hbs2/hbs2-git3/lib/HBS2/Git3/State/Index.hs

313 lines
9.6 KiB
Haskell

module HBS2.Git3.State.Index where
import HBS2.Git3.Prelude
import HBS2.System.Dir
import HBS2.CLI.Run.Internal.Merkle (getTreeContents)
import HBS2.Git3.State.Types
import HBS2.Data.Log.Structured
import Data.ByteString qualified as BS
import Data.ByteString.Lazy ( ByteString )
import Data.ByteString.Lazy qualified as LBS
import Data.List qualified as L
import Network.ByteOrder qualified as N
import System.IO.Temp as Temp
import Data.ByteString.Lazy qualified as LBS
import Data.Fixed
import Data.Maybe
import Data.HashMap.Strict (HashMap)
import Data.HashMap.Strict qualified as HM
import Data.HashSet (HashSet)
import Data.HashSet qualified as HS
import Data.Word
import Data.Vector (Vector)
import Data.Vector qualified as V
import Data.Kind
import Data.BloomFilter qualified as Bloom
import Data.BloomFilter (Bloom(..))
import Data.BloomFilter.Mutable qualified as MBloom
import Control.Monad.ST
import Control.Concurrent.STM qualified as STM
import Codec.Compression.Zstd.Lazy qualified as ZstdL
import Codec.Serialise
import Streaming.Prelude qualified as S
import Streaming hiding (run,chunksOf)
import System.TimeIt
import UnliftIO
import UnliftIO.IO.File qualified as UIO
import Data.HashPSQ qualified as HPSQ
readLogFileLBS :: forall opts m . ( MonadIO m, ReadLogOpts opts, BytesReader m )
=> opts
-> ( GitHash -> Int -> ByteString -> m () )
-> m Int
readLogFileLBS _ action = flip fix 0 \go n -> do
done <- noBytesLeft
if done then pure n
else do
ssize <- readBytesMaybe 4
>>= orThrow SomeReadLogError
<&> fromIntegral . N.word32 . LBS.toStrict
hash <- readBytesMaybe 20
>>= orThrow SomeReadLogError
<&> GitHash . BS.copy . LBS.toStrict
sdata <- readBytesMaybe ( ssize - 20 )
>>= orThrow SomeReadLogError
void $ action hash (fromIntegral ssize) sdata
go (succ n)
indexPath :: forall m . ( Git3Perks m
, MonadReader Git3Env m
) => m FilePath
indexPath = do
reflog <- getGitRemoteKey >>= orThrow Git3ReflogNotSet
getStatePath (AsBase58 reflog) <&> (</> "index")
data IndexEntry =
IndexEntry
{ entryFile :: FilePath
, entryBS :: N.ByteString
}
data Index a =
Index { entries :: [IndexEntry]
}
openIndex :: forall a m . (Git3Perks m, MonadReader Git3Env m)
=> m (Index a)
openIndex = do
files <- listObjectIndexFiles
bss <- liftIO $ for files $ \(f,_) -> (f,) <$> mmapFileByteString f Nothing
let entries = [ IndexEntry f bs | (f,bs) <- bss ]
pure $ Index entries
indexEntryLookup :: forall a m . (Git3Perks m)
=> Index a
-> GitHash
-> m (Maybe N.ByteString)
indexEntryLookup Index{..} h = do
already_ <- newTVarIO ( mempty :: HashMap GitHash N.ByteString )
forConcurrently_ entries $ \IndexEntry{..} -> do
what <- readTVarIO already_ <&> HM.lookup h
case what of
Just{} -> none
Nothing -> do
offset' <- binarySearchBS 56 ( BS.take 20 . BS.drop 4 ) (coerce h) entryBS
maybe1 offset' none $ \offset -> do
let ebs = BS.take 32 $ BS.drop (offset + 4 + 20) entryBS
atomically $ modifyTVar already_ (HM.insert h ebs)
readTVarIO already_ <&> headMay . HM.elems
indexFilterNewObjects :: forall a m . (Git3Perks m)
=> Index a
-> HashSet GitHash
-> m [GitHash]
indexFilterNewObjects Index{..} hashes = do
old_ <- newTVarIO ( mempty :: HashSet GitHash )
forConcurrently_ entries $ \IndexEntry{..} -> do
flip fix (HS.toList hashes) $ \next -> \case
[] -> none
(x:xs) -> do
old <- readTVarIO old_ <&> HS.member x
if old then
next xs
else do
off <- binarySearchBS 56 ( BS.take 20 . BS.drop 4 ) (coerce x) entryBS
when (isJust off) do
atomically $ modifyTVar old_ (HS.insert x)
next xs
old <- readTVarIO old_
pure $ HS.toList (hashes `HS.difference` old)
indexFilterNewObjectsMem :: forall a m . (Git3Perks m)
=> Index a
-> HashSet GitHash
-> m [GitHash]
indexFilterNewObjectsMem idx@Index{..} hashes = do
old_ <- newTVarIO ( mempty :: HashSet GitHash )
enumEntries idx $ \bs -> do
atomically $ modifyTVar old_ (HS.insert (coerce $ BS.take 20 bs))
old <- readTVarIO old_
pure $ HS.toList (hashes `HS.difference` old)
listObjectIndexFiles :: forall m . ( Git3Perks m
, MonadReader Git3Env m
) => m [(FilePath, Natural)]
listObjectIndexFiles = do
path <- indexPath
dirFiles path
<&> filter ( ("objects*.idx" ?==) . takeFileName )
>>= \fs -> for fs $ \f -> do
z <- fileSize f <&> fromIntegral
pure (f,z)
enumEntries :: forall a m . ( Git3Perks m
) => Index a -> ( BS.ByteString -> m () ) -> m ()
enumEntries Index{..} action = do
for_ entries $ \IndexEntry{..} -> do
scanBS entryBS action
enumEntriesFixed :: forall a m . ( Git3Perks m
)
=> Int
-> Index a
-> ( BS.ByteString -> m () )
-> m ()
enumEntriesFixed n Index{..} action = do
q <- newTQueueIO
atomically $ mapM_ (writeTQueue q) entries
replicateM_ n $ do
fix \next -> do
es' <- atomically $ tryReadTQueue q
case es' of
Nothing -> none
Just IndexEntry{..} -> do
scanBS entryBS action
next
bloomHash :: GitHash -> [Word32]
bloomHash gh = [a,b,c,d,e]
where
bs = coerce gh
a = N.word32 (BS.take 4 bs)
b = N.word32 (BS.take 4 $ BS.drop 4 bs)
c = N.word32 (BS.take 4 $ BS.drop 8 bs)
d = N.word32 (BS.take 4 $ BS.drop 12 bs)
e = N.word32 (BS.take 4 $ BS.drop 16 bs)
bloomFilterSize :: Natural -- ^ elems?
-> Natural -- ^ hash functions
-> Double -- ^ error probability
-> Natural
bloomFilterSize n k p
| p <= 0 || p >= 1 = 0
| otherwise = rnd $ negate (fromIntegral n * fromIntegral k) / log (1 - p ** (1 / fromIntegral k))
where
rnd x = 2 ** realToFrac (ceiling (logBase 2 x)) & round
writeReflogIndex :: forall m . ( Git3Perks m
, MonadReader Git3Env m
, HasClientAPI PeerAPI UNIX m
, HasClientAPI RefLogAPI UNIX m
, HasStorage m
) => m ()
writeReflogIndex = do
reflog <- getGitRemoteKey >>= orThrow Git3ReflogNotSet
api <- getClientAPI @RefLogAPI @UNIX
sto <- getStorage
idx <- openIndex
written_ <- newTVarIO mempty
(t1,_) <- timeItT do
enumEntries idx $ \bs -> do
let txh = coerce (BS.take 32 $ BS.drop 20 bs) :: HashRef
atomically $ modifyTVar written_ (HS.insert txh)
written <- readTVarIO written_
notice $ "read index at" <+> pretty (realToFrac @_ @(Fixed E2) t1)
flip runContT pure do
what' <- lift $ callRpcWaitMay @RpcRefLogGet (TimeoutSec 2) api reflog
>>= orThrow Git3RpcTimeout
what <- ContT $ maybe1 what' none
idxPath <- lift indexPath
mkdir idxPath
notice $ "STATE" <+> pretty idxPath
sink <- S.toList_ do
walkMerkle (coerce what) (getBlock sto) $ \case
Left{} -> throwIO MissedBlockError
Right (hs :: [HashRef]) -> do
for_ [h | h <- hs, not (HS.member h written)] $ \h -> void $ runMaybeT do
tx <- getBlock sto (coerce h)
>>= toMPlus
RefLogUpdate{..} <- deserialiseOrFail @(RefLogUpdate L4Proto) tx
& toMPlus
AnnotatedHashRef _ href <- deserialiseOrFail @AnnotatedHashRef (LBS.fromStrict _refLogUpdData)
& toMPlus
-- FIXME: error logging
lbs <- liftIO (runExceptT (getTreeContents sto href))
>>= orThrow MissedBlockError
pieces <- S.toList_ do
void $ runConsumeLBS (ZstdL.decompress lbs) $ readLogFileLBS () $ \o s _ -> do
lift $ S.yield o
lift $ S.yield (h, pieces)
liftIO $ forConcurrently_ sink $ \(tx, pieces) -> do
idxName <- emptyTempFile idxPath "objects-.idx"
let ss = L.sort pieces
UIO.withBinaryFileAtomic idxName WriteMode $ \wh -> do
for_ ss $ \sha1 -> do
let key = coerce @_ @N.ByteString sha1
let value = coerce @_ @N.ByteString tx
-- notice $ pretty sha1 <+> pretty tx
writeSection ( LBS.fromChunks [key,value] ) (LBS.hPutStr wh)
-- files <- lift listObjectIndexFiles
-- let num = sum (fmap snd files) `div` 56
-- let size = bloomFilterSize num 5 0.01
-- bloom <- liftIO $ stToIO (MBloom.new bloomHash (fromIntegral size))
-- lift $ enumEntries $ \bs -> do
-- liftIO $ stToIO $ MBloom.insert bloom (coerce bs)
-- let bloomIdxName = idxPath </> "filter"
-- bytes <- liftIO $ stToIO $ Bloom.freeze bloom
-- liftIO $ UIO.withBinaryFileAtomic bloomIdxName WriteMode $ \wh -> do
-- LBS.hPutStr wh "puk"
-- LBS.hPutStr wh (serialise bytes)
-- LBS.writeFile (serialise b
-- for_ ss $ \sha1 -> do
-- let key = coerce @_ @N.ByteString sha1
-- let value = coerce @_ @N.ByteString tx
-- -- notice $ pretty sha1 <+> pretty tx
-- writeSection ( LBS.fromChunks [key,value] ) (LBS.hPutStr wh)