git-annex/Annex.hs

276 lines
7.2 KiB
Haskell
Raw Normal View History

2010-10-27 20:53:54 +00:00
{- git-annex monad
-
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
- Copyright 2010-2013 Joey Hess <joey@kitenet.net>
2010-10-27 20:53:54 +00:00
-
- Licensed under the GNU GPL version 3 or higher.
-}
2010-10-10 19:04:07 +00:00
2013-11-22 16:21:37 +00:00
{-# LANGUAGE GeneralizedNewtypeDeriving, PackageImports #-}
2010-10-11 21:52:46 +00:00
module Annex (
Annex,
AnnexState(..),
PreferredContentMap,
2010-10-14 07:18:11 +00:00
new,
run,
2010-11-01 03:24:16 +00:00
eval,
getState,
changeState,
setFlag,
setField,
2012-04-30 17:59:05 +00:00
setOutput,
getFlag,
getField,
addCleanup,
gitRepo,
inRepo,
fromRepo,
calcRepo,
getGitConfig,
changeGitConfig,
changeGitRepo,
withCurrentState,
2010-10-11 21:52:46 +00:00
) where
2010-10-10 19:04:07 +00:00
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
import "mtl" Control.Monad.Reader
import "MonadCatchIO-transformers" Control.Monad.CatchIO
import Control.Concurrent
2010-10-16 20:20:49 +00:00
import Common
import qualified Git
import qualified Git.Config
import Annex.Direct.Fixup
import Git.CatFile
import Git.CheckAttr
gitignore support for the assistant and watcher Requires git 1.8.4 or newer. When it's installed, a background git check-ignore process is run, and used to efficiently check ignores whenever a new file is added. Thanks to Adam Spiers, for getting the necessary support into git for this. A complication is what to do about files that are gitignored but have been checked into git anyway. git commands assume the ignore has been overridden in this case, and not need any more overriding to commit a changed version. However, for the assistant to do the same, it would have to run git ls-files to check if the ignored file is in git. This is somewhat expensive. Or it could use the running git-cat-file process to query the file that way, but that requires transferring the whole file content over a pipe, so it can be quite expensive too, for files that are not git-annex symlinks. Now imagine if the user knows that a file or directory tree will be getting frequent changes, and doesn't want the assistant to sync it, so gitignores it. The assistant could overload the system with repeated ls-files checks! So, I've decided that the assistant will not automatically commit changes to files that are gitignored. This is a tradeoff. Hopefully it won't be a problem to adjust .gitignore settings to not ignore files you want the assistant to autocommit, or to manually git annex add files that are listed in .gitignore. (This could be revisited if git-annex gets access to an interface to check the content of the index w/o forking a git command. This could be libgit2, or perhaps a separate git cat-file --batch-check process, so it wouldn't need to ship over the whole file content.) This commit was sponsored by Francois Marier. Thanks!
2013-08-02 23:31:55 +00:00
import Git.CheckIgnore
2012-04-21 23:42:49 +00:00
import Git.SharedRepository
import qualified Git.Hook
2011-12-20 18:37:53 +00:00
import qualified Git.Queue
import Types.Key
import Types.Backend
import Types.GitConfig
import qualified Types.Remote
import Types.Crypto
import Types.BranchState
import Types.TrustLevel
import Types.Group
import Types.Messages
import Types.UUID
2013-05-25 03:07:26 +00:00
import Types.FileMatcher
import Types.NumCopies
import Types.LockPool
import Types.MetaData
import Types.DesktopNotify
2014-03-13 23:06:26 +00:00
import Types.CleanupActions
import qualified Utility.Matcher
import qualified Data.Map as M
import qualified Data.Set as S
import Utility.Quvi (QuviVersion)
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
{- git-annex's monad is a ReaderT around an AnnexState stored in a MVar.
- This allows modifying the state in an exception-safe fashion.
- The MVar is not exposed outside this module.
-}
newtype Annex a = Annex { runAnnex :: ReaderT (MVar AnnexState) IO a }
deriving (
Monad,
MonadIO,
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
MonadReader (MVar AnnexState),
MonadCatchIO,
Functor,
Applicative
)
2011-12-31 08:19:10 +00:00
type Matcher a = Either [Utility.Matcher.Token a] (Utility.Matcher.Matcher a)
type PreferredContentMap = M.Map UUID (Utility.Matcher.Matcher (S.Set UUID -> MatchInfo -> Annex Bool))
-- internal state storage
data AnnexState = AnnexState
{ repo :: Git.Repo
, gitconfig :: GitConfig
2011-12-31 08:11:39 +00:00
, backends :: [BackendA Annex]
, remotes :: [Types.Remote.RemoteA Annex]
, remoteannexstate :: M.Map UUID AnnexState
, output :: MessageState
, force :: Bool
, fast :: Bool
, auto :: Bool
, daemon :: Bool
, branchstate :: BranchState
, repoqueue :: Maybe Git.Queue.Queue
, catfilehandles :: M.Map FilePath CatFileHandle
, checkattrhandle :: Maybe CheckAttrHandle
gitignore support for the assistant and watcher Requires git 1.8.4 or newer. When it's installed, a background git check-ignore process is run, and used to efficiently check ignores whenever a new file is added. Thanks to Adam Spiers, for getting the necessary support into git for this. A complication is what to do about files that are gitignored but have been checked into git anyway. git commands assume the ignore has been overridden in this case, and not need any more overriding to commit a changed version. However, for the assistant to do the same, it would have to run git ls-files to check if the ignored file is in git. This is somewhat expensive. Or it could use the running git-cat-file process to query the file that way, but that requires transferring the whole file content over a pipe, so it can be quite expensive too, for files that are not git-annex symlinks. Now imagine if the user knows that a file or directory tree will be getting frequent changes, and doesn't want the assistant to sync it, so gitignores it. The assistant could overload the system with repeated ls-files checks! So, I've decided that the assistant will not automatically commit changes to files that are gitignored. This is a tradeoff. Hopefully it won't be a problem to adjust .gitignore settings to not ignore files you want the assistant to autocommit, or to manually git annex add files that are listed in .gitignore. (This could be revisited if git-annex gets access to an interface to check the content of the index w/o forking a git command. This could be libgit2, or perhaps a separate git cat-file --batch-check process, so it wouldn't need to ship over the whole file content.) This commit was sponsored by Francois Marier. Thanks!
2013-08-02 23:31:55 +00:00
, checkignorehandle :: Maybe (Maybe CheckIgnoreHandle)
, forcebackend :: Maybe String
, globalnumcopies :: Maybe NumCopies
, forcenumcopies :: Maybe NumCopies
, limit :: Matcher (MatchInfo -> Annex Bool)
, uuidmap :: Maybe UUIDMap
, preferredcontentmap :: Maybe PreferredContentMap
2012-04-21 23:42:49 +00:00
, shared :: Maybe SharedRepository
, forcetrust :: TrustMap
, trustmap :: Maybe TrustMap
, groupmap :: Maybe GroupMap
, ciphers :: M.Map StorableCipher Cipher
, lockpool :: LockPool
, flags :: M.Map String Bool
, fields :: M.Map String String
, modmeta :: [ModMeta]
2014-03-13 23:06:26 +00:00
, cleanup :: M.Map CleanupAction (Annex ())
, inodeschanged :: Maybe Bool
, useragent :: Maybe String
, errcounter :: Integer
, unusedkeys :: Maybe (S.Set Key)
, quviversion :: Maybe QuviVersion
, existinghooks :: M.Map Git.Hook.Hook Bool
, desktopnotify :: DesktopNotify
2011-04-16 20:41:46 +00:00
}
newState :: GitConfig -> Git.Repo -> AnnexState
newState c r = AnnexState
{ repo = r
, gitconfig = c
, backends = []
, remotes = []
, remoteannexstate = M.empty
, output = defaultMessageState
, force = False
, fast = False
, auto = False
, daemon = False
, branchstate = startBranchState
, repoqueue = Nothing
, catfilehandles = M.empty
, checkattrhandle = Nothing
gitignore support for the assistant and watcher Requires git 1.8.4 or newer. When it's installed, a background git check-ignore process is run, and used to efficiently check ignores whenever a new file is added. Thanks to Adam Spiers, for getting the necessary support into git for this. A complication is what to do about files that are gitignored but have been checked into git anyway. git commands assume the ignore has been overridden in this case, and not need any more overriding to commit a changed version. However, for the assistant to do the same, it would have to run git ls-files to check if the ignored file is in git. This is somewhat expensive. Or it could use the running git-cat-file process to query the file that way, but that requires transferring the whole file content over a pipe, so it can be quite expensive too, for files that are not git-annex symlinks. Now imagine if the user knows that a file or directory tree will be getting frequent changes, and doesn't want the assistant to sync it, so gitignores it. The assistant could overload the system with repeated ls-files checks! So, I've decided that the assistant will not automatically commit changes to files that are gitignored. This is a tradeoff. Hopefully it won't be a problem to adjust .gitignore settings to not ignore files you want the assistant to autocommit, or to manually git annex add files that are listed in .gitignore. (This could be revisited if git-annex gets access to an interface to check the content of the index w/o forking a git command. This could be libgit2, or perhaps a separate git cat-file --batch-check process, so it wouldn't need to ship over the whole file content.) This commit was sponsored by Francois Marier. Thanks!
2013-08-02 23:31:55 +00:00
, checkignorehandle = Nothing
, forcebackend = Nothing
, globalnumcopies = Nothing
, forcenumcopies = Nothing
, limit = Left []
, uuidmap = Nothing
, preferredcontentmap = Nothing
2012-04-21 23:42:49 +00:00
, shared = Nothing
, forcetrust = M.empty
, trustmap = Nothing
, groupmap = Nothing
, ciphers = M.empty
, lockpool = M.empty
, flags = M.empty
, fields = M.empty
, modmeta = []
, cleanup = M.empty
, inodeschanged = Nothing
, useragent = Nothing
, errcounter = 0
, unusedkeys = Nothing
, quviversion = Nothing
, existinghooks = M.empty
, desktopnotify = mempty
}
2010-10-14 07:18:11 +00:00
{- Makes an Annex state object for the specified git repo.
- Ensures the config is read, if it was not already. -}
new :: Git.Repo -> IO AnnexState
new r = do
r' <- Git.Config.read r
let c = extractGitConfig r'
newState c <$> if annexDirect c then fixupDirect r' else return r'
2010-10-14 07:18:11 +00:00
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
{- Performs an action in the Annex monad from a starting state,
- returning a new state. -}
2011-01-11 22:13:26 +00:00
run :: AnnexState -> Annex a -> IO (a, AnnexState)
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
run s a = do
mvar <- newMVar s
r <- runReaderT (runAnnex a) mvar
s' <- takeMVar mvar
return (r, s')
{- Performs an action in the Annex monad from a starting state,
- and throws away the new state. -}
2011-01-11 22:13:26 +00:00
eval :: AnnexState -> Annex a -> IO a
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
eval s a = do
mvar <- newMVar s
runReaderT (runAnnex a) mvar
getState :: (AnnexState -> v) -> Annex v
getState selector = do
mvar <- ask
s <- liftIO $ readMVar mvar
return $ selector s
changeState :: (AnnexState -> AnnexState) -> Annex ()
changeState modifier = do
mvar <- ask
liftIO $ modifyMVar_ mvar $ return . modifier
2010-10-14 07:18:11 +00:00
{- Sets a flag to True -}
setFlag :: String -> Annex ()
setFlag flag = changeState $ \s ->
s { flags = M.insertWith' const flag True $ flags s }
{- Sets a field to a value -}
setField :: String -> String -> Annex ()
setField field value = changeState $ \s ->
s { fields = M.insertWith' const field value $ fields s }
{- Adds a cleanup action to perform. -}
2014-03-13 23:06:26 +00:00
addCleanup :: CleanupAction -> Annex () -> Annex ()
addCleanup k a = changeState $ \s ->
s { cleanup = M.insertWith' const k a $ cleanup s }
2012-04-30 17:59:05 +00:00
{- Sets the type of output to emit. -}
setOutput :: OutputType -> Annex ()
setOutput o = changeState $ \s ->
s { output = (output s) { outputType = o } }
{- Checks if a flag was set. -}
getFlag :: String -> Annex Bool
getFlag flag = fromMaybe False . M.lookup flag <$> getState flags
{- Gets the value of a field. -}
getField :: String -> Annex (Maybe String)
getField field = M.lookup field <$> getState fields
{- Returns the annex's git repository. -}
2010-10-14 07:18:11 +00:00
gitRepo :: Annex Git.Repo
gitRepo = getState repo
{- Runs an IO action in the annex's git repository. -}
inRepo :: (Git.Repo -> IO a) -> Annex a
2011-11-12 18:24:07 +00:00
inRepo a = liftIO . a =<< gitRepo
{- Extracts a value from the annex's git repisitory. -}
fromRepo :: (Git.Repo -> a) -> Annex a
fromRepo a = a <$> gitRepo
Switch to MonadCatchIO-transformers for better handling of state while catching exceptions. As seen in this bug report, the lifted exception handling using the StateT monad throws away state changes when an action throws an exception. http://git-annex.branchable.com/bugs/git_annex_fork_bombs_on_gpg_file/ .. Which can result in cached values being redundantly calculated, or other possibly worse bugs when the annex state gets out of sync with reality. This switches from a StateT AnnexState to a ReaderT (MVar AnnexState). All changes to the state go via the MVar. So when an Annex action is running inside an exception handler, and it makes some changes, they immediately go into affect in the MVar. If it then throws an exception (or even crashes its thread!), the state changes are still in effect. The MonadCatchIO-transformers change is actually only incidental. I could have kept on using lifted-base for the exception handling. However, I'd have needed to write a new instance of MonadBaseControl for the new monad.. and I didn't write the old instance.. I begged Bas and he kindly sent it to me. Happily, MonadCatchIO-transformers is able to derive a MonadCatchIO instance for my monad. This is a deep level change. It passes the test suite! What could it break? Well.. The most likely breakage would be to code that runs an Annex action in an exception handler, and *wants* state changes to be thrown away. Perhaps the state changes leaves the state inconsistent, or wrong. Since there are relatively few places in git-annex that catch exceptions in the Annex monad, and the AnnexState is generally just used to cache calculated data, this is unlikely to be a problem. Oh yeah, this change also makes Assistant.Types.ThreadedMonad a bit redundant. It's now entirely possible to run concurrent Annex actions in different threads, all sharing access to the same state! The ThreadedMonad just adds some extra work on top of that, with its own MVar, and avoids such actions possibly stepping on one-another's toes. I have not gotten rid of it, but might try that later. Being able to run concurrent Annex actions would simplify parts of the Assistant code.
2013-05-19 18:16:36 +00:00
{- Calculates a value from an annex's git repository and its GitConfig. -}
calcRepo :: (Git.Repo -> GitConfig -> IO a) -> Annex a
calcRepo a = do
s <- getState id
liftIO $ a (repo s) (gitconfig s)
{- Gets the GitConfig settings. -}
getGitConfig :: Annex GitConfig
getGitConfig = getState gitconfig
{- Modifies a GitConfig setting. -}
changeGitConfig :: (GitConfig -> GitConfig) -> Annex ()
changeGitConfig a = changeState $ \s -> s { gitconfig = a (gitconfig s) }
{- Changing the git Repo data also involves re-extracting its GitConfig. -}
changeGitRepo :: Git.Repo -> Annex ()
changeGitRepo r = changeState $ \s -> s
{ repo = r
, gitconfig = extractGitConfig r
}
{- Converts an Annex action into an IO action, that runs with a copy
- of the current Annex state.
-
- Use with caution; the action should not rely on changing the
- state, as it will be thrown away. -}
withCurrentState :: Annex a -> Annex (IO a)
withCurrentState a = do
s <- getState id
return $ eval s a