pre-commit-annex hook script to automatically extract metadata from lots of types of files

Using the extract(1) program to do the heavy lifting.

Decided to make git-annex run pre-commit-annex when committing. Since
git-annex pre-commit also runs it, it'll be run when git commit is run too,
via the pre-commit hook. This basically gives back the pre-commit hook
that git-annex took away. The implementation avoids repeatedly looking
for the hook script when the assistant is running and committing
repeatedly; only checks if the hook is available once.

To make the script simpler, made git-annex metadata -s field?=value
only set a field when it's not already got a value.

This commit was sponsored by bak.

Annex.hs

@@ -44,6 +44,7 @@ import Git.CatFile
 import Git.CheckAttr
 import Git.CheckIgnore
 import Git.SharedRepository
+import qualified Git.Hook
 import qualified Git.Queue
 import Types.Key
 import Types.Backend
@@ -118,6 +119,7 @@ data AnnexState = AnnexState
 	, errcounter :: Integer
 	, unusedkeys :: Maybe (S.Set Key)
 	, quviversion :: Maybe QuviVersion
+	, existinghooks :: M.Map Git.Hook.Hook Bool
 	}
 
 newState :: GitConfig -> Git.Repo -> AnnexState
@@ -157,6 +159,7 @@ newState c r = AnnexState
 	, errcounter = 0
 	, unusedkeys = Nothing
 	, quviversion = Nothing
+	, existinghooks = M.empty
 	}
 
 {- Makes an Annex state object for the specified git repo.

Annex/Hook.hs

@@ -1,9 +1,10 @@
 {- git-annex git hooks
  -
- - Note that it's important that the scripts not change, otherwise
- - removing old hooks using an old version of the script would fail.
+ - Note that it's important that the scripts installed by git-annex
+ - not change, otherwise removing old hooks using an old version of
+ - the script would fail.
  -
- - Copyright 2013 Joey Hess <joey@kitenet.net>
+ - Copyright 2013-2014 Joey Hess <joey@kitenet.net>
  -
  - Licensed under the GNU GPL version 3 or higher.
  -}
@@ -12,12 +13,19 @@ module Annex.Hook where
 
 import Common.Annex
 import qualified Git.Hook as Git
-import Utility.Shell
 import Config
+import qualified Annex
+import Utility.Shell
+import Utility.FileMode
+
+import qualified Data.Map as M
 
 preCommitHook :: Git.Hook
 preCommitHook = Git.Hook "pre-commit" (mkHookScript "git annex pre-commit .")
 
+preCommitAnnexHook :: Git.Hook
+preCommitAnnexHook = Git.Hook "pre-commit-annex" ""
+
 mkHookScript :: String -> String
 mkHookScript s = unlines
 	[ shebang_local
@@ -40,3 +48,23 @@ hookWarning :: Git.Hook -> String -> Annex ()
 hookWarning h msg = do
 	r <- gitRepo
 	warning $ Git.hookName h ++ " hook (" ++ Git.hookFile h r ++ ") " ++ msg
+
+{- Runs a hook. To avoid checking if the hook exists every time,
+ - the existing hooks are cached. -}
+runAnnexHook :: Git.Hook -> Annex ()
+runAnnexHook hook = do
+	cmd <- fromRepo $ Git.hookFile hook
+	m <- Annex.getState Annex.existinghooks
+	case M.lookup hook m of
+		Just True -> run cmd
+		Just False -> noop
+		Nothing -> do
+			exists <- hookexists cmd
+			Annex.changeState $ \s -> s
+				{ Annex.existinghooks = M.insert hook exists m }
+			when exists $
+				run cmd
+  where
+	hookexists f = liftIO $ isExecutable . fileMode <$> getFileStatus f
+	run cmd = unlessM (liftIO $ boolSystem cmd []) $
+		warning $ cmd ++ " failed"

Command/PreCommit.hs

@@ -13,6 +13,7 @@ import Config
 import qualified Command.Add
 import qualified Command.Fix
 import Annex.Direct
+import Annex.Hook
 import Annex.View
 import Annex.View.ViewedFile
 import Logs.View
@@ -28,13 +29,16 @@ def = [command "pre-commit" paramPaths seek SectionPlumbing
 
 seek :: CommandSeek
 seek ps = ifM isDirect
-	-- update direct mode mappings for committed files
-	( withWords startDirect ps
+	( do
+		-- update direct mode mappings for committed files
+		withWords startDirect ps
+		runAnnexHook preCommitAnnexHook
 	, do
 		-- fix symlinks to files being committed
 		withFilesToBeCommitted (whenAnnexed Command.Fix.start) ps
 		-- inject unlocked files into the annex
 		withFilesUnlockedToBeCommitted startIndirect ps
+		runAnnexHook preCommitAnnexHook
 		-- committing changes to a view updates metadata
 		mv <- currentView
 		case mv of
@@ -43,6 +47,7 @@ seek ps = ifM isDirect
 				(addViewMetaData v)
 				(removeViewMetaData v)
 	)
+	
 
 startIndirect :: FilePath -> CommandStart
 startIndirect f = next $ do

Command/Sync.hs

@@ -18,6 +18,7 @@ import qualified Types.Remote as Remote
 import Annex.Direct
 import Annex.CatFile
 import Annex.Link
+import Annex.Hook
 import qualified Git.Command
 import qualified Git.LsFiles as LsFiles
 import qualified Git.Merge
@@ -156,6 +157,7 @@ commitStaged commitmessage = go =<< inRepo Git.Branch.currentUnsafe
   where
 	go Nothing = return False
 	go (Just branch) = do
+		runAnnexHook preCommitAnnexHook
 		parent <- inRepo $ Git.Ref.sha branch
 		void $ inRepo $ Git.Branch.commit False commitmessage branch
 			(maybeToList parent)

Git/Hook.hs

@@ -15,6 +15,10 @@ data Hook = Hook
 	{ hookName :: FilePath
 	, hookScript :: String
 	}
+	deriving (Ord)
+
+instance Eq Hook where
+	a == b = hookName a == hookName b
 
 hookFile :: Hook -> Repo -> FilePath
 hookFile h r = localGitDir r </> "hooks" </> hookName h

Types/MetaData.hs

@@ -219,6 +219,7 @@ data ModMeta
 	= AddMeta MetaField MetaValue
 	| DelMeta MetaField MetaValue
 	| SetMeta MetaField MetaValue -- removes any existing values
+	| MaybeSetMeta MetaField MetaValue -- when field has no existing value
 
 {- Applies a ModMeta, generating the new MetaData.
  - Note that the new MetaData does not include all the 
@@ -229,12 +230,16 @@ modMeta _ (DelMeta f oldv) = updateMetaData f (unsetMetaValue oldv) emptyMetaDat
 modMeta m (SetMeta f v) = updateMetaData f v $
 	foldr (updateMetaData f) emptyMetaData $
 		map unsetMetaValue $ S.toList $ currentMetaDataValues f m
+modMeta m (MaybeSetMeta f v)
+	| S.null (currentMetaDataValues f m) = updateMetaData f v emptyMetaData
+	| otherwise = emptyMetaData
 
-{- Parses field=value, field+=value, field-=value -}
+{- Parses field=value, field+=value, field-=value, field?=value -}
 parseModMeta :: String -> Either String ModMeta
 parseModMeta p = case lastMaybe f of
 	Just '+' -> AddMeta <$> mkMetaField f' <*> v
 	Just '-' -> DelMeta <$> mkMetaField f' <*> v
+	Just '?' -> MaybeSetMeta <$> mkMetaField f' <*> v
 	_ -> SetMeta <$> mkMetaField f <*> v
   where
 	(f, sv) = separate (== '=') p

debian/changelog

@@ -16,6 +16,9 @@ git-annex (5.20140228) UNRELEASED; urgency=medium
     using !tag and field!=value.
   * vadd: Allow listing multiple desired values for a field.
   * view: Refuse to enter a view when no branch is currently checked out.
+  * metadata: To only set a field when it's not already got a value, use
+    -s field?=value
+  * Run .git/hooks/pre-commit-annex whenever a commit is made.
 
  -- Joey Hess <joeyh@debian.org>  Fri, 28 Feb 2014 14:52:15 -0400
 

doc/design/metadata.mdwn

@@ -56,14 +56,7 @@ once, and can be left alone when refining a view.
 When annex.genmetadata is set, git annex add automatically attaches
 some metadata to a file. Currently year and month fields, from its mtime.
 
-TODO A git hook could be run by git annex add to gather more metadata.
-For example, by examining file permisions or MP3 metadata.
-Alternatively, this could be a
-regular post-commit hook, that examines the files committed, and runs git
-annex metadata to add metadata. No extra git-annex support is needed
-to do that!
-However, in direct mode, or when using the assistant, git-annex does its
-own committing, not using git commit, so bypassing the commit hooks.
+There's also a post-commit-annex hook script.
 
 ## directory hierarchy metadata
 

doc/git-annex.mdwn

@@ -717,6 +717,9 @@ subdirectories).
 
   To remove a value, use -s field-=value.
 
+  To set a value, only if the field does not already have a value,
+  use -s field?=value
+
   To set a tag, use -t tag, and use -u tag to remove a tag.
 
   For example, to set some tags on a file and also its author:
@@ -1678,6 +1681,10 @@ used by git-annex.
 `~/.config/git-annex/autostart` is a list of git repositories
 to start the git-annex assistant in.
 
+`.git/hooks/pre-commit-annex` in your git repsitory will be run whenever
+a commit is made, either by git commit, git-annex sync, or the git-annex
+assistant.
+
 # SEE ALSO
 
 Most of git-annex's documentation is available on its web site,

doc/metadata.mdwn

@@ -30,7 +30,8 @@ Here are some recommended metadata fields to use:
   being.
   
 To make git-annex automatically set the year and month when adding files,
-run `git config annex.genmetadata true`
+run `git config annex.genmetadata true`. Also, see
+[[tips/automatically_adding_metadata]].
 
 git-annex's metadata can be updated in a distributed fashion. For example,
 two users, each with their own clone of a repository, can set and unset

doc/tips/automatically_adding_metadata.mdwn

@@ -0,0 +1,24 @@
+git-annex's [[metadata]] works best when files have a lot of useful
+metadata attached to them.
+
+To make git-annex automatically set the year and month when adding files,
+run `git config annex.genmetadata true`.
+
+A git commit hook can be set up to extract lots of metadata from files
+like photos, mp3s, etc.
+
+* Install the `extract` utility, from <http://www.gnu.org/software/libextractor/>  
+  `apt-get install extract`
+* Download [[pre-commit-annex]] and install it in your git-annex repository
+  as `.git/hooks/pre-commit-annex`.  
+  Remember to make the script executable!
+* Run: `git config metadata.extract "artist album title camera_make video_dimensions"`
+
+Now any fields you list in metadata.extract to will be extracted and
+stored when files are committed.
+
+To get a list of all possible fields, run: `extract -L | sed ' ' _`
+
+By default, if a git-annex already has a metadata field for a file,
+its value will not be overwritten with metadata taken from files.
+To allow overwriting, run: `git config metadata.overwrite true`

doc/tips/automatically_adding_metadata/pre-commit-annex

@@ -0,0 +1,61 @@
+#!/bin/sh
+# This script can be used to add git-annex metadata to files when they're
+# committed.
+#
+# Copyright 2014 Joey Hess <id@joeyh.name>
+# License: GPL-3+
+
+extract="$(git config metadata.extract || true)" 
+want="$(perl -e 'print (join("|", map {s/_/ /g; "^$_ - "} (split " ", shift())))' "$extract")"
+
+if [ -z "$want" ]; then
+	exit 0
+fi
+
+echo "$want"
+
+case "$(git config --bool metadata.overwrite || true)" in
+	true)
+		overwrite=1
+	;;
+	*)
+		overwrite=""
+	;;
+esac
+
+# By default, any existing metadata git-annex has stored is not overwritten
+# with metadata taken from files. Uncomment this to enable.
+#overwrite_metadata=1
+
+addmeta () {
+	file="$1"
+	field="$2"
+	value="$3"
+	afield="$(echo "$field" | tr ' ' _)"
+	if [ "$overwrite" ]; then
+		p="$afield=$value"
+
+	else
+		p="$afield?=$value"
+	fi
+	git -c annex.alwayscommit=false annex metadata "$file" -s "$p" --quiet
+}
+
+if git rev-parse --verify HEAD >/dev/null 2>&1; then
+	against=HEAD
+else
+	# Initial commit: diff against an empty tree object
+	against=4b825dc642cb6eb9a060e54bf8d69288fbee4904
+fi
+
+IFS="
+"
+for f in $(git diff-index --name-only --cached $against); do
+	if [ -e "$f" ]; then
+		for l in $(extract "$f" | egrep "$want"); do
+			field="${l%% - *}"
+			value="${l#* - }"
+			addmeta "$f" "$field" "$value"
+		done
+	fi
+done