git-annex/Git.hs

186 lines
5.4 KiB
Haskell
Raw Normal View History

2010-10-12 03:22:38 +00:00
{- git repository handling
-
- This is written to be completely independant of git-annex and should be
- suitable for other uses.
-
- Copyright 2010-2021 Joey Hess <id@joeyh.name>
2010-10-27 20:53:54 +00:00
-
- Licensed under the GNU AGPL version 3 or higher.
2010-10-14 06:36:41 +00:00
-}
2010-10-10 01:06:46 +00:00
{-# LANGUAGE OverloadedStrings #-}
2013-05-10 21:29:59 +00:00
{-# LANGUAGE CPP #-}
module Git (
2011-12-14 19:30:14 +00:00
Repo(..),
Ref(..),
fromRef,
fromRef',
Branch,
Sha,
Tag,
2010-10-22 18:05:30 +00:00
repoIsUrl,
2010-10-22 17:40:19 +00:00
repoIsSsh,
2011-08-16 23:23:56 +00:00
repoIsHttp,
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
repoIsLocal,
repoIsLocalBare,
repoIsLocalUnknown,
2010-10-14 06:36:41 +00:00
repoDescribe,
repoLocation,
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
repoPath,
repoWorkTree,
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
localGitDir,
2010-10-14 06:36:41 +00:00
attributes,
attributesLocal,
hookPath,
assertLocal,
adjustPath,
relPath,
2010-10-11 21:52:46 +00:00
) where
2010-10-10 01:06:46 +00:00
import qualified Data.ByteString as B
import Network.URI (uriPath, uriScheme, unEscapeString)
2013-08-04 17:07:55 +00:00
#ifndef mingw32_HOST_OS
2012-03-14 16:17:38 +00:00
import System.Posix.Files
2013-05-10 21:29:59 +00:00
#endif
import qualified System.FilePath.ByteString as P
2010-10-16 20:20:49 +00:00
import Common
import Git.Types
2013-08-04 17:12:18 +00:00
#ifndef mingw32_HOST_OS
2012-03-14 16:17:38 +00:00
import Utility.FileMode
2013-08-04 17:12:18 +00:00
#endif
{- User-visible description of a git repo. -}
2010-10-31 19:38:47 +00:00
repoDescribe :: Repo -> String
repoDescribe Repo { remoteName = Just name } = name
repoDescribe Repo { location = Url url } = show url
repoDescribe Repo { location = UnparseableUrl url } = url
repoDescribe Repo { location = Local { worktree = Just dir } } = fromRawFilePath dir
repoDescribe Repo { location = Local { gitdir = dir } } = fromRawFilePath dir
repoDescribe Repo { location = LocalUnknown dir } = fromRawFilePath dir
repoDescribe Repo { location = Unknown } = "UNKNOWN"
2010-10-13 18:40:56 +00:00
{- Location of the repo, either as a path or url. -}
repoLocation :: Repo -> String
repoLocation Repo { location = Url url } = show url
repoLocation Repo { location = UnparseableUrl url } = url
repoLocation Repo { location = Local { worktree = Just dir } } = fromRawFilePath dir
repoLocation Repo { location = Local { gitdir = dir } } = fromRawFilePath dir
repoLocation Repo { location = LocalUnknown dir } = fromRawFilePath dir
repoLocation Repo { location = Unknown } = error "unknown repoLocation"
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
{- Path to a repository. For non-bare, this is the worktree, for bare,
- it's the gitdir, and for URL repositories, is the path on the remote
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
- host. -}
repoPath :: Repo -> RawFilePath
repoPath Repo { location = Url u } = toRawFilePath $ unEscapeString $ uriPath u
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
repoPath Repo { location = Local { worktree = Just d } } = d
repoPath Repo { location = Local { gitdir = d } } = d
repoPath Repo { location = LocalUnknown dir } = dir
repoPath Repo { location = Unknown } = error "unknown repoPath"
repoPath Repo { location = UnparseableUrl _u } = error "unknwon repoPath"
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
repoWorkTree :: Repo -> Maybe RawFilePath
repoWorkTree Repo { location = Local { worktree = Just d } } = Just d
repoWorkTree _ = Nothing
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
{- Path to a local repository's .git directory. -}
localGitDir :: Repo -> RawFilePath
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
localGitDir Repo { location = Local { gitdir = d } } = d
localGitDir _ = error "unknown localGitDir"
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
2010-10-22 18:05:30 +00:00
{- Some code needs to vary between URL and normal repos,
2010-10-22 16:38:20 +00:00
- or bare and non-bare, these functions help with that. -}
2010-10-31 19:38:47 +00:00
repoIsUrl :: Repo -> Bool
repoIsUrl Repo { location = Url _ } = True
repoIsUrl Repo { location = UnparseableUrl _ } = True
repoIsUrl _ = False
2010-10-31 19:38:47 +00:00
repoIsSsh :: Repo -> Bool
repoIsSsh Repo { location = Url url }
2011-12-14 19:30:14 +00:00
| scheme == "ssh:" = True
-- git treats these the same as ssh
2011-12-14 19:30:14 +00:00
| scheme == "git+ssh:" = True
| scheme == "ssh+git:" = True
| otherwise = False
2012-11-11 04:51:07 +00:00
where
scheme = uriScheme url
repoIsSsh _ = False
2011-08-16 23:23:56 +00:00
repoIsHttp :: Repo -> Bool
repoIsHttp Repo { location = Url url }
| uriScheme url == "http:" = True
| uriScheme url == "https:" = True
| otherwise = False
repoIsHttp _ = False
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
repoIsLocal :: Repo -> Bool
repoIsLocal Repo { location = Local { } } = True
repoIsLocal _ = False
repoIsLocalBare :: Repo -> Bool
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
repoIsLocalBare Repo { location = Local { worktree = Nothing } } = True
repoIsLocalBare _ = False
repoIsLocalUnknown :: Repo -> Bool
repoIsLocalUnknown Repo { location = LocalUnknown { } } = True
repoIsLocalUnknown _ = False
2010-10-31 19:38:47 +00:00
assertLocal :: Repo -> a -> a
2012-03-16 05:59:07 +00:00
assertLocal repo action
| repoIsUrl repo = error $ unwords
[ "acting on non-local git repo"
, repoDescribe repo
, "not supported"
]
| otherwise = action
{- Path to a repository's gitattributes file. -}
attributes :: Repo -> RawFilePath
attributes repo
| repoIsLocalBare repo = attributesLocal repo
| otherwise = repoPath repo P.</> ".gitattributes"
attributesLocal :: Repo -> RawFilePath
attributesLocal repo = localGitDir repo P.</> "info" P.</> "attributes"
2010-10-10 06:29:58 +00:00
{- Path to a given hook script in a repository, only if the hook exists
- and is executable. -}
hookPath :: String -> Repo -> IO (Maybe FilePath)
hookPath script repo = do
let hook = fromRawFilePath (localGitDir repo) </> "hooks" </> script
2012-03-16 05:59:07 +00:00
ifM (catchBoolIO $ isexecutable hook)
( return $ Just hook , return Nothing )
2012-11-11 04:51:07 +00:00
where
2013-08-04 17:07:55 +00:00
#if mingw32_HOST_OS
2013-05-10 21:29:59 +00:00
isexecutable f = doesFileExist f
#else
2012-11-11 04:51:07 +00:00
isexecutable f = isExecutable . fileMode <$> getFileStatus f
2013-05-10 21:29:59 +00:00
#endif
{- Makes the path to a local Repo be relative to the cwd. -}
relPath :: Repo -> IO Repo
relPath = adjustPath torel
where
torel p = do
p' <- relPathCwdToFile p
return $ if B.null p' then "." else p'
{- Adusts the path to a local Repo using the provided function. -}
adjustPath :: (RawFilePath -> IO RawFilePath) -> Repo -> IO Repo
adjustPath f r@(Repo { location = l@(Local { gitdir = d, worktree = w }) }) = do
d' <- f d
w' <- maybe (pure Nothing) (Just <$$> f) w
return $ r
{ location = l
{ gitdir = d'
, worktree = w'
}
}
where
adjustPath f r@(Repo { location = LocalUnknown d }) = do
d' <- f d
return $ r { location = LocalUnknown d' }
adjustPath _ r = pure r