Files
Quytelda Kahja e621f9a576 Don't add Zip64 extra fields when copying entries without them
When copying an entry from another archive, there is no reason to add
Zip64 extra fields if the source entry doesn't have them. This change
checks whether a copied entry had Zip64 extra fields before
appending/updating the field in the local header.
2025-06-02 16:49:33 +02:00

1246 lines
44 KiB
Haskell
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
{-# LANGUAGE CPP #-}
{-# LANGUAGE RecordWildCards #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE TupleSections #-}
-- |
-- Module : Codec.Archive.Zip.Internal
-- Copyright : © 2016present Mark Karpov
-- License : BSD 3 clause
--
-- Maintainer : Mark Karpov <markkarpov92@gmail.com>
-- Stability : experimental
-- Portability : portable
--
-- Low-level, non-public types and operations.
module Codec.Archive.Zip.Internal where
import Codec.Archive.Zip.CP437 (decodeCP437)
import Codec.Archive.Zip.Internal.Type
import Conduit (PrimMonad)
import Control.Applicative (many, (<|>))
import Control.Exception (bracketOnError, catchJust)
import Control.Monad
import Control.Monad.Catch (MonadThrow (..))
import Control.Monad.Trans.Maybe
import Control.Monad.Trans.Resource (MonadResource, ResourceT)
import Data.Bits
import Data.Bool (bool)
import Data.ByteString (ByteString)
import Data.ByteString qualified as B
import Data.Char (ord)
import Data.Conduit (ConduitT, ZipSink (..), (.|))
import Data.Conduit qualified as C
import Data.Conduit.Binary qualified as CB
import Data.Conduit.List qualified as CL
import Data.Conduit.Zlib qualified as Z
import Data.Digest.CRC32 (crc32Update)
import Data.Fixed (Fixed (..))
import Data.Map.Strict (Map)
import Data.Map.Strict qualified as M
import Data.Maybe (catMaybes, fromJust, isNothing)
import Data.Sequence (Seq, (><), (|>))
import Data.Sequence qualified as S
import Data.Serialize
import Data.Set qualified as E
import Data.Text (Text)
import Data.Text qualified as T
import Data.Text.Encoding qualified as T
import Data.Time
import Data.Version
import Data.Void
import Data.Word (Word16, Word32)
import Numeric.Natural (Natural)
import System.Directory
import System.FilePath
import System.IO
import System.IO.Error (isDoesNotExistError)
#ifndef mingw32_HOST_OS
import qualified Codec.Archive.Zip.Unix as Unix
#endif
#ifdef ENABLE_BZIP2
import qualified Data.Conduit.BZlib as BZ
#endif
#ifdef ENABLE_ZSTD
import qualified Data.Conduit.Zstd as Zstandard
#endif
#if !MIN_VERSION_base(4,20,0)
import Data.Foldable (foldl')
#endif
----------------------------------------------------------------------------
-- Data types
-- | The sum type describes all possible actions that can be performed on an
-- archive.
data PendingAction
= -- | Add an entry given its 'Source'
SinkEntry
CompressionMethod
(ConduitT () ByteString (ResourceT IO) ())
EntrySelector
| -- | Add an entry given its content as a string ByteString.
StrictEntry CompressionMethod ByteString EntrySelector
| -- | Copy an entry from another archive without re-compression
CopyEntry FilePath EntrySelector EntrySelector
| -- | Change the name of the entry inside archive
RenameEntry EntrySelector EntrySelector
| -- | Delete an entry from archive
DeleteEntry EntrySelector
| -- | Change the compression method on an entry
Recompress CompressionMethod EntrySelector
| -- | Set the comment for a particular entry
SetEntryComment Text EntrySelector
| -- | Delete theh comment of a particular entry
DeleteEntryComment EntrySelector
| -- | Set the modification time of a particular entry
SetModTime UTCTime EntrySelector
| -- | Add an extra field to the specified entry
AddExtraField Word16 ByteString EntrySelector
| -- | Delete an extra filed of the specified entry
DeleteExtraField Word16 EntrySelector
| -- | Set the comment for the entire archive
SetArchiveComment Text
| -- | Delete the comment of the entire archive
DeleteArchiveComment
| -- | Set an external file attribute for the specified entry
SetExternalFileAttributes Word32 EntrySelector
-- | A collection of maps describing how to produce entries in the resulting
-- archive.
data ProducingActions = ProducingActions
{ paCopyEntry :: Map FilePath (Map EntrySelector EntrySelector),
paSinkEntry :: Map EntrySelector (EntryOrigin, ConduitT () ByteString (ResourceT IO) ())
}
-- | A collection of editing actions, that is, actions that modify already
-- existing entries.
data EditingActions = EditingActions
{ eaCompression :: Map EntrySelector CompressionMethod,
eaEntryComment :: Map EntrySelector Text,
eaDeleteComment :: Map EntrySelector (),
eaModTime :: Map EntrySelector UTCTime,
eaExtraField :: Map EntrySelector (Map Word16 ByteString),
eaDeleteField :: Map EntrySelector (Map Word16 ()),
eaExtFileAttr :: Map EntrySelector Word32
}
-- | The origin of entries that can be streamed into archive.
data EntryOrigin
= GenericOrigin
| StrictOrigin Natural -- uncompressed length
| Borrowed EntryDescription
deriving (Eq)
-- | The type of the file header: local or central directory.
data HeaderType
= LocalHeader EntryOrigin
| CentralDirHeader
deriving (Eq)
-- | The data descriptor representation.
data DataDescriptor = DataDescriptor
{ ddCRC32 :: Word32,
ddCompressedSize :: Natural,
ddUncompressedSize :: Natural
}
-- | A temporary data structure to hold Zip64 extra data field information.
data Zip64ExtraField = Zip64ExtraField
{ z64efUncompressedSize :: Natural,
z64efCompressedSize :: Natural,
z64efOffset :: Natural
}
-- | MS-DOS date-time: a pair of 'Word16' (date, time) with the following
-- structure:
--
-- > DATE bit 0 - 4 5 - 8 9 - 15
-- > value day (1 - 31) month (1 - 12) years from 1980
-- > TIME bit 0 - 4 5 - 10 11 - 15
-- > value seconds* minute hour
-- > *stored in two-second increments
data MsDosTime = MsDosTime
{ msDosDate :: Word16,
msDosTime :: Word16
}
----------------------------------------------------------------------------
-- Constants
-- | “Version created by” to specify when writing archive data.
zipVersion :: Version
zipVersion = Version [6, 3] []
----------------------------------------------------------------------------
-- Higher-level operations
-- | Scan the central directory of an archive and return its description
-- 'ArchiveDescription' as well as a collection of its entries.
--
-- This operation may fail with:
--
-- * @isAlreadyInUseError@ if the file is already open and cannot be
-- reopened;
--
-- * @isDoesNotExistError@ if the file does not exist;
--
-- * @isPermissionError@ if the user does not have permission to open
-- the file;
--
-- * 'ParsingFailed' when specified archive is something this library
-- cannot parse (this includes multi-disk archives, for example).
--
-- Please note that entries with invalid (non-portable) file names may be
-- missing in the list of entries. Files that are compressed with
-- unsupported compression methods are skipped as well. Also, if several
-- entries would collide on some operating systems (such as Windows, because
-- of its case-insensitivity), only one of them will be available, because
-- 'EntrySelector' is case-insensitive. These are the consequences of the
-- design decision to make it impossible to create non-portable archives
-- with this library.
scanArchive ::
-- | Path to archive to scan
FilePath ->
IO (ArchiveDescription, Map EntrySelector EntryDescription)
scanArchive path = withBinaryFile path ReadMode $ \h -> do
mecdOffset <- locateECD path h
case mecdOffset of
Just ecdOffset -> do
hSeek h AbsoluteSeek ecdOffset
ecdSize <- subtract ecdOffset <$> hFileSize h
ecdRaw <- B.hGet h (fromIntegral ecdSize)
case runGet getECD ecdRaw of
Left msg -> throwM (ParsingFailed path msg)
Right ecd -> do
hSeek h AbsoluteSeek $ fromIntegral (adCDOffset ecd)
cdRaw <- B.hGet h $ fromIntegral (adCDSize ecd)
case runGet getCD cdRaw of
Left msg -> throwM (ParsingFailed path msg)
Right cd -> return (ecd, cd)
Nothing ->
throwM (ParsingFailed path "Cannot locate end of central directory")
-- | Given location of the archive and information about a specific archive
-- entry 'EntryDescription', return 'Source' of its data. The actual data
-- can be compressed or uncompressed depending on the third argument.
sourceEntry ::
(PrimMonad m, MonadThrow m, MonadResource m) =>
-- | Path to archive that contains the entry
FilePath ->
-- | Information needed to extract entry of interest
EntryDescription ->
-- | Should we stream uncompressed data?
Bool ->
-- | Source of uncompressed data
ConduitT () ByteString m ()
sourceEntry path EntryDescription {..} d =
source .| CB.isolate (fromIntegral edCompressedSize) .| decompress
where
source = CB.sourceIOHandle $ do
h <- openFile path ReadMode
hSeek h AbsoluteSeek (fromIntegral edOffset)
localHeader <- B.hGet h 30
case runGet getLocalHeaderGap localHeader of
Left msg -> throwM (ParsingFailed path msg)
Right gap -> do
hSeek h RelativeSeek gap
return h
decompress =
if d
then decompressingPipe edCompression
else C.awaitForever C.yield
-- | Undertake /all/ actions specified as the fourth argument of the
-- function. This transforms the given pending actions so they can be
-- performed in one pass, and then they are applied in the most efficient
-- way.
commit ::
-- | Location of archive file to edit or create
FilePath ->
-- | Archive description
ArchiveDescription ->
-- | Current list of entries
Map EntrySelector EntryDescription ->
-- | Collection of pending actions
Seq PendingAction ->
IO ()
commit path ArchiveDescription {..} entries xs =
withNewFile path $ \h -> do
let (ProducingActions copying sinking, editing) =
optimize (toRecreatingActions path entries >< xs)
comment = predictComment adComment xs
copiedCD <-
M.unions
<$> M.traverseWithKey
( \srcPath m ->
copyEntries h srcPath m editing
)
copying
sunkCD <-
M.traverseWithKey
( \selector (origin, source) ->
sinkEntry h selector origin source editing
)
(sinking `M.difference` copiedCD)
writeCD h comment (copiedCD `M.union` sunkCD)
-- | Create a new file with the guarantee that in the case of an exception
-- the old file will be intact. The file is only updated\/replaced if the
-- second argument finishes without exceptions.
withNewFile ::
-- | Name of file to create
FilePath ->
-- | Action that writes to given 'Handle'
(Handle -> IO ()) ->
IO ()
withNewFile fpath action =
bracketOnError allocate release $ \(path, h) -> do
action h
hClose h
renameFile path fpath
where
allocate = openBinaryTempFile (takeDirectory fpath) ".zip"
release (path, h) = do
hClose h
-- Despite using `bracketOnError` the file is not guaranteed to exist
-- here since we could be interrupted with an async exception after
-- the file has been renamed. Therefore, we silentely ignore
-- `DoesNotExistError`.
catchJust (guard . isDoesNotExistError) (removeFile path) (const $ pure ())
-- | Determine what comment in new archive will look like given its original
-- value and a collection of pending actions.
predictComment :: Maybe Text -> Seq PendingAction -> Maybe Text
predictComment original xs =
case S.index xs <$> S.findIndexR (isNothing . targetEntry) xs of
Nothing -> original
Just DeleteArchiveComment -> Nothing
Just (SetArchiveComment txt) -> Just txt
Just _ -> Nothing
-- | Transform a map representing existing entries into a collection of
-- actions that re-create those entires.
toRecreatingActions ::
-- | Name of the archive file where entires are found
FilePath ->
-- | Actual list of entires
Map EntrySelector EntryDescription ->
-- | Actions that recreate the archive entries
Seq PendingAction
toRecreatingActions path entries = E.foldl' f S.empty (M.keysSet entries)
where
f s e = s |> CopyEntry path e e
-- | Transform a collection of 'PendingAction's into 'ProducingActions' and
-- 'EditingActions'—data that describes how to create resulting archive.
optimize ::
-- | Collection of pending actions
Seq PendingAction ->
-- | Optimized data
(ProducingActions, EditingActions)
optimize =
foldl'
f
( ProducingActions M.empty M.empty,
EditingActions M.empty M.empty M.empty M.empty M.empty M.empty M.empty
)
where
f (pa, ea) a = case a of
SinkEntry m src s ->
( pa
{ paSinkEntry = M.insert s (GenericOrigin, src) (paSinkEntry pa),
paCopyEntry = M.map (M.filter (/= s)) (paCopyEntry pa)
},
(clearEditingFor s ea)
{ eaCompression = M.insert s m (eaCompression ea)
}
)
StrictEntry m bs s ->
( pa
{ paSinkEntry = M.insert s (origin, C.yield bs) (paSinkEntry pa),
paCopyEntry = M.map (M.filter (/= s)) (paCopyEntry pa)
},
(clearEditingFor s ea)
{ eaCompression = M.insert s m (eaCompression ea)
}
)
where
origin = StrictOrigin (fromIntegral $ B.length bs)
CopyEntry path os ns ->
( pa
{ paSinkEntry = M.delete ns (paSinkEntry pa),
paCopyEntry = M.alter (ef os ns) path (paCopyEntry pa)
},
clearEditingFor ns ea
)
RenameEntry os ns ->
( pa
{ paCopyEntry = M.map (M.map $ re os ns) (paCopyEntry pa),
paSinkEntry = renameKey os ns (paSinkEntry pa)
},
ea
{ eaCompression = renameKey os ns (eaCompression ea),
eaEntryComment = renameKey os ns (eaEntryComment ea),
eaDeleteComment = renameKey os ns (eaDeleteComment ea),
eaModTime = renameKey os ns (eaModTime ea),
eaExtraField = renameKey os ns (eaExtraField ea),
eaDeleteField = renameKey os ns (eaDeleteField ea)
}
)
DeleteEntry s ->
( pa
{ paSinkEntry = M.delete s (paSinkEntry pa),
paCopyEntry = M.map (M.delete s) (paCopyEntry pa)
},
clearEditingFor s ea
)
Recompress m s ->
(pa, ea {eaCompression = M.insert s m (eaCompression ea)})
SetEntryComment txt s ->
( pa,
ea
{ eaEntryComment = M.insert s txt (eaEntryComment ea),
eaDeleteComment = M.delete s (eaDeleteComment ea)
}
)
DeleteEntryComment s ->
( pa,
ea
{ eaEntryComment = M.delete s (eaEntryComment ea),
eaDeleteComment = M.insert s () (eaDeleteComment ea)
}
)
SetModTime time s ->
(pa, ea {eaModTime = M.insert s time (eaModTime ea)})
AddExtraField n b s ->
( pa,
ea
{ eaExtraField = M.alter (ef n b) s (eaExtraField ea),
eaDeleteField = M.delete s (eaDeleteField ea)
}
)
DeleteExtraField n s ->
( pa,
ea
{ eaExtraField = M.alter (er n) s (eaExtraField ea),
eaDeleteField = M.alter (ef n ()) s (eaDeleteField ea)
}
)
SetExternalFileAttributes b s ->
( pa,
ea {eaExtFileAttr = M.insert s b (eaExtFileAttr ea)}
)
_ -> (pa, ea)
clearEditingFor s ea =
ea
{ eaCompression = M.delete s (eaCompression ea),
eaEntryComment = M.delete s (eaEntryComment ea),
eaDeleteComment = M.delete s (eaDeleteComment ea),
eaModTime = M.delete s (eaModTime ea),
eaExtraField = M.delete s (eaExtraField ea),
eaDeleteField = M.delete s (eaDeleteField ea),
eaExtFileAttr = M.delete s (eaExtFileAttr ea)
}
re o n x = if x == o then n else x
ef k v (Just m) = Just (M.insert k v m)
ef k v Nothing = Just (M.singleton k v)
er k (Just m) =
let n = M.delete k m
in if M.null n then Nothing else Just n
er _ Nothing = Nothing
-- | Copy entries from another archive and write them into the file
-- associated with the given handle. This can throw 'EntryDoesNotExist' if
-- there is no such entry in that archive.
copyEntries ::
-- | Opened 'Handle' of zip archive file
Handle ->
-- | Path to the file to copy the entries from
FilePath ->
-- | 'Map' from original name to name to use in new archive
Map EntrySelector EntrySelector ->
-- | Additional info that can influence result
EditingActions ->
-- | Info to generate central directory file headers later
IO (Map EntrySelector EntryDescription)
copyEntries h path names editing = do
entries <- snd <$> scanArchive path
M.foldlWithKey
( \acc oldName newName ->
case M.lookup oldName entries of
Nothing -> throwM (EntryDoesNotExist path oldName)
Just oldDesc ->
M.insert newName
<$> sinkEntry h newName (Borrowed oldDesc) src editing
<*> acc
where
src = sourceEntry path oldDesc False
)
mempty
names
-- | Sink an entry from the given stream into the file associated with the
-- given 'Handle'.
sinkEntry ::
-- | Opened 'Handle' of zip archive file
Handle ->
-- | Name of the entry to add
EntrySelector ->
-- | Origin of the entry (can contain additional info)
EntryOrigin ->
-- | Source of the entry contents
ConduitT () ByteString (ResourceT IO) () ->
-- | Additional info that can influence result
EditingActions ->
-- | Info to generate the central directory file headers later
IO EntryDescription
sinkEntry h s o src EditingActions {..} = do
currentTime <- getCurrentTime
offset <- hTell h
let compressed = case o of
Borrowed ed -> edCompression ed
_ -> Store
compression = M.findWithDefault compressed s eaCompression
recompression = compression /= compressed
modTime = case o of
Borrowed ed -> edModTime ed
_ -> currentTime
extFileAttr = M.findWithDefault defaultFileMode s eaExtFileAttr
oldExtraFields = case o of
Borrowed ed -> edExtraField ed
_ -> M.empty
extraField =
(M.findWithDefault M.empty s eaExtraField `M.union` oldExtraFields)
`M.difference` M.findWithDefault M.empty s eaDeleteField
oldComment = case (o, M.lookup s eaDeleteComment) of
(Borrowed ed, Nothing) -> edComment ed
_ -> Nothing
desc0 =
EntryDescription -- to write in local header
{ edVersionMadeBy = zipVersion,
edVersionNeeded = zipVersion,
edCompression = compression,
edModTime = M.findWithDefault modTime s eaModTime,
edCRC32 = 0, -- to be overwritten after streaming
edCompressedSize = 0, -- ↑
edUncompressedSize = 0, -- ↑
edOffset = fromIntegral offset,
edComment = M.lookup s eaEntryComment <|> oldComment,
edExtraField = extraField,
edExternalFileAttrs = extFileAttr
}
B.hPut h (runPut (putHeader (LocalHeader o) s desc0))
DataDescriptor {..} <-
C.runConduitRes $
if recompression
then
if compressed == Store
then src .| sinkData h compression
else src .| decompressingPipe compressed .| sinkData h compression
else src .| sinkData h Store
afterStreaming <- hTell h
let desc1 = case o of
Borrowed ed ->
desc0
{ edCRC32 =
bool (edCRC32 ed) ddCRC32 recompression,
edCompressedSize =
bool (edCompressedSize ed) ddCompressedSize recompression,
edUncompressedSize =
bool (edUncompressedSize ed) ddUncompressedSize recompression
}
_ ->
desc0
{ edCRC32 = ddCRC32,
edCompressedSize = ddCompressedSize,
edUncompressedSize = ddUncompressedSize
}
desc2 =
desc1
{ edVersionNeeded =
getZipVersion (needsZip64 desc1) (Just compression)
}
hSeek h AbsoluteSeek offset
B.hPut h (runPut (putHeader (LocalHeader o) s desc2))
hSeek h AbsoluteSeek afterStreaming
return desc2
{- ORMOLU_DISABLE -}
-- | Create a 'Sink' to stream data there. Once streaming is finished,
-- return 'DataDescriptor' for the streamed data. The action /does not/
-- close the given 'Handle'.
sinkData ::
-- | Opened 'Handle' of zip archive file
Handle ->
-- | Compression method to apply
CompressionMethod ->
-- | 'Sink' where to stream data
ConduitT ByteString Void (ResourceT IO) DataDescriptor
sinkData h compression = do
let sizeSink = CL.fold (\acc input -> fromIntegral (B.length input) + acc) 0
dataSink =
getZipSink $
ZipSink sizeSink <* ZipSink (CB.sinkHandle h)
withCompression sink =
getZipSink $
(,,) <$> ZipSink sizeSink
<*> ZipSink crc32Sink
<*> ZipSink sink
(uncompressedSize, crc32, compressedSize) <-
case compression of
Store ->
withCompression
dataSink
Deflate ->
withCompression $
Z.compress 9 (Z.WindowBits (-15)) .| dataSink
#ifdef ENABLE_BZIP2
BZip2 ->
withCompression $
BZ.bzip2 .| dataSink
#else
BZip2 -> throwM (UnsupportedCompressionMethod BZip2)
#endif
#ifdef ENABLE_ZSTD
Zstd ->
withCompression $
Zstandard.compress 1 .| dataSink
#else
Zstd -> throwM (UnsupportedCompressionMethod Zstd)
#endif
return
DataDescriptor
{ ddCRC32 = fromIntegral crc32,
ddCompressedSize = compressedSize,
ddUncompressedSize = uncompressedSize
}
{- ORMOLU_ENABLE -}
-- | Append central directory entries and the end of central directory
-- record to the file that given 'Handle' is associated with. Note that this
-- automatically writes Zip64 end of central directory record and Zip64 end
-- of central directory locator when necessary.
writeCD ::
-- | Opened handle of zip archive file
Handle ->
-- | Commentary to the entire archive
Maybe Text ->
-- | Info about already written local headers and entry data
Map EntrySelector EntryDescription ->
IO ()
writeCD h comment m = do
let cd = runPut (putCD m)
cdOffset <- fromIntegral <$> hTell h
B.hPut h cd -- write central directory
let totalCount = fromIntegral (M.size m)
cdSize = fromIntegral (B.length cd)
needZip64 =
totalCount >= ffff
|| cdSize >= ffffffff
|| cdOffset >= ffffffff
when needZip64 $ do
zip64ecdOffset <- fromIntegral <$> hTell h
(B.hPut h . runPut) (putZip64ECD totalCount cdSize cdOffset)
(B.hPut h . runPut) (putZip64ECDLocator zip64ecdOffset)
(B.hPut h . runPut) (putECD totalCount cdSize cdOffset comment)
----------------------------------------------------------------------------
-- Binary serialization
-- | Extract the number of bytes between the start of file name in local
-- header and the start of actual data.
getLocalHeaderGap :: Get Integer
getLocalHeaderGap = do
getSignature 0x04034b50
skip 2 -- version needed to extract
skip 2 -- general purpose bit flag
skip 2 -- compression method
skip 2 -- last mod file time
skip 2 -- last mod file date
skip 4 -- crc-32 check sum
skip 4 -- compressed size
skip 4 -- uncompressed size
fileNameSize <- fromIntegral <$> getWord16le -- file name length
extraFieldSize <- fromIntegral <$> getWord16le -- extra field length
return (fileNameSize + extraFieldSize)
-- | Parse central directory file headers and put them into a 'Map'.
getCD :: Get (Map EntrySelector EntryDescription)
getCD = M.fromList . catMaybes <$> many getCDHeader
-- | Parse a single central directory file header. If it's a directory or
-- file compressed with unsupported compression method, 'Nothing' is
-- returned.
getCDHeader :: Get (Maybe (EntrySelector, EntryDescription))
getCDHeader = do
getSignature 0x02014b50 -- central file header signature
versionMadeBy <- toVersion <$> getWord16le -- version made by
versionNeeded <- toVersion <$> getWord16le -- version needed to extract
when (versionNeeded > zipVersion) . fail $
"Version required to extract the archive is "
++ showVersion versionNeeded
++ " (can do "
++ showVersion zipVersion
++ ")"
bitFlag <- getWord16le -- general purpose bit flag
when (any (testBit bitFlag) [0, 6, 13]) . fail $
"Encrypted archives are not supported"
let needUnicode = testBit bitFlag 11
mcompression <- toCompressionMethod <$> getWord16le -- compression method
modTime <- getWord16le -- last mod file time
modDate <- getWord16le -- last mod file date
crc32 <- getWord32le -- CRC32 check sum
compressed <- fromIntegral <$> getWord32le -- compressed size
uncompressed <- fromIntegral <$> getWord32le -- uncompressed size
fileNameSize <- getWord16le -- file name length
extraFieldSize <- getWord16le -- extra field length
commentSize <- getWord16le -- file comment size
skip 4 -- disk number start, internal file attributes
externalFileAttrs <- getWord32le -- external file attributes
offset <- fromIntegral <$> getWord32le -- offset of local header
fileName <-
decodeText needUnicode
<$> getBytes (fromIntegral fileNameSize) -- file name
extraField <-
M.fromList
<$> isolate (fromIntegral extraFieldSize) (many getExtraField)
-- ↑ extra fields in their raw form
comment <- decodeText needUnicode <$> getBytes (fromIntegral commentSize)
-- ↑ file comment
let dfltZip64 =
Zip64ExtraField
{ z64efUncompressedSize = uncompressed,
z64efCompressedSize = compressed,
z64efOffset = offset
}
z64ef = case M.lookup 1 extraField of
Nothing -> dfltZip64
Just b -> parseZip64ExtraField dfltZip64 b
case mcompression of
Nothing -> return Nothing
Just compression ->
let desc =
EntryDescription
{ edVersionMadeBy = versionMadeBy,
edVersionNeeded = versionNeeded,
edCompression = compression,
edModTime = fromMsDosTime (MsDosTime modDate modTime),
edCRC32 = crc32,
edCompressedSize = z64efCompressedSize z64ef,
edUncompressedSize = z64efUncompressedSize z64ef,
edOffset = z64efOffset z64ef,
edComment = if commentSize == 0 then Nothing else comment,
edExtraField = extraField,
edExternalFileAttrs = externalFileAttrs
}
in return $ (,desc) <$> (fileName >>= mkEntrySelector . T.unpack)
-- | Parse an extra-field.
getExtraField :: Get (Word16, ByteString)
getExtraField = do
header <- getWord16le -- header id
size <- getWord16le -- data size
body <- getBytes (fromIntegral size) -- content
return (header, body)
-- | Get signature. If the extracted data is not equal to the provided
-- signature, fail.
getSignature :: Word32 -> Get ()
getSignature sig = do
x <- getWord32le -- grab 4-byte signature
unless (x == sig) . fail $
"Expected signature " ++ show sig ++ ", but got: " ++ show x
-- | Parse 'Zip64ExtraField' from its binary representation.
parseZip64ExtraField ::
-- | What is read from central directory file header
Zip64ExtraField ->
-- | Actual binary representation
ByteString ->
-- | Result
Zip64ExtraField
parseZip64ExtraField dflt@Zip64ExtraField {..} b =
either (const dflt) id . flip runGet b $ do
let ifsat v =
if v >= ffffffff
then fromIntegral <$> getWord64le
else return v
uncompressed <- ifsat z64efUncompressedSize -- uncompressed size
compressed <- ifsat z64efCompressedSize -- compressed size
offset <- ifsat z64efOffset -- offset of local file header
return (Zip64ExtraField uncompressed compressed offset)
-- | Produce binary representation of 'Zip64ExtraField'.
makeZip64ExtraField ::
-- | Is this for local or central directory header?
HeaderType ->
-- | Zip64 extra field's data
Zip64ExtraField ->
-- | Resulting representation
ByteString
makeZip64ExtraField headerType Zip64ExtraField {..} = runPut $ do
case headerType of
LocalHeader _ -> do
putWord64le (fromIntegral z64efUncompressedSize) -- uncompressed size
putWord64le (fromIntegral z64efCompressedSize) -- compressed size
CentralDirHeader -> do
when (z64efUncompressedSize >= ffffffff) $
putWord64le (fromIntegral z64efUncompressedSize) -- uncompressed size
when (z64efCompressedSize >= ffffffff) $
putWord64le (fromIntegral z64efCompressedSize) -- compressed size
when (z64efOffset >= ffffffff) $
putWord64le (fromIntegral z64efOffset) -- offset of local file header
-- | Create 'ByteString' representing an extra field.
putExtraField :: Map Word16 ByteString -> Put
putExtraField = M.foldMapWithKey $ \headerId bs -> do
let b = B.take 0xffff bs
putWord16le headerId
putWord16le (fromIntegral $ B.length b)
putByteString b
-- | Create 'ByteString' representing the entire central directory.
putCD :: Map EntrySelector EntryDescription -> Put
putCD = M.foldMapWithKey (putHeader CentralDirHeader)
-- | Create 'ByteString' representing either a local file header or a
-- central directory file header.
putHeader ::
-- | Type of header to generate
HeaderType ->
-- | Name of entry to write
EntrySelector ->
-- | Description of entry
EntryDescription ->
Put
putHeader headerType s entry@EntryDescription {..} = do
let isCentralDirHeader = headerType == CentralDirHeader
putWord32le (bool 0x04034b50 0x02014b50 isCentralDirHeader)
-- ↑ local/central file header signature
when isCentralDirHeader $
putWord16le (fromVersion edVersionMadeBy) -- version made by
putWord16le (fromVersion edVersionNeeded) -- version needed to extract
let entryName = getEntryName s
rawName = T.encodeUtf8 entryName
comment = B.take 0xffff (maybe B.empty T.encodeUtf8 edComment)
unicode =
needsUnicode entryName
|| maybe False needsUnicode edComment
modTime = toMsDosTime edModTime
putWord16le (if unicode then setBit 0 11 else 0)
-- ↑ general purpose bit-flag
putWord16le (fromCompressionMethod edCompression) -- compression method
putWord16le (msDosTime modTime) -- last mod file time
putWord16le (msDosDate modTime) -- last mod file date
putWord32le edCRC32 -- CRC-32 checksum
putWord32le (withSaturation edCompressedSize) -- compressed size
putWord32le (withSaturation edUncompressedSize) -- uncompressed size
putWord16le (fromIntegral $ B.length rawName) -- file name length
let zip64ef =
makeZip64ExtraField
headerType
Zip64ExtraField
{ z64efUncompressedSize = edUncompressedSize,
z64efCompressedSize = edCompressedSize,
z64efOffset = edOffset
}
appendZip64 =
case headerType of
LocalHeader (StrictOrigin size) -> size >= ffffffff
LocalHeader (Borrowed ed) -> entryUsesZip64 ed
LocalHeader GenericOrigin -> True
CentralDirHeader -> needsZip64 entry
extraField =
B.take 0xffff . runPut . putExtraField $
if appendZip64
then M.insert 1 zip64ef edExtraField
else edExtraField
putWord16le (fromIntegral $ B.length extraField) -- extra field length
when isCentralDirHeader $ do
putWord16le (fromIntegral $ B.length comment) -- file comment length
putWord16le 0 -- disk number start
putWord16le 0 -- internal file attributes
putWord32le edExternalFileAttrs -- external file attributes
putWord32le (withSaturation edOffset) -- relative offset of local header
putByteString rawName -- file name (variable size)
putByteString extraField -- extra field (variable size)
when isCentralDirHeader (putByteString comment) -- file comment (variable size)
-- | Create 'ByteString' representing Zip64 end of central directory record.
putZip64ECD ::
-- | Total number of entries
Natural ->
-- | Size of the central directory
Natural ->
-- | Offset of central directory record
Natural ->
Put
putZip64ECD totalCount cdSize cdOffset = do
putWord32le 0x06064b50 -- zip64 end of central dir signature
putWord64le 44 -- size of zip64 end of central dir record
putWord16le (fromVersion zipVersion) -- version made by
putWord16le (fromVersion $ getZipVersion True Nothing)
-- ↑ version needed to extract
putWord32le 0 -- number of this disk
putWord32le 0 -- number of the disk with the start of the central directory
putWord64le (fromIntegral totalCount) -- total number of entries (this disk)
putWord64le (fromIntegral totalCount) -- total number of entries
putWord64le (fromIntegral cdSize) -- size of the central directory
putWord64le (fromIntegral cdOffset) -- offset of central directory
-- | Create 'ByteString' representing Zip64 end of the central directory
-- locator.
putZip64ECDLocator ::
-- | Offset of Zip64 end of central directory
Natural ->
Put
putZip64ECDLocator ecdOffset = do
putWord32le 0x07064b50 -- zip64 end of central dir locator signature
putWord32le 0 -- number of the disk with the start of the zip64 end of
-- central directory
putWord64le (fromIntegral ecdOffset) -- relative offset of the zip64 end
-- of central directory record
putWord32le 1 -- total number of disks
-- | Parse end of the central directory record or Zip64 end of the central
-- directory record depending on signature binary data begins with.
getECD :: Get ArchiveDescription
getECD = do
sig <- getWord32le -- end of central directory signature
let zip64 = sig == 0x06064b50
unless (sig == 0x06054b50 || sig == 0x06064b50) $
fail "Cannot locate end of central directory"
zip64size <-
if zip64
then do
x <- getWord64le -- size of zip64 end of central directory record
skip 2 -- version made by
skip 2 -- version needed to extract
return (Just x)
else return Nothing
thisDisk <- bool (fromIntegral <$> getWord16le) getWord32le zip64
-- ↑ number of this disk
cdDisk <- bool (fromIntegral <$> getWord16le) getWord32le zip64
-- ↑ number of the disk with the start of the central directory
unless (thisDisk == 0 && cdDisk == 0) $
fail "No support for multi-disk archives"
skip (bool 2 8 zip64)
-- ↑ total number of entries in the central directory on this disk
skip (bool 2 8 zip64)
-- ↑ total number of entries in the central directory
cdSize <- bool (fromIntegral <$> getWord32le) getWord64le zip64
-- ↑ size of the central directory
cdOffset <- bool (fromIntegral <$> getWord32le) getWord64le zip64
-- ↑ offset of start of central directory with respect to the starting
-- disk number
when zip64 . skip . fromIntegral $ fromJust zip64size - 4 -- obviously
commentSize <- getWord16le -- .ZIP file comment length
comment <- decodeText True <$> getBytes (fromIntegral commentSize)
-- ↑ archive comment, it's uncertain how we should decide on encoding here
return
ArchiveDescription
{ adComment = if commentSize == 0 then Nothing else comment,
adCDOffset = fromIntegral cdOffset,
adCDSize = fromIntegral cdSize
}
-- | Create a 'ByteString' representing the end of central directory record.
putECD ::
-- | Total number of entries
Natural ->
-- | Size of the central directory
Natural ->
-- | Offset of central directory record
Natural ->
-- | Zip file comment
Maybe Text ->
Put
putECD totalCount cdSize cdOffset mcomment = do
putWord32le 0x06054b50 -- end of central dir signature
putWord16le 0 -- number of this disk
putWord16le 0 -- number of the disk with the start of the central directory
putWord16le (withSaturation totalCount)
-- ↑ total number of entries on this disk
putWord16le (withSaturation totalCount) -- total number of entries
putWord32le (withSaturation cdSize) -- size of central directory
putWord32le (withSaturation cdOffset) -- offset of start of central directory
let comment = maybe B.empty T.encodeUtf8 mcomment
putWord16le (fromIntegral $ B.length comment)
putByteString comment
-- | Find the absolute offset of the end of central directory record or, if
-- present, Zip64 end of central directory record.
locateECD :: FilePath -> Handle -> IO (Maybe Integer)
locateECD path h = sizeCheck
where
sizeCheck = do
fsize <- hFileSize h
let limit = max 0 (fsize - 0xffff - 22)
if fsize < 22
then return Nothing
else hSeek h SeekFromEnd (-22) >> loop limit
loop limit = do
sig <- getNum getWord32le 4
pos <- subtract 4 <$> hTell h
let again = hSeek h AbsoluteSeek (pos - 1) >> loop limit
done = pos <= limit
if sig == 0x06054b50
then do
result <-
runMaybeT $
MaybeT (checkComment pos)
>>= MaybeT . checkCDSig
>>= MaybeT . checkZip64
case result of
Nothing -> bool again (return Nothing) done
Just ecd -> return (Just ecd)
else bool again (return Nothing) done
checkComment pos = do
size <- hFileSize h
hSeek h AbsoluteSeek (pos + 20)
l <- fromIntegral <$> getNum getWord16le 2
return $
if l + 22 == size - pos
then Just pos
else Nothing
checkCDSig pos = do
hSeek h AbsoluteSeek (pos + 16)
sigPos <- fromIntegral <$> getNum getWord32le 4
if sigPos == 0xffffffff -- Zip64 is probably used
then return (Just pos)
else do
hSeek h AbsoluteSeek sigPos
cdSig <- getNum getWord32le 4
return $
if cdSig == 0x02014b50
||
-- ↑ normal case: central directory file header signature
cdSig == 0x06064b50
||
-- ↑ happens when zip 64 archive is empty
cdSig == 0x06054b50
then -- ↑ happens when vanilla archive is empty
Just pos
else Nothing
checkZip64 pos =
if pos < 20
then return (Just pos)
else do
hSeek h AbsoluteSeek (pos - 20)
zip64locatorSig <- getNum getWord32le 4
if zip64locatorSig == 0x07064b50
then do
hSeek h AbsoluteSeek (pos - 12)
Just . fromIntegral <$> getNum getWord64le 8
else return (Just pos)
getNum f n = do
result <- runGet f <$> B.hGet h n
case result of
Left msg -> throwM (ParsingFailed path msg)
Right val -> return val
----------------------------------------------------------------------------
-- Helpers
-- | Rename an entry (key) in a 'Map'.
renameKey :: (Ord k) => k -> k -> Map k a -> Map k a
renameKey ok nk m = case M.lookup ok m of
Nothing -> m
Just e -> M.insert nk e (M.delete ok m)
-- | Like 'fromIntegral', but with saturation when converting to bounded
-- types.
withSaturation :: forall a b. (Integral a, Integral b, Bounded b) => a -> b
withSaturation x =
if (fromIntegral x :: Integer) > (fromIntegral bound :: Integer)
then bound
else fromIntegral x
where
bound = maxBound :: b
-- | Determine the target entry of an action.
targetEntry :: PendingAction -> Maybe EntrySelector
targetEntry (SinkEntry _ _ s) = Just s
targetEntry (StrictEntry _ _ s) = Just s
targetEntry (CopyEntry _ _ s) = Just s
targetEntry (RenameEntry s _) = Just s
targetEntry (DeleteEntry s) = Just s
targetEntry (Recompress _ s) = Just s
targetEntry (SetEntryComment _ s) = Just s
targetEntry (DeleteEntryComment s) = Just s
targetEntry (SetModTime _ s) = Just s
targetEntry (AddExtraField _ _ s) = Just s
targetEntry (DeleteExtraField _ s) = Just s
targetEntry (SetExternalFileAttributes _ s) = Just s
targetEntry (SetArchiveComment _) = Nothing
targetEntry DeleteArchiveComment = Nothing
-- | Decode a 'ByteString'. The first argument indicates whether we should
-- treat it as UTF-8 (in case bit 11 of general-purpose bit flag is set),
-- otherwise the function assumes CP437. Note that since not every stream of
-- bytes constitutes valid UTF-8 text, this function can fail. In that case
-- 'Nothing' is returned.
decodeText ::
-- | Whether bit 11 of general-purpose bit flag is set
Bool ->
-- | Binary data to decode
ByteString ->
-- | Decoded 'Text' in case of success
Maybe Text
decodeText False = Just . decodeCP437
decodeText True = either (const Nothing) Just . T.decodeUtf8'
-- | Detect if the given text needs newer Unicode-aware features to be
-- properly encoded in the archive.
needsUnicode :: Text -> Bool
needsUnicode = not . T.all validCP437
where
validCP437 x = ord x <= 127
-- | Convert numeric representation (as per the .ZIP specification) of
-- version into 'Version'.
toVersion :: Word16 -> Version
toVersion x = makeVersion [major, minor]
where
(major, minor) = quotRem (fromIntegral $ x .&. 0x00ff) 10
-- | Covert 'Version' to its numeric representation as per the .ZIP
-- specification.
fromVersion :: Version -> Word16
fromVersion v = fromIntegral ((ZIP_OS `shiftL` 8) .|. (major * 10 + minor))
where
(major, minor) =
case versionBranch v of
v0 : v1 : _ -> (v0, v1)
v0 : _ -> (v0, 0)
[] -> (0, 0)
-- | Get the compression method form its numeric representation.
toCompressionMethod :: Word16 -> Maybe CompressionMethod
toCompressionMethod 0 = Just Store
toCompressionMethod 8 = Just Deflate
toCompressionMethod 12 = Just BZip2
toCompressionMethod 93 = Just Zstd
toCompressionMethod _ = Nothing
-- | Convert 'CompressionMethod' to its numeric representation as per the
-- .ZIP specification.
fromCompressionMethod :: CompressionMethod -> Word16
fromCompressionMethod Store = 0
fromCompressionMethod Deflate = 8
fromCompressionMethod BZip2 = 12
fromCompressionMethod Zstd = 93
-- | Check if an entry with these parameters needs the Zip64 extension.
needsZip64 :: EntryDescription -> Bool
needsZip64 EntryDescription {..} =
any
(>= ffffffff)
[edOffset, edCompressedSize, edUncompressedSize]
-- | Check if an entry has Zip64 extra fields.
entryUsesZip64 :: EntryDescription -> Bool
entryUsesZip64 EntryDescription {..} = M.member 1 edExtraField
-- | Determine “version needed to extract” that should be written to the
-- headers given the need of the Zip64 feature and the compression method.
getZipVersion :: Bool -> Maybe CompressionMethod -> Version
getZipVersion zip64 m = max zip64ver mver
where
zip64ver = makeVersion (if zip64 then [4, 5] else [2, 0])
mver = makeVersion $ case m of
Nothing -> [2, 0]
Just Store -> [2, 0]
Just Deflate -> [2, 0]
Just BZip2 -> [4, 6]
Just Zstd -> [6, 3]
-- | Return a decompressing 'Conduit' corresponding to the given compression
-- method.
decompressingPipe ::
(PrimMonad m, MonadThrow m, MonadResource m) =>
CompressionMethod ->
ConduitT ByteString ByteString m ()
decompressingPipe Store = C.awaitForever C.yield
decompressingPipe Deflate = Z.decompress $ Z.WindowBits (-15)
#ifdef ENABLE_BZIP2
decompressingPipe BZip2 = BZ.bunzip2
#else
decompressingPipe BZip2 = throwM (UnsupportedCompressionMethod BZip2)
#endif
#ifdef ENABLE_ZSTD
decompressingPipe Zstd = Zstandard.decompress
#else
decompressingPipe Zstd = throwM (UnsupportedCompressionMethod Zstd)
#endif
-- | A sink that calculates the CRC32 check sum for an incoming stream.
crc32Sink :: ConduitT ByteString Void (ResourceT IO) Word32
crc32Sink = CL.fold crc32Update 0
-- | Convert 'UTCTime' to the MS-DOS time format.
toMsDosTime :: UTCTime -> MsDosTime
toMsDosTime UTCTime {..} = MsDosTime dosDate dosTime
where
dosTime = fromIntegral (seconds + shiftL minutes 5 + shiftL hours 11)
dosDate = fromIntegral (day + shiftL month 5 + shiftL year 9)
seconds =
let (MkFixed x) = todSec tod
in fromIntegral (x `quot` 2000000000000)
minutes = todMin tod
hours = todHour tod
tod = timeToTimeOfDay utctDayTime
year = fromIntegral year' - 1980
(year', month, day) = toGregorian utctDay
-- | Convert MS-DOS date-time to 'UTCTime'.
fromMsDosTime :: MsDosTime -> UTCTime
fromMsDosTime MsDosTime {..} =
UTCTime
(fromGregorian year month day)
(secondsToDiffTime $ hours * 3600 + minutes * 60 + seconds)
where
seconds = fromIntegral $ 2 * (msDosTime .&. 0x1f)
minutes = fromIntegral (shiftR msDosTime 5 .&. 0x3f)
hours = fromIntegral (shiftR msDosTime 11 .&. 0x1f)
day = fromIntegral (msDosDate .&. 0x1f)
month = fromIntegral $ shiftR msDosDate 5 .&. 0x0f
year = 1980 + fromIntegral (shiftR msDosDate 9)
-- We use the constants of the type 'Natural' instead of literals to protect
-- ourselves from overflows on 32 bit systems.
--
-- If we're in the development mode, use lower values so the tests get a
-- chance to check all cases (otherwise we would need to generate way too
-- big archives on CI).
ffff, ffffffff :: Natural
#ifdef HASKELL_ZIP_DEV_MODE
ffff = 25
ffffffff = 250
#else
ffff = 0xffff
ffffffff = 0xffffffff
#endif
-- | The default permissions for the files, permissions not set on Windows,
-- and are set to rw on Unix. This mimics the behavior of the zip utility.
defaultFileMode :: Word32
#ifdef mingw32_HOST_OS
defaultFileMode = 0
#else
defaultFileMode = Unix.fromFileMode 0o600
#endif