Handle installing Pandoc

lib/mix/tasks/pandoc.ex

@@ -1,13 +1,44 @@
 defmodule Mix.Tasks.Pandoc do
-  use Mix.Task
+  @moduledoc """
+  Invokes pandoc with the given args.
 
-  @ext ".md"
+  Usage:
+
+      $ mix pandoc TASK_OPTIONS PROFILE PANDOC_ARGS
+
+  Example:
+
+      $ mix pandoc default documents/hello.md -o priv/static/posts/hello.html
+
+  If pandoc is not installed, it is automatically downloaded. Note the
+  arguments given to this task will be appended to any configured arguments.
+
+  ## Options
+
+    * `--runtime-config` - load the runtime configuration
+      before executing command
+
+  Note flags to control this Mix task must be given before the profile:
+
+      $ mix pandoc --runtime-config default documents/hello.md
+
+  """
+
+  @shortdoc "Invokes pandoc with the profile and args"
+  @compile {:no_warn_undefined, Mix}
+
+  use Mix.Task
 
   @impl true
   def run(args) do
     switches = [runtime_config: :boolean]
     {opts, remaining_args} = OptionParser.parse_head!(args, switches: switches)
 
+    if function_exported?(Mix, :ensure_application!, 1) do
+      Mix.ensure_application!(:inets)
+      Mix.ensure_application!(:ssl)
+    end
+
     if opts[:runtime_config] do
       Mix.Task.run("app.config")
     else
@@ -19,28 +50,11 @@ defmodule Mix.Tasks.Pandoc do
     install_and_run(remaining_args)
   end
 
-  defp install_and_run([profile | _args] = all) do
-    IO.puts("Converting markdown...")
-
-    profile = String.to_atom(profile)
-    config = Application.get_env(:pandoc, profile)
-    args = config[:args] || []
-    opts = [cd: config[:cd] || File.cwd!()]
-
-    out_path = List.last(args)
-    full_out_path = [opts[:cd], out_path] |> Path.join() |> Path.expand()
-    File.rm_rf!(full_out_path)
-    File.mkdir_p!(full_out_path)
-
-    opts[:cd]
-    |> Path.join("*#{@ext}")
-    |> Path.wildcard()
-    |> Enum.each(fn path ->
-      case Pandoc.run(profile, path) do
-        0 -> :ok
-        status -> Mix.raise("`mix pandoc #{Enum.join(all, " ")}` exited with #{status}")
-      end
-    end)
+  defp install_and_run([profile | args] = all) do
+    case Pandoc.install_and_run(String.to_atom(profile), args) do
+      0 -> :ok
+      status -> Mix.raise("`mix pandoc #{Enum.join(all, " ")}` exited with #{status}")
+    end
   end
 
   defp install_and_run([]) do

lib/mix/tasks/pandoc.install.ex

@@ -0,0 +1,70 @@
+defmodule Mix.Tasks.Pandoc.Install do
+  @moduledoc """
+  Installs pandoc under `_build`.
+
+  ```bash
+  $ mix pandoc.install
+  $ mix pandoc.install --if-missing
+  ```
+
+  By default, it installs #{Pandoc.latest_version()} but you can configure it
+  in your config files, such as:
+
+      config :pandoc, :version, "#{Pandoc.latest_version()}"
+
+  ## Options
+
+    * `--runtime-config` - load the runtime configuration before executing
+      command
+
+    * `--if-missing` - install only if the given version does not exist
+  """
+
+  @shortdoc "Installs pandoc under _build"
+  @compile {:no_warn_undefined, Mix}
+
+  use Mix.Task
+
+  @impl true
+  def run(args) do
+    valid_options = [runtime_config: :boolean, if_missing: :boolean]
+
+    {opts, base_url} =
+      case OptionParser.parse_head!(args, strict: valid_options) do
+        {opts, []} ->
+          {opts, Pandoc.default_base_url()}
+
+        {opts, [base_url]} ->
+          {opts, base_url}
+
+        {_, _} ->
+          Mix.raise("""
+          Invalid arguments to pandoc.install, expected one of:
+
+              mix pandoc.install
+              mix pandoc.install 'https://github.com/jgm/pandoc/releases/download/$version/pandoc-$version-$target.tar.gz'
+              mix pandoc.install --runtime-config
+              mix pandoc.install --if-missing
+          """)
+      end
+
+    if opts[:runtime_config], do: Mix.Task.run("app.config")
+
+    if opts[:if_missing] && latest_version?() do
+      :ok
+    else
+      if function_exported?(Mix, :ensure_application!, 1) do
+        Mix.ensure_application!(:inets)
+        Mix.ensure_application!(:ssl)
+      end
+
+      Mix.Task.run("loadpaths")
+      Pandoc.install(base_url)
+    end
+  end
+
+  defp latest_version?() do
+    version = Pandoc.configured_version()
+    match?({:ok, ^version}, Pandoc.bin_version())
+  end
+end

lib/pandoc.ex

@@ -1,17 +1,130 @@
 defmodule Pandoc do
+  # https://github.com/jgm/pandoc/releases
+  @latest_version "3.6.1"
+
   @moduledoc """
-  Documentation for `Pandoc`.
+  Pandoc is an installer, runner and watcher for [pandoc](https://pandoc.org).
+
+  ## Profiles
+
+  You can define multiple pandoc profiles. By default, there is a profile
+  called `:default` which you can configure its args, current directory and
+  environment. You can make the args dynamic by defining a function.
+
+      config :pandoc,
+        version: "#{@latest_version}",
+        default: [
+          args: fn extra_args ->
+            {_, [input_file], _} = OptionParser.parse(extra_args, switches: [])
+            ~w(--output=../priv/static/posts/\#{Path.rootname(input_file)}.html)
+          end,
+          cd: Path.expand("../documents", __DIR__)
+        ]
+
+  ## Pandoc configuration
+
+  There are four global configurations for the pandoc application:
+
+    * `:version` - the expected pandoc version
+
+    * `:version_check` - whether to perform the version check or not.
+      Useful when you manage the pandoc executable with an external
+      tool
+
+    * `:cacerts_path` - the directory to find certificates for
+      https connections
+
+    * `:path` - the path to find the pandoc executable at. By
+      default, it is automatically downloaded and placed inside
+      the `_build` directory of your current app
+
+  Overriding the `:path` is not recommended, as we will automatically download
+  and manage `pandoc` for you. But in case you can't download it, you may want
+  to set the `:path` to a configurable system location.
   """
 
-  def run(profile, ["--watch" | _]) do
-    config = Application.get_env(:pandoc, profile)
-    opts = [cd: config[:cd] || File.cwd!()]
-    dirs = [opts[:cd], Path.join(opts[:cd], "_drafts")]
+  require Logger
+
+  @doc false
+  # Latest known version at the time of publishing.
+  def latest_version, do: @latest_version
+
+  @doc """
+  Returns the configured pandoc version.
+  """
+  def configured_version do
+    Application.get_env(:pandoc, :version, latest_version())
+  end
+
+  @doc """
+  Returns the configuration for the given profile.
+
+  Returns nil if the profile does not exist.
+  """
+  def config_for!(profile) when is_atom(profile) do
+    Application.get_env(:pandoc, profile) ||
+      raise ArgumentError, """
+      unknown pandoc profile. Make sure the profile is defined in your config/config.exs file, such as:
+
+          config :pandoc,
+            version: "#{@latest_version}",
+            #{profile}: [
+              cd: Path.expand("../documents", __DIR__)
+            ]
+      """
+  end
+
+  @doc """
+  Returns the path to the executable.
+
+  The executable may not be available if it was not yet installed.
+  """
+  def bin_path do
+    name = "pandoc-#{target()}"
+
+    Application.get_env(:pandoc, :path) ||
+      if Code.ensure_loaded?(Mix.Project) do
+        relative_build_dir = Mix.Project.build_path() |> Path.dirname() |> Path.relative_to_cwd()
+        project_dir = Path.dirname(Mix.Project.project_file())
+        Path.join([project_dir, relative_build_dir, name])
+      else
+        Path.expand("_build/#{name}")
+      end
+  end
+
+  @doc """
+  Returns the version of the pandoc executable.
+
+  Returns `{:ok, version_string}` on success or `:error` when the executable
+  is not available.
+  """
+  def bin_version do
+    path = bin_path()
+
+    with true <- File.exists?(path),
+         {out, 0} <- System.cmd(path, ["--version"]),
+         [vsn] <- Regex.run(~r/#{Path.basename(path)} ([^\s]+)/, out, capture: :all_but_first) do
+      {:ok, vsn}
+    else
+      _ -> :error
+    end
+  end
+
+  @doc """
+  Starts a file system watcher that runs the given command with `args` when a
+  file event is received for a file that matches the given pattern.
+
+  The given args will be appended to the configured args. The task output will
+  be streamed directly to stdio.
+  """
+  def watch(profile, extra_args \\ [], pattern \\ ~r/\.md$/) when is_atom(profile) do
+    config = config_for!(profile)
+    opts = [dirs: [config[:cd] || File.cwd!()]]
 
     ref =
       __MODULE__.Supervisor
       |> Supervisor.start_child(
-        Supervisor.child_spec({Pandoc.Watcher, [profile, dirs: dirs]},
+        Supervisor.child_spec({Pandoc.Watcher, [profile, opts, pattern, extra_args]},
           restart: :transient,
           id: __MODULE__.Watcher
         )
@@ -27,9 +140,24 @@ defmodule Pandoc do
     end
   end
 
-  def run(profile, path) do
-    config = Application.get_env(:pandoc, profile)
-    args = config[:args] || []
+  @doc """
+  Runs the given command with `args`.
+
+  The given args will be appended to the configured args. The task output will
+  be streamed directly to stdio. It returns the status of the underlying call.
+  """
+  def run(profile, extra_args) when is_atom(profile) and is_list(extra_args) do
+    config = config_for!(profile)
+
+    args =
+      case config[:args] do
+        args_fn when is_function(args_fn) -> args_fn.(extra_args)
+        args -> args || []
+      end
+
+    if args == [] and extra_args == [] do
+      raise "no arguments passed to pandoc"
+    end
 
     opts = [
       cd: config[:cd] || File.cwd!(),
@@ -37,17 +165,215 @@ defmodule Pandoc do
       stderr_to_stdout: true
     ]
 
-    new_filename =
-      path |> Path.basename() |> String.replace_suffix(".md", ".html") |> String.slice(11..-1//1)
+    {parsed_args, _, _} = OptionParser.parse(args, switches: [output: :string])
+    {_, input_files, _} = OptionParser.parse(extra_args, switches: [])
 
-    new_path = args |> List.last() |> Path.join(new_filename)
-    out_path = Path.join(opts[:cd], new_path) |> Path.expand()
-
-    if File.exists?(path) do
-      args = List.replace_at(args, -1, out_path)
-      "pandoc" |> System.cmd(args ++ [path], opts) |> elem(1)
+    if parsed_args[:output] &&
+         not File.cd!(opts[:cd], fn ->
+           input_files |> Enum.map(&File.exists?(&1)) |> Enum.all?()
+         end) do
+      parsed_args[:output] |> Path.expand(opts[:cd]) |> File.rm!()
     else
-      File.rm(out_path)
+      bin_path() |> System.cmd(args ++ extra_args, opts) |> elem(1)
     end
   end
+
+  defp start_unique_install_worker() do
+    ref =
+      __MODULE__.Supervisor
+      |> Supervisor.start_child(
+        Supervisor.child_spec({Task, &install/0}, restart: :transient, id: __MODULE__.Installer)
+      )
+      |> case do
+        {:ok, pid} -> pid
+        {:error, {:already_started, pid}} -> pid
+      end
+      |> Process.monitor()
+
+    receive do
+      {:DOWN, ^ref, _, _, _} -> :ok
+    end
+  end
+
+  @doc """
+  Installs, if not available, and then runs `pandoc`.
+
+  This task may be invoked concurrently and it will avoid concurrent installs.
+
+  Returns the same as `run/2`.
+  """
+  def install_and_run(profile, args) do
+    File.exists?(bin_path()) || start_unique_install_worker()
+
+    run(profile, args)
+  end
+
+  @doc """
+  The default URL to install Pandoc from.
+  """
+  def default_base_url do
+    "https://github.com/jgm/pandoc/releases/download/$version/pandoc-$version-$target.tar.gz"
+  end
+
+  @doc """
+  Installs pandoc with `configured_version/0`.
+
+  If invoked concurrently, this task will perform concurrent installs.
+  """
+  def install(base_url \\ default_base_url()) do
+    version = configured_version()
+    tmp_opts = if System.get_env("MIX_XDG"), do: %{os: :linux}, else: %{}
+
+    tmp_dir =
+      freshdir_p(:filename.basedir(:user_cache, "phx-pandoc", tmp_opts)) ||
+        freshdir_p(Path.join(System.tmp_dir!(), "phx-pandoc")) ||
+        raise "could not install pandoc. Set MIX_XGD=1 and then set XDG_CACHE_HOME to the path you want to use as cache"
+
+    url = get_url(base_url)
+    tar = fetch_body!(url)
+
+    case :erl_tar.extract({:binary, tar}, [:compressed, cwd: to_charlist(tmp_dir)]) do
+      :ok -> :ok
+      other -> raise "couldn't unpack archive: #{inspect(other)}"
+    end
+
+    bin_path = bin_path()
+    File.mkdir_p!(Path.dirname(bin_path))
+    [tmp_dir, "pandoc-" <> version, "bin", "pandoc"] |> Path.join() |> File.cp!(bin_path)
+  end
+
+  defp freshdir_p(path) do
+    with {:ok, _} <- File.rm_rf(path),
+         :ok <- File.mkdir_p(path) do
+      path
+    else
+      _ -> nil
+    end
+  end
+
+  defp fetch_body!(url, retry \\ true) do
+    scheme = URI.parse(url).scheme
+    url = String.to_charlist(url)
+    Logger.debug("Downloading pandoc from #{url}")
+
+    {:ok, _} = Application.ensure_all_started(:inets)
+    {:ok, _} = Application.ensure_all_started(:ssl)
+
+    if proxy = proxy_for_scheme(scheme) do
+      %{host: host, port: port} = URI.parse(proxy)
+      Logger.debug("Using #{String.upcase(scheme)}_PROXY: #{proxy}")
+      set_option = if "https" == scheme, do: :https_proxy, else: :proxy
+      :httpc.set_options([{set_option, {{String.to_charlist(host), port}, []}}])
+    end
+
+    # https://erlef.github.io/security-wg/secure_coding_and_deployment_hardening/inets
+    cacertfile = cacertfile() |> String.to_charlist()
+
+    http_options =
+      [
+        ssl: [
+          verify: :verify_peer,
+          cacertfile: cacertfile,
+          depth: 2,
+          customize_hostname_check: [
+            match_fun: :public_key.pkix_verify_hostname_match_fun(:https)
+          ],
+          versions: protocol_versions()
+        ]
+      ]
+      |> maybe_add_proxy_auth(scheme)
+
+    options = [body_format: :binary]
+
+    case {retry, :httpc.request(:get, {url, []}, http_options, options)} do
+      {_, {:ok, {{_, 200, _}, _headers, body}}} ->
+        body
+
+      {true, {:error, {:failed_connect, [{:to_address, _}, {inet, _, reason}]}}}
+      when inet in [:inet, :inet6] and
+             reason in [:ehostunreach, :enetunreach, :eprotonosupport, :nxdomain] ->
+        :httpc.set_options(ipfamily: fallback(inet))
+        fetch_body!(url, false)
+
+      other ->
+        raise """
+        Couldn't fetch #{url}: #{inspect(other)}
+
+        This typically means we cannot reach the source or you are behind a proxy.
+        You can try again later and, if that does not work, you might:
+
+          1. If behind a proxy, ensure your proxy is configured and that
+             your certificates are set via the cacerts_path configuration
+
+          2. Manually download the executable from the URL above and
+             place it inside "_build/pandoc-#{target()}"
+        """
+    end
+  end
+
+  defp fallback(:inet), do: :inet6
+  defp fallback(:inet6), do: :inet
+
+  defp proxy_for_scheme("http") do
+    System.get_env("HTTP_PROXY") || System.get_env("http_proxy")
+  end
+
+  defp proxy_for_scheme("https") do
+    System.get_env("HTTPS_PROXY") || System.get_env("https_proxy")
+  end
+
+  defp maybe_add_proxy_auth(http_options, scheme) do
+    case proxy_auth(scheme) do
+      nil -> http_options
+      auth -> [{:proxy_auth, auth} | http_options]
+    end
+  end
+
+  defp proxy_auth(scheme) do
+    with proxy when is_binary(proxy) <- proxy_for_scheme(scheme),
+         %{userinfo: userinfo} when is_binary(userinfo) <- URI.parse(proxy),
+         [username, password] <- String.split(userinfo, ":") do
+      {String.to_charlist(username), String.to_charlist(password)}
+    else
+      _ -> nil
+    end
+  end
+
+  defp cacertfile() do
+    Application.get_env(:pandoc, :cacerts_path) || CAStore.file_path()
+  end
+
+  defp protocol_versions do
+    if otp_version() < 25, do: [:"tlsv1.2"], else: [:"tlsv1.2", :"tlsv1.3"]
+  end
+
+  defp otp_version do
+    :erlang.system_info(:otp_release) |> List.to_integer()
+  end
+
+  # Available targets: https://github.com/jgm/pandoc/releases
+  # We support only linux-amd64, for now.
+  defp target do
+    case :os.type() do
+      # Assuming it's an x86 CPU
+      {:win32, _} ->
+        raise "pandoc does not currently support OS family: Windows"
+
+      {:unix, osname} ->
+        arch_str = :erlang.system_info(:system_architecture)
+        [arch | _] = arch_str |> List.to_string() |> String.split("-")
+
+        case arch do
+          "amd64" -> "#{osname}-amd64"
+          "x86_64" -> "#{osname}-amd64"
+          _ -> raise "pandoc does not currently support architecture: #{arch_str}"
+        end
+    end
+  end
+
+  defp get_url(base_url) do
+    base_url
+    |> String.replace("$version", configured_version())
+    |> String.replace("$target", target())
+  end
 end

lib/pandoc/application.ex

@@ -5,16 +5,37 @@ defmodule Pandoc.Application do
 
   use Application
 
-  @impl true
-  def start(_type, _args) do
-    children = [
-      # Starts a worker by calling: Pandoc.Worker.start_link(arg)
-      # {Pandoc.Worker, arg}
-    ]
+  require Logger
+  import Pandoc, only: [latest_version: 0, configured_version: 0, bin_version: 0]
 
-    # See https://hexdocs.pm/elixir/Supervisor.html
-    # for other strategies and supported options
-    opts = [strategy: :one_for_one, name: Pandoc.Supervisor]
-    Supervisor.start_link(children, opts)
+  @doc false
+  def start(_, _) do
+    if Application.get_env(:pandoc, :version_check, true) do
+      unless Application.get_env(:pandoc, :version) do
+        Logger.warning("""
+        pandoc version is not configured. Please set it in your config files:
+
+            config :pandoc, :version, "#{latest_version()}"
+        """)
+      end
+
+      configured_version = configured_version()
+
+      case bin_version() do
+        {:ok, ^configured_version} ->
+          :ok
+
+        {:ok, version} ->
+          Logger.warning("""
+          Outdated pandoc version. Expected #{configured_version}, got #{version}. \
+          Please run `mix pandoc.install` or update the version in your config files.\
+          """)
+
+        :error ->
+          :ok
+      end
+    end
+
+    Supervisor.start_link([], strategy: :one_for_one, name: Pandoc.Supervisor)
   end
 end

lib/pandoc/watcher.ex

@@ -1,22 +1,29 @@
 defmodule Pandoc.Watcher do
-  use GenServer
+  @moduledoc false
 
-  @ext ".md"
+  use GenServer
 
   def start_link(args) do
     GenServer.start_link(__MODULE__, args)
   end
 
-  def init([profile | args]) do
-    {:ok, watcher_pid} = FileSystem.start_link(args)
+  # Callbacks
+
+  @impl true
+  def init([profile, options, pattern, extra_args]) do
+    {:ok, watcher_pid} = FileSystem.start_link(options)
     FileSystem.subscribe(watcher_pid)
-    {:ok, %{watcher_pid: watcher_pid, profile: profile}}
+    {:ok, %{watcher_pid: watcher_pid, profile: profile, pattern: pattern, extra_args: extra_args}}
   end
 
+  @impl true
   def handle_info({:file_event, watcher_pid, {path, events}}, %{watcher_pid: watcher_pid} = state) do
-    case {Path.extname(path), :closed in events or :deleted in events} do
-      {@ext, true} -> Pandoc.run(state[:profile], path)
-      _ -> nil
+    case {String.match?(path, state[:pattern]), :closed in events or :deleted in events} do
+      {true, true} ->
+        Pandoc.install_and_run(state[:profile], [Path.basename(path) | state[:extra_args]])
+
+      _ ->
+        nil
     end
 
     {:noreply, state}