropMsg Nothing = dropMsg' ""
dropMsg (Just r) = dropMsg' $ " --from " ++ Remote.name r
dropMsg' :: String -> String
dropMsg' s = "\nTo remove unwanted data: git-annex dropunused" ++ s ++ " NUMBER\n"
2011-04-03 00:59:41 +00:00
2012-03-12 19:21:20 +00:00
{- Finds keys in the list that are not referenced in the git repository.
-
- Strategy:
-
- * Build a bloom filter of all keys referenced by symlinks. This
- is the fastest one to build and will filter out most keys.
- * If keys remain, build a second bloom filter of keys referenced by
- all branches.
- * The list is streamed through these bloom filters lazily, so both will
- exist at the same time. This means that twice the memory is used,
- but they're relatively small, so the added complexity of using a
- mutable bloom filter does not seem worthwhile.
- * Generating the second bloom filter can take quite a while, since
- it needs enumerating all keys in all git branches. But, the common
- case, if the second filter is needed, is for some keys to be globally
- unused, and in that case, no short-circuit is possible.
- Short-circuiting if the first filter filters all the keys handles the
- other common case.
-}
excludeReferenced :: [Key] -> Annex [Key]
2012-03-12 19:21:20 +00:00
excludeReferenced ks = runfilter firstlevel ks >>= runfilter secondlevel
where
2012-03-12 19:21:20 +00:00
runfilter _ [] = return [] -- optimisation
runfilter a l = bloomFilter show l <$> genBloomFilter show a
firstlevel = withKeysReferencedM
secondlevel = withKeysReferencedInGit
2011-04-03 00:59:41 +00:00
{- Finds items in the first, smaller list, that are not
- present in the second, larger list.
-
- Constructing a single set, of the list that tends to be
- smaller, appears more efficient in both memory and CPU
- than constructing and taking the S.difference of two sets. -}
exclude :: Ord a => [a] -> [a] -> [a]
exclude [] _ = [] -- optimisation
exclude smaller larger = S.toList $ remove larger $ S.fromList smaller
where
remove a b = foldl (flip S.delete) b a
{- A bloom filter capable of holding half a million keys with a
- false positive rate of 1 in 1000 uses around 8 mb of memory,
- so will easily fit on even my lowest memory systems.
-}
bloomCapacity :: Annex Int
bloomCapacity = fromMaybe 500000 . readish
<$> getConfig (annexConfig "bloomcapacity") ""
bloomAccuracy :: Annex Int
bloomAccuracy = fromMaybe 1000 . readish
<$> getConfig (annexConfig "bloomaccuracy") ""
bloomBitsHashes :: Annex (Int, Int)
bloomBitsHashes = do
capacity <- bloomCapacity
accuracy <- bloomAccuracy
return $ suggestSizing capacity (1/ fromIntegral accuracy)
2012-03-12 19:21:20 +00:00
{- Creates a bloom filter, and runs an action, such as withKeysReferenced,
- to populate it.
-
- The action is passed a callback that it can use to feed values into the
- bloom filter.
-
- Once the action completes, the mutable filter is frozen
- for later use.
-}
2012-03-12 19:21:20 +00:00
genBloomFilter :: Hashable t => (v -> t) -> ((v -> Annex ()) -> Annex b) -> Annex (Bloom t)
genBloomFilter convert populate = do
(numbits, numhashes) <- bloomBitsHashes
bloom <- lift $ newMB (cheapHashes numhashes) numbits
2012-03-12 19:21:20 +00:00
_ <- populate $ \v -> lift $ insertMB bloom (convert v)
lift $ unsafeFreezeMB bloom
where
lift = liftIO . stToIO
2012-03-12 19:21:20 +00:00
bloomFilter :: Hashable t => (v -> t) -> [v] -> Bloom t -> [v]
bloomFilter convert l bloom = filter (\k -> convert k `notElemB` bloom) l
2012-03-12 19:21:20 +00:00
{- Given an initial value, folds it with each key referenced by
- symlinks in the git repo. -}
withKeysReferenced :: v -> (Key -> v -> v) -> Annex v
2012-03-12 19:21:20 +00:00
withKeysReferenced initial a = withKeysReferenced' initial folda
where
folda k v = return $ a k v
{- Runs an action on each referenced key in the git repo. -}
withKeysReferencedM :: (Key -> Annex ()) -> Annex ()
withKeysReferencedM a = withKeysReferenced' () calla
where
2012-03-12 19:21:20 +00:00
calla k _ = a k
withKeysReferenced' :: v -> (Key -> v -> Annex v) -> Annex v
withKeysReferenced' initial a = go initial =<< files
where
files = do
Clean up handling of git directory and git worktree. Baked into the code was an assumption that a repository's git directory could be determined by adding ".git" to its work tree (or nothing for bare repos). That fails when core.worktree, or GIT_DIR and GIT_WORK_TREE are used to separate the two. This was attacked at the type level, by storing the gitdir and worktree separately, so Nothing for the worktree means a bare repo. A complication arose because we don't learn where a repository is bare until its configuration is read. So another Location type handles repositories that have not had their config read yet. I am not entirely happy with this being a Location type, rather than representing them entirely separate from the Git type. The new code is not worse than the old, but better types could enforce more safety. Added support for core.worktree. Overriding it with -c isn't supported because it's not really clear what to do if a git repo's config is read, is not bare, and is then overridden to bare. What is the right git directory in this case? I will worry about this if/when someone has a use case for overriding core.worktree with -c. (See Git.Config.updateLocation) Also removed and renamed some functions like gitDir and workTree that misused git's terminology. One minor regression is known: git annex add in a bare repository does not print a nice error message, but runs git ls-files in a way that fails earlier with a less nice error message. This is because before --work-tree was always passed to git commands, even in a bare repo, while now it's not.
2012-05-18 20:38:26 +00:00
top <- fromRepo Git.repoPath
inRepo $ LsFiles.inRepo [top]
go v [] = return v
go v (f:fs) = do
x <- Backend.lookupFile f
case x of
Nothing -> go v fs
Just (k, _) -> do
!v' <- a k v
go v' fs
2012-03-12 19:21:20 +00:00
withKeysReferencedInGit :: (Key -> Annex ()) -> Annex ()
withKeysReferencedInGit a = do
rs <- relevantrefs <$> showref
forM_ rs (withKeysReferencedInGitRef a)
where
showref = inRepo $ Git.Command.pipeRead [Param "show-ref"]
relevantrefs = map (Git.Ref . snd) .
nubBy uniqref .
filter ourbranches .
map (separate (== ' ')) . lines
uniqref (x, _) (y, _) = x == y
ourbranchend = '/' : show Annex.Branch.name
ourbranches (_, b) = not $ ourbranchend `isSuffixOf` b
withKeysReferencedInGitRef :: (Key -> Annex ()) -> Git.Ref -> Annex ()
withKeysReferencedInGitRef a ref = do
2011-12-12 22:23:24 +00:00
showAction $ "checking " ++ Git.Ref.describe ref
2012-03-12 19:21:20 +00:00
go =<< inRepo (LsTree.lsTree ref)
where
2012-04-22 03:32:33 +00:00
go [] = noop
2012-03-12 19:21:20 +00:00
go (l:ls)
2011-09-29 00:12:11 +00:00
| isSymLink (LsTree.mode l) = do
content <- L.decodeUtf8 <$> catFile ref (LsTree.file l)
case fileKey (takeFileName $ L.unpack content) of
2012-03-12 19:21:20 +00:00
Nothing -> go ls
Just k -> do
2012-03-12 19:21:20 +00:00
a k
go ls
| otherwise = go ls
{- Looks in the specified directory for bad/tmp keys, and returns a list
- of those that might still have value, or might be stale and removable.
-
- Also, stale keys that can be proven to have no value are deleted.
-}
staleKeysPrune :: (Git.Repo -> FilePath) -> Annex [Key]
staleKeysPrune dirspec = do
contents <- staleKeys dirspec
2011-09-23 22:13:24 +00:00
dups <- filterM inAnnex contents
let stale = contents `exclude` dups
dir <- fromRepo dirspec
liftIO $ forM_ dups $ \t -> removeFile $ dir </> keyFile t
return stale
staleKeys :: (Git.Repo -> FilePath) -> Annex [Key]
staleKeys dirspec = do
dir <- fromRepo dirspec
ifM (liftIO $ doesDirectoryExist dir)
( do
contents <- liftIO $ getDirectoryContents dir
files <- liftIO $ filterM doesFileExist $
map (dir </>) contents
2011-07-15 16:47:14 +00:00
return $ mapMaybe (fileKey . takeFileName) files
, return []
)