defmodule Unzip do @moduledoc """ Module to get files out of a zip. Works with local and remote files ## Overview Unzip tries to solve problem of accessing files from a zip which is not local (Aws S3, sftp etc). It does this by simply separating file system and zip implementation. Anything which implements `Unzip.FileAccess` can be used to get zip contents. Unzip relies on the ability to seek and read of the file, This is due to the nature of zip file. Files from the zip are read on demand. ## Usage # Unzip.LocalFile implements Unzip.FileAccess zip_file = Unzip.LocalFile.open("foo/bar.zip") # `new` reads list of files by reading central directory found at the end of the zip {:ok, unzip} = Unzip.new(zip_file) # Alternatively if you have the zip file in memory as binary you can # directly pass it to `Unzip.new(binary)` to unzip # # {:ok, unzip} = Unzip.new(<>) # returns list of files along with metadata file_entries = Unzip.list_entries(unzip) # returns decompressed file stream stream = Unzip.file_stream!(unzip, "baz.png") Supports STORED and DEFLATE compression methods. Does not support zip64 specification yet """ require Logger alias Unzip.FileAccess alias Unzip.FileBuffer alias Unzip.RangeTree @chunk_size 65_000 @typedoc """ Struct holding zip related metadata, returned by the `new/1`. Public fields: * `zip` - Zip struct passed to `new/1` Remaining fields are private and must not be accessed directly. """ @type t :: %__MODULE__{ zip: struct(), cd_list: map() } defstruct [:zip, :cd_list] defmodule Error do defexception [:message] end defmodule Entry do @moduledoc """ File metadata returned by `Unzip.list_entries/1` * `:file_name` - (string) File name with complete path. Directory files will have `/` at the end of their name * `:last_modified_datetime` - (NaiveDateTime) last modified date and time of the file. It will be set to `nil` if datetime is invalid * `:compressed_size` - (positive integer) Compressed file size in bytes * `:uncompressed_size` - (positive integer) Uncompressed file size in bytes """ @type t :: %__MODULE__{ file_name: String.t(), last_modified_datetime: NaiveDateTime.t() | nil, compressed_size: pos_integer(), uncompressed_size: pos_integer() } defstruct [ :file_name, :last_modified_datetime, :compressed_size, :uncompressed_size ] end @doc """ Reads zip metadata from the passed zip file. `zip` must implement `Unzip.FileAccess` protocol. Fetches the list of files present in the zip and other metadata by reading central directory found at the end of the zip. Returns `Unzip` struct which contains passed `zip` struct and zip metadata. """ @spec new(Unzip.FileAccess.t()) :: {:ok, t} | {:error, term()} def new(zip) do with {:ok, eocd} <- find_eocd(zip), {:ok, entries} <- read_cd_entries(zip, eocd) do {:ok, %Unzip{zip: zip, cd_list: entries}} end end @doc """ Returns list of files metadata. This does not make `pread` call as metadata is already by `new/1`. See `Unzip.Entry` for metadata fields """ @spec list_entries(t) :: list(Entry.t()) def list_entries(unzip) do Enum.map(unzip.cd_list, fn {_, entry} -> %Entry{ file_name: entry.file_name, last_modified_datetime: entry.last_modified_datetime, compressed_size: entry.compressed_size, uncompressed_size: entry.uncompressed_size } end) end @type stream_options :: {:chunk_size, pos_integer()} @doc """ Returns decompressed file as a stream of stream of [iodata](https://hexdocs.pm/elixir/IO.html#module-io-data). `file_path` *must* be the complete file path within the zip. The file entry is read in chunks, then decompressed in a streaming fashion. ### Options * `chunk_size` - Chunks are read from the source of the size specified by `chunk_size`. This is *not* the size of the chunk returned by `file_stream!` since the chunk size varies after decompressing the. Useful when reading from the source is expensive and you want optimize by increasing the chunk size. Defaults to `65_000` """ @spec file_stream!(t, String.t()) :: Enumerable.t() @spec file_stream!(t, String.t(), [stream_options]) :: Enumerable.t() def file_stream!(%Unzip{zip: zip, cd_list: cd_list}, file_path, opts \\ []) do unless Map.has_key?(cd_list, file_path) do raise Error, message: "File #{inspect(file_path)} not present in the zip" end entry = Map.fetch!(cd_list, file_path) local_header = pread!(zip, entry.local_header_offset, 30) case local_header do <<0x04034B50::little-32, _::little-32, compression_method::little-16, _::little-128, file_path_length::little-16, extra_field_length::little-16>> -> offset = entry.local_header_offset + 30 + file_path_length + extra_field_length stream!(zip, offset, entry.compressed_size, opts) |> decompress(compression_method) |> crc_check(entry.crc) _ -> raise Error, message: "Invalid local zip header" end end defp stream!(file, offset, size, opts) do end_offset = offset + size Stream.unfold(offset, fn offset when offset >= end_offset -> nil offset -> chunk_size = Keyword.get(opts, :chunk_size, @chunk_size) next_offset = min(offset + chunk_size, end_offset) data = pread!(file, offset, next_offset - offset) {data, next_offset} end) end @end_of_stream :__end__ defp decompress(stream, 0x8) do stream |> Stream.concat([@end_of_stream]) |> Stream.transform( fn -> z = :zlib.open() :ok = :zlib.inflateInit(z, -15) z end, fn @end_of_stream, z -> :zlib.inflateEnd(z) {[], z} data, z -> {[:zlib.inflate(z, data)], z} end, fn z -> :zlib.close(z) end ) end defp decompress(stream, 0x0), do: stream defp decompress(_stream, compression_method) do raise Error, message: "Compression method #{compression_method} is not supported" end defp crc_check(stream, expected_crc) do stream |> Stream.concat([@end_of_stream]) |> Stream.transform(:erlang.crc32(<<>>), fn @end_of_stream, crc -> unless crc == expected_crc do raise Error, message: "CRC mismatch. expected: #{expected_crc} got: #{crc}" end {[], :ok} data, crc -> {[data], :erlang.crc32(crc, data)} end) end defp read_cd_entries(zip, eocd) do with {:ok, file_buffer} <- FileBuffer.new( zip, @chunk_size, eocd.cd_offset + eocd.cd_size, eocd.cd_offset, :forward ) do parse_cd(file_buffer, %{entries: %{}, range_tree: RangeTree.new()}) end end defp parse_cd(%FileBuffer{buffer_position: pos, limit: limit}, %{entries: entries}) when pos >= limit, do: {:ok, entries} defp parse_cd(buffer, acc) do with {:ok, chunk, buffer} <- FileBuffer.next_chunk(buffer, 46), {:match_chunk, <<0x02014B50::little-32, _::little-32, flag::little-16, compression_method::little-16, mtime::little-16, mdate::little-16, crc::little-32, compressed_size::little-32, uncompressed_size::little-32, file_name_length::little-16, extra_field_length::little-16, comment_length::little-16, _::little-64, local_header_offset::little-32>>} <- {:match_chunk, chunk}, {:ok, buffer} <- FileBuffer.move_forward_by(buffer, 46), {:ok, file_name, buffer} <- FileBuffer.next_chunk(buffer, file_name_length), {:ok, buffer} <- FileBuffer.move_forward_by(buffer, file_name_length), {:ok, extra_fields, buffer} <- FileBuffer.next_chunk(buffer, extra_field_length), {:ok, buffer} <- FileBuffer.move_forward_by(buffer, extra_field_length), {:ok, _file_comment, buffer} <- FileBuffer.next_chunk(buffer, comment_length), {:ok, buffer} <- FileBuffer.move_forward_by(buffer, comment_length) do entry = %{ bit_flag: flag, compression_method: compression_method, last_modified_datetime: to_datetime(<>, <>), crc: crc, compressed_size: compressed_size, uncompressed_size: uncompressed_size, local_header_offset: local_header_offset, # TODO: we should treat binary as "IBM Code Page 437" encoded string if GP flag 11 is not set file_name: file_name } entry = if need_zip64_extra?(entry) do merge_zip64_extra(entry, extra_fields) else entry end case add_entry(acc, file_name, entry) do {:error, _} = error -> error acc -> parse_cd(buffer, acc) end else {:error, :invalid_count} -> {:error, "Invalid zip file, invalid central directory"} {:match_chunk, _binary} -> {:error, "Invalid zip file, invalid central directory file header"} {:error, _error} = error -> error end end defp add_entry(%{entries: entries, range_tree: range_tree}, file_name, entry) do if RangeTree.overlap?(range_tree, entry.local_header_offset, entry.compressed_size) do {:error, "Invalid zip file, found overlapping zip entries"} else %{ entries: Map.put(entries, file_name, entry), range_tree: RangeTree.insert(range_tree, entry.local_header_offset, entry.compressed_size) } end end defp need_zip64_extra?(%{ compressed_size: cs, uncompressed_size: ucs, local_header_offset: offset }) do Enum.any?([cs, ucs, offset], &(&1 == 0xFFFFFFFF)) end @zip64_extra_field_id 0x0001 defp merge_zip64_extra(entry, extra) do zip64_extra = find_extra_fields(extra) |> Map.fetch!(@zip64_extra_field_id) {entry, zip64_extra} = if entry[:uncompressed_size] == 0xFFFFFFFF do <> = zip64_extra {%{entry | uncompressed_size: uncompressed_size}, zip64_extra} else {entry, zip64_extra} end {entry, zip64_extra} = if entry[:compressed_size] == 0xFFFFFFFF do <> = zip64_extra {%{entry | compressed_size: compressed_size}, zip64_extra} else {entry, zip64_extra} end {entry, _zip64_extra} = if entry[:local_header_offset] == 0xFFFFFFFF do <> = zip64_extra {%{entry | local_header_offset: local_header_offset}, zip64_extra} else {entry, zip64_extra} end entry end defp find_extra_fields(extra, result \\ %{}) defp find_extra_fields(<<>>, result), do: result defp find_extra_fields( <>, result ) do find_extra_fields(rest, Map.put(result, id, data)) end defp find_eocd(zip) do with {:ok, file_buffer} <- FileBuffer.new(zip, @chunk_size), {:ok, eocd, file_buffer} <- find_eocd(file_buffer, 0) do case find_zip64_eocd(file_buffer) do {:ok, zip64_eocd} -> {:ok, zip64_eocd} _ -> {:ok, eocd} end end end @zip64_eocd_locator_size 20 @zip64_eocd_size 56 defp find_zip64_eocd(file_buffer) do with {:ok, chunk, file_buffer} <- FileBuffer.next_chunk(file_buffer, @zip64_eocd_locator_size), true <- zip64?(chunk) do <<0x07064B50::little-32, _::little-32, eocd_offset::little-64, _::little-32>> = chunk {:ok, <<0x06064B50::little-32, _::64, _::16, _::16, _::32, _::32, _::64, total_entries::little-64, cd_size::little-64, cd_offset::little-64>>} = pread(file_buffer.file, eocd_offset, @zip64_eocd_size) {:ok, %{total_entries: total_entries, cd_size: cd_size, cd_offset: cd_offset}} else _ -> false end end defp zip64?(<<0x07064B50::little-32, _::little-128>>), do: true defp zip64?(_), do: false # Spec has variable length comment at the end of zip after EOCD, so # EOCD can anywhere in the zip file. To avoid exhaustive search, we # limit search space to last 5Mb. If we don't find EOCD within that # we assume it's an invalid zip @eocd_seach_limit 5 * 1024 * 1024 defp find_eocd(_file_buffer, consumed) when consumed > @eocd_seach_limit, do: {:error, "Invalid zip file, missing EOCD record"} @eocd_header_size 22 defp find_eocd(file_buffer, consumed) do with {:ok, chunk, file_buffer} <- FileBuffer.next_chunk(file_buffer, @eocd_header_size) do case chunk do <<0x06054B50::little-32, _ignore::little-48, total_entries::little-16, cd_size::little-32, cd_offset::little-32, ^consumed::little-16>> -> {:ok, buffer} = FileBuffer.move_backward_by(file_buffer, @eocd_header_size) {:ok, %{total_entries: total_entries, cd_size: cd_size, cd_offset: cd_offset}, buffer} chunk when byte_size(chunk) < @eocd_header_size -> {:error, "Invalid zip file, missing EOCD record"} _ -> {:ok, buffer} = FileBuffer.move_backward_by(file_buffer, 1) find_eocd(buffer, consumed + 1) end end end defp to_datetime(<>, <>) do case NaiveDateTime.new(1980 + year, month, day, hour, minute, second * 2) do {:ok, datetime} -> datetime _ -> nil end end defp pread!(file, offset, length) do case pread(file, offset, length) do {:ok, term} -> term {:error, reason} when is_binary(reason) -> raise Error, message: reason {:error, reason} -> raise Error, message: inspect(reason) end end defp pread(file, offset, length) do case FileAccess.pread(file, offset, length) do {:ok, term} when is_binary(term) -> {:ok, term} {:error, reason} -> {:error, reason} _ -> {:error, "Invalid data returned by pread/3. Expected binary"} end end end