org-garden/lib/org_garden/export.ex

defmodule OrgGarden.Export do
  @moduledoc """
  Org-to-Markdown export via Emacs batch + ox-hugo.

  Provides both single-file and batch export, plus a helper to compute
  the expected `.md` output path for a given `.org` source file.
  """

  require Logger

  @doc """
  Export a single `.org` file to Markdown via `emacs --batch` + ox-hugo.

  Accepts an optional `id_locations_file` path for pre-built org-id database.
  If not provided, builds the ID database inline (slower for batch exports).

  Returns `{:ok, exit_code}` with the emacs exit code (0 = success),
  or `{:error, reason}` if the command could not be executed.
  """
  @spec export_file(String.t(), String.t(), String.t(), String.t() | nil) ::
          {:ok, non_neg_integer()} | {:error, term()}
  def export_file(orgfile, notes_dir, output_dir, id_locations_file \\ nil) do
    OrgGarden.Telemetry.span_export(orgfile, fn ->
      do_export_file(orgfile, notes_dir, output_dir, id_locations_file)
    end)
  end

  defp do_export_file(orgfile, notes_dir, output_dir, id_locations_file) do
    section =
      orgfile
      |> Path.dirname()
      |> Path.relative_to(notes_dir)

    # ox-hugo requires static/ to exist for image asset copying
    File.mkdir_p!(Path.join(output_dir, "static"))

    # Build the org-id setup commands based on whether we have a pre-built file
    id_setup_args =
      if id_locations_file do
        # Use pre-built ID locations file (faster for parallel exports)
        [
          "--eval", ~s[(setq org-id-locations-file "#{id_locations_file}")],
          "--eval", "(org-id-locations-load)"
        ]
      else
        # Build ID locations inline (for single file exports)
        [
          "--eval", ~s[(setq org-id-extra-files (directory-files-recursively "#{notes_dir}" "\\\\.org$"))],
          "--eval", "(org-id-update-id-locations)"
        ]
      end

    {output, exit_code} =
      System.cmd(
        "emacs",
        [
          "--batch",
          "--eval", "(require 'ox-hugo)"
        ] ++
          id_setup_args ++
          [
            # Allow export to proceed even if some links cannot be resolved
            "--eval", "(setq org-export-with-broken-links 'mark)",
            # Prevent errors when file links point to non-existent files/headlines
            "--eval", "(advice-add 'org-link-search :around (lambda (orig-fn &rest args) (condition-case nil (apply orig-fn args) (error nil))))",
            "--eval", """
            (org-cite-register-processor 'passthrough
              :export-citation
              (lambda (citation _style _backend _info)
                (let ((keys (mapcar (lambda (ref)
                                      (concat "@" (org-element-property :key ref)))
                                    (org-cite-get-references citation))))
                  (format "[cite:%s]" (string-join keys ";")))))
            """,
            "--eval", "(setq org-cite-export-processors '((t passthrough)))",
            # Use YAML frontmatter instead of TOML to avoid date parsing issues
            # (TOML interprets bare dates like 2022-10-31 as date literals requiring time)
            "--eval", "(setq org-hugo-front-matter-format 'yaml)",
            "--eval", ~s[(setq org-hugo-base-dir "#{output_dir}")],
            "--eval", ~s[(setq org-hugo-default-section-directory "#{section}")],
            "--visit", orgfile,
            "--funcall", "org-hugo-export-to-md"
          ],
        stderr_to_stdout: true
      )

    # Log raw emacs output at debug level for troubleshooting
    if output != "", do: Logger.debug("emacs output:\n#{output}")

    if exit_code == 0 do
      {:ok, exit_code}
    else
      {:error, {:emacs_exit, exit_code}}
    end
  rescue
    e -> {:error, e}
  end

  @default_max_concurrency 8

  @doc """
  Export all `.org` files found under `notes_dir`.

  Exports files in parallel for improved performance. The concurrency level
  can be configured via the `:export_concurrency` application config or
  the `EXPORT_CONCURRENCY` environment variable. Defaults to #{@default_max_concurrency}.

  Returns `{:ok, success_count, failures}` where `success_count` is the number
  of successfully exported files and `failures` is a list of `{file, {:error, reason}}`
  tuples for files that failed to export. The pipeline continues even if some
  files fail.
  """
  @spec export_all(String.t(), String.t()) :: {:ok, non_neg_integer(), list()}
  def export_all(notes_dir, output_dir) do
    org_files =
      Path.join(notes_dir, "**/*.org")
      |> Path.wildcard()

    if org_files == [] do
      Logger.warning("No .org files found in #{notes_dir}")
      {:ok, 0, []}
    else
      max_concurrency = get_concurrency()
      Logger.info("Exporting #{length(org_files)} org file(s) from #{notes_dir} (concurrency: #{max_concurrency})")

      # Build org-id locations database once before parallel export
      id_locations_file = build_id_locations(notes_dir)

      results =
        org_files
        |> Task.async_stream(
          fn orgfile ->
            Logger.info("  exporting: #{orgfile}")
            result = export_file(orgfile, notes_dir, output_dir, id_locations_file)

            # Log failure inline at warning level
            case result do
              {:ok, _} ->
                :ok

              {:error, {:emacs_exit, code}} ->
                Logger.warning("  failed: #{Path.basename(orgfile)} (emacs exit code #{code})")
            end

            {orgfile, result}
          end,
          max_concurrency: max_concurrency,
          timeout: :infinity,
          ordered: false
        )
        |> Enum.map(fn {:ok, result} -> result end)

      # Clean up temp file
      if id_locations_file, do: File.rm(id_locations_file)

      {successes, failures} =
        Enum.split_with(results, fn
          {_, {:ok, _}} -> true
          {_, {:error, _}} -> false
        end)

      {:ok, length(successes), failures}
    end
  end

  # Build org-id locations database file by scanning all org files once
  defp build_id_locations(notes_dir) do
    id_file = Path.join(System.tmp_dir!(), "org-id-locations-#{:erlang.unique_integer([:positive])}")

    {_output, exit_code} =
      System.cmd(
        "emacs",
        [
          "--batch",
          "--eval", ~s[(setq org-id-locations-file "#{id_file}")],
          "--eval", ~s[(setq org-id-extra-files (directory-files-recursively "#{notes_dir}" "\\\\.org$"))],
          "--eval", "(org-id-update-id-locations)"
        ],
        stderr_to_stdout: true
      )

    if exit_code == 0 do
      Logger.debug("Built org-id locations database: #{id_file}")
      id_file
    else
      Logger.warning("Failed to build org-id locations database")
      nil
    end
  end

  defp get_concurrency do
    Application.get_env(:org_garden, :export_concurrency, @default_max_concurrency)
  end

  @doc """
  Compute the expected `.md` path for a given `.org` file.

  Uses the same section-mapping logic as ox-hugo: the relative directory
  of the `.org` file within `notes_dir` becomes the section directory
  under `content_dir`.

  ## Examples

      iex> OrgGarden.Export.expected_md_path("/notes/bus/emt.org", "/notes", "/out/content")
      "/out/content/bus/emt.md"

      iex> OrgGarden.Export.expected_md_path("/notes/top-level.org", "/notes", "/out/content")
      "/out/content/top-level.md"
  """
  @spec expected_md_path(String.t(), String.t(), String.t()) :: String.t()
  def expected_md_path(orgfile, notes_dir, content_dir) do
    section =
      orgfile
      |> Path.dirname()
      |> Path.relative_to(notes_dir)

    basename = Path.basename(orgfile, ".org") <> ".md"

    case section do
      "." -> Path.join(content_dir, basename)
      _ -> Path.join([content_dir, section, basename])
    end
  end
end