defmodule NativeElixirPdfUtilities.HtmlToPdf.HtmlParser do
  @moduledoc """
  Strict HTML parser for the native HTML-to-PDF renderer.

  Supports a strict subset of document-oriented HTML: structural
  html/head/body/style tags, paragraphs, headings, inline emphasis/color
  containers, div containers for flex/grid layouts, lists, links, tables,
  images, and CSS-targeting attributes. Unsupported or malformed markup returns
  an error instead of guessing at browser behavior.
  """

  @type text_node :: %{type: :text, text: String.t()}
  @type element_node :: %{
          type: :element,
          tag: String.t(),
          attributes: map(),
          children: [text_node() | element_node()]
        }
  @type dom_tree :: %{type: :document, children: [element_node()]}

  @structural_tags ~w(html head body style title meta)
  @block_tags ~w(article aside div footer header main nav p h1 h2 h3 h4 h5 h6 section ul ol table img)
  @inline_tags ~w(strong b em i span a br)
  @list_tags ~w(ul ol)
  @table_structure_tags ~w(table thead tbody tfoot tr)
  @table_content_tags ~w(caption th td)
  @void_tags ~w(meta br img)

  @doc """
  Parses an HTML binary into a renderer DOM tree.
  """
  @spec parse(String.t()) :: {:ok, dom_tree()} | {:error, :invalid_html | :unsupported_html}
  def parse(html) do
    case parse_detailed(html) do
      {:ok, dom} -> {:ok, dom}
      {:error, {reason, _detail}} -> {:error, reason}
    end
  end

  @doc """
  Parses an HTML binary and returns source-location details when parsing fails.
  """
  @spec parse_detailed(String.t()) ::
          {:ok, dom_tree()} | {:error, {:invalid_html | :unsupported_html, map()}}
  def parse_detailed(html) do
    case html do
      html when is_binary(html) ->
        case parse_document(html) do
          {:ok, dom} -> {:ok, dom}
          {:error, reason} -> {:error, {reason, html_error_detail(reason, html)}}
        end

      _ ->
        {:error,
         {:invalid_html,
          %{
            stage: :html,
            reason: :invalid_html,
            message: "HTML input must be a string"
          }}}
    end
  end

  defp parse_document(html) do
    tokens = html |> tokenize() |> List.flatten()

    with true <- Enum.join(tokens) == html,
         {:ok, children, []} <- parse_children(tokens, :document, nil, []),
         true <- children != [] do
      {:ok, %{type: :document, children: children}}
    else
      _ -> {:error, :unsupported_html}
    end
  end

  defp tokenize(html) do
    Regex.scan(~r/<[^>]*>|[^<]+/u, html)
  end

  defp parse_children(tokens, context, closing_tag, children) do
    case tokens do
      [] when is_nil(closing_tag) ->
        {:ok, children, []}

      [] ->
        {:error, :unsupported_html}

      [token | remaining] ->
        parse_child(token, remaining, context, closing_tag, children)
    end
  end

  defp parse_child(token, remaining, context, closing_tag, children) do
    cond do
      context == :document and doctype_token?(token) ->
        parse_children(remaining, context, closing_tag, children)

      String.starts_with?(token, "</") ->
        case parse_closing_tag(token) do
          {:ok, tag} when tag == closing_tag -> {:ok, children, remaining}
          _ -> {:error, :unsupported_html}
        end

      String.starts_with?(token, "<") ->
        with {:ok, tag, attributes, self_closing?} <- parse_opening_tag(token),
             true <- allowed_child?(context, tag),
             {:ok, element_children, rest} <- element_children(tag, remaining, self_closing?),
             true <- valid_element?(tag, element_children) do
          element = %{
            type: :element,
            tag: tag,
            attributes: attributes,
            children: element_children
          }

          parse_children(rest, context, closing_tag, children ++ [element])
        else
          _ -> {:error, :unsupported_html}
        end

      context == :document and String.trim(token) == "" ->
        parse_children(remaining, context, closing_tag, children)

      context in @list_tags and String.trim(token) == "" ->
        parse_children(remaining, context, closing_tag, children)

      context in @table_structure_tags and String.trim(token) == "" ->
        parse_children(remaining, context, closing_tag, children)

      context in ~w(html head body) and String.trim(token) == "" ->
        parse_children(remaining, context, closing_tag, children)

      context == :document ->
        {:error, :unsupported_html}

      true ->
        parse_children(
          remaining,
          context,
          closing_tag,
          children ++ [%{type: :text, text: decode_entities(token)}]
        )
    end
  end

  defp parse_opening_tag(token) do
    captures =
      Regex.named_captures(
        ~r/^<\s*(?<tag>[a-zA-Z][a-zA-Z0-9]*)\s*(?<attributes>[^<>]*?)(?<self_closing>\/?)\s*>$/u,
        token
      )

    case captures do
      %{"tag" => tag, "attributes" => attributes, "self_closing" => self_closing} ->
        tag = String.downcase(tag)

        with true <- supported_tag?(tag),
             {:ok, attributes} <- parse_attributes(attributes, tag) do
          {:ok, tag, attributes, self_closing == "/" or tag in @void_tags}
        else
          _ -> {:error, :unsupported_html}
        end

      _ ->
        {:error, :unsupported_html}
    end
  end

  defp parse_closing_tag(token) do
    case Regex.named_captures(~r/^<\/\s*(?<tag>[a-zA-Z][a-zA-Z0-9]*)\s*>$/u, token) do
      %{"tag" => tag} ->
        tag = String.downcase(tag)

        case supported_tag?(tag) do
          true -> {:ok, tag}
          false -> {:error, :unsupported_html}
        end

      _ ->
        {:error, :unsupported_html}
    end
  end

  defp parse_attributes(attributes, tag) do
    case String.trim(attributes) do
      "" ->
        {:ok, %{}}

      attributes ->
        captures =
          Regex.scan(~r/\s*([a-zA-Z][a-zA-Z0-9_-]*)\s*=\s*("[^"]*"|'[^']*')/u, attributes)

        parsed_source = captures |> Enum.map_join("", &List.first/1) |> String.trim()

        with true <- parsed_source == attributes,
             {:ok, parsed} <- attributes_to_map(captures, tag) do
          {:ok, parsed}
        else
          _ -> {:error, :unsupported_html}
        end
    end
  end

  defp attributes_to_map(captures, tag) do
    Enum.reduce_while(captures, {:ok, %{}}, fn [_, name, quoted_value], {:ok, acc} ->
      name = String.downcase(name)
      value = quoted_value |> String.slice(1..-2//1) |> decode_entities()

      case {name, tag, Map.has_key?(acc, name)} do
        {"style", _, false} ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {"id", _, false} ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {"class", _, false} ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {"lang", "html", false} ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {"href", "a", false} ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {"src", "img", false} ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {"alt", "img", false} ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {name, "meta", false} when name in ~w(charset content http-equiv name property) ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        {name, tag, false} when tag in ~w(td th) and name in ~w(colspan rowspan) ->
          {:cont, {:ok, Map.put(acc, name, value)}}

        _ ->
          {:halt, {:error, :unsupported_html}}
      end
    end)
  end

  defp element_children(tag, remaining, self_closing?) do
    case self_closing? do
      true -> {:ok, [], remaining}
      false -> parse_children(remaining, tag, tag, [])
    end
  end

  defp allowed_child?(context, tag) do
    cond do
      context == :document ->
        tag in @block_tags or tag in ~w(html head body style meta title)

      context == "html" ->
        tag in @block_tags or tag in ~w(head body style meta title)

      context == "head" ->
        tag in ~w(style meta title)

      context == "body" ->
        tag in @block_tags or tag in ~w(style meta title)

      context in @list_tags ->
        tag == "li"

      context == "style" ->
        false

      context == "title" ->
        false

      context == "table" ->
        tag in ~w(caption thead tbody tfoot tr)

      context in ~w(thead tbody tfoot) ->
        tag == "tr"

      context == "tr" ->
        tag in ~w(th td)

      context == "a" ->
        tag in @inline_tags and tag != "a"

      context in @table_content_tags ->
        tag in @inline_tags or tag in @block_tags

      context in ~w(article aside div footer header main nav section) ->
        tag in @block_tags or tag in @inline_tags or tag == "style"

      context == "li" or context in @block_tags or context in @inline_tags ->
        tag in @inline_tags
    end
  end

  defp supported_tag?(tag) do
    tag in @structural_tags or tag in @block_tags or tag in @inline_tags or tag == "li" or
      tag in ~w(caption thead tbody tfoot tr th td)
  end

  defp valid_element?(tag, children) do
    case tag do
      "head" ->
        Enum.all?(children, &match?(%{tag: tag} when tag in ["style", "meta", "title"], &1))

      "style" ->
        Enum.all?(children, &match?(%{type: :text}, &1))

      "title" ->
        Enum.all?(children, &match?(%{type: :text}, &1))

      tag when tag in @void_tags ->
        children == []

      "table" ->
        caption_count = Enum.count(children, &match?(%{tag: "caption"}, &1))
        caption_first? = caption_count == 0 or hd(children).tag == "caption"
        caption_count <= 1 and caption_first? and Enum.any?(children, &table_row_container?/1)

      tag when tag in ~w(thead tbody tfoot) ->
        children != [] and Enum.all?(children, &match?(%{tag: "tr"}, &1))

      "tr" ->
        children != [] and Enum.all?(children, &match?(%{tag: tag} when tag in ["th", "td"], &1))

      _ ->
        true
    end
  end

  defp table_row_container?(child) do
    case child do
      %{tag: tag} when tag in ~w(thead tbody tfoot tr) -> true
      _ -> false
    end
  end

  defp doctype_token?(token) do
    Regex.match?(~r/^<!doctype\s+html\s*>$/iu, token)
  end

  defp html_error_detail(reason, html) do
    source = html_error_source(html)
    {line, column} = source_location(html, source)

    %{
      stage: :html,
      reason: reason,
      message: html_error_message(reason, line, source, html),
      line: line,
      column: column,
      source: source
    }
  end

  defp html_error_source(html) do
    cond do
      String.trim(html) == "" ->
        ""

      malformed_tag = malformed_tag_source(html) ->
        malformed_tag

      unsupported = unsupported_tag_source(html) ->
        unsupported

      invalid_attribute = invalid_attribute_source(html) ->
        invalid_attribute

      true ->
        html |> String.split("\n") |> Enum.find(&(String.trim(&1) != "")) |> Kernel.||("")
    end
    |> String.trim()
  end

  defp malformed_tag_source(html) do
    html
    |> String.split("\n")
    |> Enum.find(fn line ->
      String.contains?(line, "<") and not Regex.match?(~r/<[^>]*>/u, line)
    end)
  end

  defp unsupported_tag_source(html) do
    Regex.scan(~r/<\/?\s*([a-zA-Z][a-zA-Z0-9]*)[^>]*>/u, html)
    |> Enum.find_value(fn [source, tag] ->
      case supported_tag?(String.downcase(tag)) do
        true -> nil
        false -> source
      end
    end)
  end

  defp invalid_attribute_source(html) do
    Regex.scan(~r/<\s*([a-zA-Z][a-zA-Z0-9]*)\s*([^<>]*?)(?:\/?)\s*>/u, html)
    |> Enum.find_value(fn [source, tag, attributes] ->
      tag = String.downcase(tag)

      case supported_tag?(tag) and String.trim(attributes) != "" and
             parse_attributes(attributes, tag) != {:ok, attribute_map(source)} do
        true -> source
        false -> nil
      end
    end)
  end

  defp attribute_map(source) do
    ~r/\s+([a-zA-Z][a-zA-Z0-9_-]*)\s*=\s*("[^"]*"|'[^']*')/u
    |> Regex.scan(source)
    |> Enum.reduce(%{}, fn [_, name, quoted_value], acc ->
      Map.put(
        acc,
        String.downcase(name),
        quoted_value |> String.slice(1..-2//1) |> decode_entities()
      )
    end)
  end

  defp html_error_message(_reason, line, source, html) do
    cond do
      String.trim(html) == "" ->
        "line 1: document is empty"

      unsupported_tag_source(source) ->
        ~s(line #{line}: HTML tag "#{source}" is unsupported)

      invalid_attribute_source(source) ->
        ~s(line #{line}: HTML tag "#{source}" has invalid, duplicate, or unsupported attributes)

      true ->
        ~s(line #{line}: HTML source "#{source}" is unsupported or malformed)
    end
  end

  defp source_location(source, snippet) do
    case snippet == "" do
      true ->
        {1, 1}

      false ->
        {index, _length} = :binary.match(source, snippet)
        prefix = binary_part(source, 0, index)
        lines = String.split(prefix, "\n", trim: false)
        line = length(lines)
        column = String.length(List.last(lines) || "") + 1
        {line, column}
    end
  end

  defp decode_entities(text) do
    text
    |> String.replace("&amp;", "&")
    |> String.replace("&nbsp;", " ")
    |> String.replace("&lt;", "<")
    |> String.replace("&gt;", ">")
    |> String.replace("&quot;", "\"")
    |> String.replace("&#39;", "'")
    |> String.replace("&apos;", "'")
  end
end