defmodule Zap do
  @moduledoc """
  Native ZIP archive creation with chunked input and output.

  Erlang/OTP provides the powerful `:zip` and `:zlib` modules, but they can only create an archive
  all at once. That requires _all_ of the data to be kept in memory or written to disk. What if
  you don't have enough space to keep the file in memory or on disk? With Zap you can add files
  one at a time while writing chunks of data at the same time.

  ## Examples

  Create a ZIP by adding a single entry at a time:

  ```elixir
  iodata =
    Zap.new()
    |> Zap.entry("a.txt", a_binary)
    |> Zap.entry("b.txt", some_iodata)
    |> Zap.entry("c.txt", more_iodata)
    |> Zap.to_iodata()

  File.write!("archive.zip", iodata, [:binary, :raw])
  ```

  Use `into` support from the `Collectable` protocol to build a ZIP dynamically:

  ```elixir
  iodata =
    "*.*"
    |> Path.wildcard()
    |> Enum.map(fn path -> {Path.basename(path), File.read!(path)} end)
    |> Enum.into(Zap.new())
    |> Zap.to_iodata()

  File.write!("files.zip", iodata, [:binary, :raw])
  ```

  Use `Zap.into_stream/2` to incrementally build a ZIP by chunking files into an archive:

  ```elixir
  one_mb = 1024 * 1024

  write_fun = &File.write!("streamed.zip", &1, [:append, :binary, :raw])

  file_list
  |> Stream.map(fn path -> {Path.basename(path), File.read!(path)} end)
  |> Zap.into_stream(one_mb)
  |> Stream.each(write_fun)
  |> Stream.run()
  ```

  ## Glossary

  The entry and header bytes are composed based on the [ZIP specification provided by
  PKWare](https://pkware.cachefly.net/webdocs/casestudies/APPNOTE.TXT). Some helpful terms
  that you may encounter in the function documentation:

  * LFH (Local File Header) — Included before each file in the archive. The header contains
  details about the name and size of the entry.
  * CDH (Central Directory Header) — The final bits of an archive, this contains summary
  information about the files contained within the archive.
  """

  alias Zap.{Directory, Entry}

  @type t :: %__MODULE__{entries: [Entry.t()]}

  defstruct compression: 0, entries: []

  @doc """
  Initialize a new Zap struct.

  The struct is used to accumulate entries, which can then be flushed as parts of a zip file.
  """
  @doc since: "0.1.0"
  @spec new() :: t()
  def new(opts \\ []) do
    struct!(__MODULE__, opts)
  end

  @doc """
  Add a named entry to a zap struct.

  ## Example

      Zap.new()
      |> Zap.entry("a.jpg", jpg_data)
      |> Zap.entry("b.png", png_data)
  """
  @doc since: "0.1.0"
  @spec entry(zap :: t(), name :: binary(), data :: binary()) :: t()
  def entry(%__MODULE__{} = zap, name, data) when is_binary(name) and is_binary(data) do
    %{zap | entries: List.flatten([zap.entries | [Entry.new(name, data)]])}
  end

  @doc """
  Check the total number of un-flushed bytes available.

  ## Example

      iex> Zap.bytes(Zap.new())
      0

      iex> Zap.new()
      ...> |> Zap.entry("a.txt", "a")
      ...> |> Zap.bytes()
      52

      iex> zap = Zap.new()
      ...> zap = Zap.entry(zap, "a.txt", "a")
      ...> {zap, _bytes} = Zap.flush(zap, :all)
      ...> Zap.bytes(zap)
      0
  """
  @doc since: "0.1.0"
  @spec bytes(zap :: t()) :: non_neg_integer()
  def bytes(%__MODULE__{entries: entries}) do
    Enum.reduce(entries, 0, &(&1.size + &2))
  end

  @doc """
  Output a complete iolist of data from a Zap struct.

  This is a convenient way of combining the output from `flush/1` and `final/1`.

  Though the function is called `to_iodata` it _also_ returns a zap struct because the struct is
  modified when it is flushed.

  ## Example

      iex> Zap.new()
      ...> |> Zap.entry("a.txt", "aaaa")
      ...> |> Zap.entry("b.txt", "bbbb")
      ...> |> Zap.to_iodata()
      ...> |> IO.iodata_length()
      248
  """
  @doc since: "0.1.0"
  @spec to_iodata(zap :: t()) :: iolist()
  def to_iodata(%__MODULE__{} = zap) do
    {zap, flush} = flush(zap)
    {_ap, final} = final(zap)

    [flush, final]
  end

  @doc """
  Flush a fixed number of bytes from the stored entries.

  Flushing is stateful, meaning the same data won't be flushed on successive calls.

  ## Example

      iex> Zap.new()
      ...> |> Zap.entry("a.txt", "aaaa")
      ...> |> Zap.entry("b.txt", "bbbb")
      ...> |> Zap.flush()
      ...> |> elem(1)
      ...> |> IO.iodata_length()
      110
  """
  @doc since: "0.1.0"
  @spec flush(zap :: t(), bytes :: pos_integer() | :all) :: {t(), iodata()}
  def flush(%__MODULE__{entries: entries} = zap, bytes \\ :all) do
    {flushed, entries, _} =
      Enum.reduce(entries, {[], [], bytes}, fn entry, {iodata, entries, bytes} ->
        {entry, binary} = Entry.consume(entry, bytes)

        next_bytes =
          cond do
            bytes == :all -> :all
            bytes - byte_size(binary) > 0 -> bytes - byte_size(binary)
            true -> 0
          end

        {[iodata | binary], [entry | entries], next_bytes}
      end)

    {%{zap | entries: Enum.reverse(entries)}, flushed}
  end

  @doc """
  Generate the final CDH (Central Directory Header), required to complete an archive.
  """
  @doc since: "0.1.0"
  @spec final(zap :: t()) :: {t(), iodata()}
  def final(%__MODULE__{entries: entries} = zap) do
    {zap, Directory.encode(entries)}
  end

  @doc """
  Stream an enumerable of `name`/`data` pairs into a zip structure and emit chunks of zip data.

  The chunked output will be _at least_ the size of `chunk_size`, but they may be much larger. The
  last emitted chunk automatically includes the central directory header, the closing set of
  bytes.

  ## Example

      iex> %{"a.txt" => "aaaa", "b.txt" => "bbbb"}
      ...> |> Zap.into_stream(8)
      ...> |> Enum.to_list()
      ...> |> IO.iodata_to_binary()
      ...> |> :zip.table()
      ...> |> elem(0)
      :ok
  """
  @doc since: "0.1.0"
  @spec into_stream(enum :: Enumerable.t(), chunk_size :: pos_integer()) :: Enumerable.t()
  def into_stream(enum, chunk_size \\ 1024 * 1024) when is_integer(chunk_size) do
    chunk_fun = fn {name, data}, zap ->
      zap = entry(zap, name, data)

      if bytes(zap) >= chunk_size do
        {zap, flushed} = flush(zap, :all)

        {:cont, flushed, zap}
      else
        {:cont, zap}
      end
    end

    after_fun = fn zap ->
      iodata = to_iodata(zap)

      {:cont, iodata, zap}
    end

    Stream.chunk_while(enum, Zap.new(), chunk_fun, after_fun)
  end

  defimpl Collectable do
    def into(original) do
      fun = fn
        zap, {:cont, {name, data}} -> Zap.entry(zap, name, data)
        zap, :done -> zap
        _zap, :halt -> :ok
      end

      {original, fun}
    end
  end

  defimpl Inspect do
    import Inspect.Algebra

    alias Inspect.{List, Opts}

    def inspect(zap, opts) do
      opts = %Opts{opts | charlists: :as_lists}

      concat(["#Zap<", List.inspect(names(zap), opts), ">"])
    end

    defp names(%{entries: entries}) do
      entries
      |> Enum.reverse()
      |> Enum.map(& &1.header.name)
    end
  end
end