don't frontload reconcileStaged in git-annex init

init: Avoid scanning for annexed files, which can be lengthy in a
large repository. Instead that scan is done on demand. This lets git-annex
init be run and some query commands be used in a repository without
waiting.

Note that autoinit already behaved this way, so while this will mean some
commands like git-annex get/unlock/add will do the scan the first time run,
that is not really a significant behavior change.

And, it's really better to have a consistent behavior. The reason for
the inconsistency was a strange bug discussed in
b3c4579c79. Avoiding reconcileStaged in
init will keep avoiding whatever that was.

Sponsored-by: Dartmouth College's DANDI project

Annex/Init.hs

@@ -37,7 +37,6 @@ import Types.RepoVersion
 import Annex.Version
 import Annex.Difference
 import Annex.UUID
-import Annex.WorkTree
 import Annex.Fixup
 import Annex.Path
 import Config
@@ -102,8 +101,8 @@ genDescription Nothing = do
 		Right username -> [username, at, hostname, ":", reldir]
 		Left _ -> [hostname, ":", reldir]
 
-initialize :: Bool -> Maybe String -> Maybe RepoVersion -> Annex ()
-initialize autoinit mdescription mversion = checkInitializeAllowed $ \initallowed -> do
+initialize :: Maybe String -> Maybe RepoVersion -> Annex ()
+initialize mdescription mversion = checkInitializeAllowed $ \initallowed -> do
 	{- Has to come before any commits are made as the shared
 	 - clone heuristic expects no local objects. -}
 	sharedclone <- checkSharedClone
@@ -113,7 +112,7 @@ initialize autoinit mdescription mversion = checkInitializeAllowed $ \initallowe
 	ensureCommit $ Annex.Branch.create
 
 	prepUUID
-	initialize' autoinit mversion initallowed
+	initialize' mversion initallowed
 	
 	initSharedClone sharedclone
 	
@@ -125,8 +124,8 @@ initialize autoinit mdescription mversion = checkInitializeAllowed $ \initallowe
 
 -- Everything except for uuid setup, shared clone setup, and initial
 -- description.
-initialize' :: Bool -> Maybe RepoVersion -> InitializeAllowed -> Annex ()
-initialize' autoinit mversion _initallowed = do
+initialize' :: Maybe RepoVersion -> InitializeAllowed -> Annex ()
+initialize' mversion _initallowed = do
 	checkLockSupport
 	checkFifoSupport
 	checkCrippledFileSystem
@@ -143,8 +142,6 @@ initialize' autoinit mversion _initallowed = do
 	unlessM isBareRepo $ do
 		hookWrite postCheckoutHook
 		hookWrite postMergeHook
-		unless autoinit $
-			scanAnnexedFiles
 
 	AdjustedBranch.checkAdjustedClone >>= \case
 		AdjustedBranch.InAdjustedClone -> return ()
@@ -206,7 +203,7 @@ ensureInitialized remotelist = getInitializedVersion >>= maybe needsinit checkUp
   where
 	needsinit = ifM autoInitializeAllowed
 		( do
-			tryNonAsync (initialize True Nothing Nothing) >>= \case
+			tryNonAsync (initialize Nothing Nothing) >>= \case
 				Right () -> noop
 				Left e -> giveup $ show e ++ "\n" ++
 					"git-annex: automatic initialization failed due to above problems"
@@ -259,7 +256,7 @@ autoInitialize remotelist = getInitializedVersion >>= maybe needsinit checkUpgra
   where
 	needsinit =
 		whenM (initializeAllowed <&&> autoInitializeAllowed) $ do
-			initialize True Nothing Nothing
+			initialize Nothing Nothing
 			autoEnableSpecialRemotes remotelist
 
 {- Checks if a repository is initialized. Does not check version for ugrade. -}

Assistant/MakeRepo.hs

@@ -85,7 +85,7 @@ initRepo False _ dir desc mgroup = inDir dir $ do
 
 initRepo' :: Maybe String -> Maybe StandardGroup -> Annex ()
 initRepo' desc mgroup = unlessM isInitialized $ do
-	initialize False desc Nothing
+	initialize desc Nothing
 	u <- getUUID
 	maybe noop (defaultStandardGroup u) mgroup
 	{- Ensure branch gets committed right away so it is

CHANGELOG

@@ -1,7 +1,9 @@
 git-annex (10.20221105) UNRELEASED; urgency=medium
 
   * Support quettabyte and yottabyte.
-  * Sped up the initial scanning for annexed files by 21%.
+  * Sped up the initial scan for annexed files by 21%.
+  * init: Avoid scanning for annexed files, which can be lengthy in a
+    large repository. Instead that scan is done on demand.
 
  -- Joey Hess <id@joeyh.name>  Fri, 18 Nov 2022 12:58:06 -0400
 

Command/ConfigList.hs

@@ -47,7 +47,7 @@ findOrGenUUID = do
 		else ifM (Annex.Branch.hasSibling <||> (isJust <$> Fields.getField Fields.autoInit))
 			( do
 				liftIO checkNotReadOnly
-				initialize True Nothing Nothing
+				initialize Nothing Nothing
 				getUUID
 			, return NoUUID
 			)

Command/Init.hs

@@ -75,7 +75,7 @@ perform os = do
 			Just v | v /= wantversion ->
 				giveup $ "This repository is already a initialized with version " ++ show (fromRepoVersion v) ++ ", not changing to requested version."
 			_ -> noop
-	initialize False
+	initialize
 		(if null (initDesc os) then Nothing else Just (initDesc os))
 		(initVersion os)
 	unless (noAutoEnable os)

Command/Reinit.hs

@@ -35,6 +35,6 @@ perform s = do
 		then return $ toUUID s
 		else Remote.nameToUUID s
 	storeUUID u
-	checkInitializeAllowed $ initialize' False Nothing
+	checkInitializeAllowed $ initialize' Nothing
 	Annex.SpecialRemote.autoEnable
 	next $ return True

Command/Upgrade.hs

@@ -45,6 +45,6 @@ start (UpgradeOptions { autoOnly = True }) =
 start _ =
 	starting "upgrade" (ActionItemOther Nothing) (SeekInput []) $ do
 		whenM (isNothing <$> getVersion) $ do
-			initialize False Nothing Nothing
+			initialize Nothing Nothing
 		r <- upgrade False latestVersion
 		next $ return r

doc/bugs/performance_regression__63___init_takes_times_more/comment_14_8c3b13806adb731435b346a64990527b._comment

@@ -5,4 +5,15 @@
  content="""
 Implemented the two optimisations discussed above, and init in that
 repository dropped from 24 seconds to 19 seconds, a 21% speedup.
+
+I think that's as fast as reconcileStaged is likely to get without
+some deep optimisation of the persistent library.
+
+Then I realized that `git-annex init` does not really need to scan for
+associated files. That can be done later, when running a command that needs
+to access the keys database. Indeed, when git-annex is used in a clone of
+an annexed repo without explicitly running `git-annex init`, that's what
+it already did. I've implemented that, so now `git-annex init` takes 3
+seconds or so. The price will be paid later, the first time running a
+`git-annex add` or `git-annex unlock` or `git-annex get`.
 """]]