postcss

defmodule PostCSS.Parser do
  @moduledoc """
  CSS Parser for PostCSS.

  Converts tokens from the tokenizer into an Abstract Syntax Tree (AST).
  """

  alias PostCSS.AtRule
  alias PostCSS.Comment
  alias PostCSS.CssSyntaxError
  alias PostCSS.Declaration
  alias PostCSS.Raws
  alias PostCSS.Root
  alias PostCSS.Rule
  alias PostCSS.Tokenizer

  @doc """
  Parses CSS string into a PostCSS AST.
  """
  def parse(css) when is_binary(css) do
    css
    |> Tokenizer.tokenize()
    |> parse_tokens()
  end

  defp parse_tokens(tokens) do
    tokens
    |> filter_meaningful_tokens()
    |> parse_root()
  end

  defp filter_meaningful_tokens(tokens) do
    # Keep spaces for selector parsing, remove only standalone spaces not between words
    tokens
  end

  defp parse_root(tokens) do
    {nodes, remaining} = parse_nodes(tokens)

    # Collect any trailing whitespace
    trailing_whitespace = collect_trailing_whitespace(remaining, "")

    unless Enum.empty?(remaining) and trailing_whitespace == "" do
      # If we have trailing whitespace, that's fine
      # If we have other tokens, that's an error
      non_whitespace_remaining =
        Enum.reject(remaining, fn
          {:space, _, _} -> true
          {:space, _, _, _} -> true
          _ -> false
        end)

      unless Enum.empty?(non_whitespace_remaining) do
        raise CssSyntaxError.new("Unexpected tokens at end of input")
      end
    end

    root = %Root{nodes: nodes}

    # Add trailing whitespace to raws.after if present
    if trailing_whitespace != "" do
      Raws.put(root, :after, trailing_whitespace)
    else
      root
    end
  end

  defp collect_trailing_whitespace([], acc), do: acc

  defp collect_trailing_whitespace([{:space, space, _} | rest], acc) do
    collect_trailing_whitespace(rest, acc <> space)
  end

  defp collect_trailing_whitespace([{:space, space, _, _} | rest], acc) do
    collect_trailing_whitespace(rest, acc <> space)
  end

  defp collect_trailing_whitespace(_, acc), do: acc

  defp parse_nodes(tokens) do
    parse_nodes_with_whitespace(tokens, [], "")
  end

  defp parse_nodes_with_whitespace([], acc, whitespace) do
    # If we have trailing whitespace, preserve it as remaining space tokens
    remaining =
      if whitespace != "" do
        [{:space, whitespace, 0}]
      else
        []
      end

    {Enum.reverse(acc), remaining}
  end

  defp parse_nodes_with_whitespace([{:close_brace, _, _} | _] = tokens, acc, _whitespace) do
    {Enum.reverse(acc), tokens}
  end

  defp parse_nodes_with_whitespace([{:space, space, _} | rest], acc, whitespace) do
    # Collect whitespace for the next node
    parse_nodes_with_whitespace(rest, acc, whitespace <> space)
  end

  defp parse_nodes_with_whitespace([{:space, space, _, _} | rest], acc, whitespace) do
    # Collect whitespace for the next node
    parse_nodes_with_whitespace(rest, acc, whitespace <> space)
  end

  defp parse_nodes_with_whitespace(tokens, acc, whitespace) do
    case parse_node(tokens) do
      {node, remaining} ->
        # Add collected whitespace to node's raws.before
        node_with_raws =
          if whitespace != "" do
            Raws.put(node, :before, whitespace)
          else
            node
          end

        parse_nodes_with_whitespace(remaining, [node_with_raws | acc], "")

      :error ->
        raise CssSyntaxError.new("Failed to parse node")
    end
  end

  defp parse_node(tokens) do
    case tokens do
      # Comment
      [{:comment, comment_text, _, _} | rest] ->
        parse_comment(comment_text, rest)

      # At-rule: @import, @media, etc.
      [{:at_word, at_rule_text, _, _} | rest] ->
        parse_at_rule(at_rule_text, rest)

      # Simple declaration: word : value
      [{:word, prop, _, _} | rest] ->
        rest = skip_spaces(rest)

        case rest do
          [{:colon, _, _} | value_tokens] ->
            # This is a declaration
            case parse_declaration_value(value_tokens) do
              {value, important, remaining} ->
                decl = %Declaration{prop: prop, value: value, important: important}
                {decl, remaining}

              :error ->
                raise CssSyntaxError.new("Failed to parse declaration value")
            end

          _ ->
            # This might be a selector, collect all tokens until we find an open brace
            case collect_selector_tokens(tokens) do
              {[_ | _] = selector_tokens, [{:open_brace, _, _} | rule_rest]} ->
                selector = extract_selector_from_tokens(selector_tokens)
                {declarations, remaining} = parse_declarations(rule_rest, [])

                case remaining do
                  [{:close_brace, _, _} | rest_tokens] ->
                    rule = %Rule{selector: selector, nodes: declarations}
                    {rule, rest_tokens}

                  [] ->
                    raise CssSyntaxError.new("Unclosed rule: missing '}'")

                  _ ->
                    raise CssSyntaxError.new("Expected '}' after rule declarations")
                end

              {[_ | _] = tokens_found, _} ->
                [first_token | _] = tokens_found
                raise CssSyntaxError.new("Unexpected token: #{inspect(first_token)}")

              _ ->
                :error
            end
        end

      # Other tokens - try to parse as selector
      _ ->
        case collect_selector_tokens(tokens) do
          {[_ | _] = selector_tokens, [{:open_brace, _, _} | rule_rest]} ->
            selector = extract_selector_from_tokens(selector_tokens)
            {declarations, remaining} = parse_declarations(rule_rest, [])

            case remaining do
              [{:close_brace, _, _} | rest_tokens] ->
                rule = %Rule{selector: selector, nodes: declarations}
                {rule, rest_tokens}

              [] ->
                raise CssSyntaxError.new("Unclosed rule: missing '}'")

              _ ->
                raise CssSyntaxError.new("Expected '}' after rule declarations")
            end

          {[_ | _] = tokens_found, _} ->
            [first_token | _] = tokens_found
            raise CssSyntaxError.new("Unexpected token: #{inspect(first_token)}")

          _ ->
            :error
        end
    end
  end

  defp collect_selector_tokens(tokens) do
    collect_selector_tokens(tokens, [])
  end

  defp collect_selector_tokens([], acc), do: {Enum.reverse(acc), []}

  defp collect_selector_tokens([{:open_brace, _, _} | _] = tokens, acc),
    do: {Enum.reverse(acc), tokens}

  defp collect_selector_tokens([token | rest], acc),
    do: collect_selector_tokens(rest, [token | acc])

  defp extract_selector_from_tokens(tokens) do
    tokens
    |> Enum.map(fn
      {:word, value, _, _} -> value
      {:at_word, value, _, _} -> value
      {:space, value, _} -> value
      {:space, value, _, _} -> value
      {:comma, value, _} -> value
      {:colon, value, _} -> value
      {_, value, _} -> value
      {_, value, _, _} -> value
    end)
    |> Enum.join("")
    |> String.trim()
  end

  defp parse_declarations(tokens, acc) do
    parse_declarations_with_whitespace(tokens, acc, "")
  end

  defp parse_declarations_with_whitespace(tokens, acc, whitespace) do
    case tokens do
      [{:close_brace, _, _} | _] ->
        {Enum.reverse(acc), tokens}

      [] ->
        {Enum.reverse(acc), []}

      [{:space, space, _} | rest] ->
        # Collect whitespace for the next node
        parse_declarations_with_whitespace(rest, acc, whitespace <> space)

      [{:space, space, _, _} | rest] ->
        # Collect whitespace for the next node
        parse_declarations_with_whitespace(rest, acc, whitespace <> space)

      [{:comment, comment_text, _, _} | rest] ->
        # Handle comments inside rule blocks
        {comment, remaining} = parse_comment(comment_text, rest)

        comment_with_raws =
          if whitespace != "" do
            Raws.put(comment, :before, whitespace)
          else
            comment
          end

        parse_declarations_with_whitespace(remaining, [comment_with_raws | acc], "")

      _ ->
        case parse_declaration(tokens) do
          {decl, remaining} ->
            # Add collected whitespace to declaration's raws.before
            decl_with_raws =
              if whitespace != "" do
                Raws.put(decl, :before, whitespace)
              else
                decl
              end

            # Skip semicolon if present, but preserve whitespace for next declaration
            remaining = skip_semicolon_preserve_whitespace(remaining)
            parse_declarations_with_whitespace(remaining, [decl_with_raws | acc], "")

          :error ->
            {Enum.reverse(acc), tokens}
        end
    end
  end

  defp parse_declaration([{:word, prop, _, _} | rest]) do
    # Collect spaces before colon
    {before_colon, after_spaces} = collect_spaces_before_colon(rest, "")

    case after_spaces do
      [{:colon, _, _} | value_tokens] ->
        # Collect spaces after colon
        {after_colon, value_tokens} = collect_spaces(value_tokens, "")

        case parse_declaration_value(value_tokens) do
          {value, important, remaining} ->
            # Build the between string (spaces + colon + spaces)
            between = before_colon <> ":" <> after_colon

            decl = %Declaration{prop: prop, value: value, important: important}
            decl_with_raws = Raws.put(decl, :between, between)
            {decl_with_raws, remaining}

          :error ->
            :error
        end

      _ ->
        :error
    end
  end

  defp collect_spaces_before_colon([{:space, space, _} | rest], acc) do
    collect_spaces_before_colon(rest, acc <> space)
  end

  defp collect_spaces_before_colon([{:space, space, _, _} | rest], acc) do
    collect_spaces_before_colon(rest, acc <> space)
  end

  defp collect_spaces_before_colon(tokens, acc), do: {acc, tokens}

  defp collect_spaces([{:space, space, _} | rest], acc) do
    collect_spaces(rest, acc <> space)
  end

  defp collect_spaces([{:space, space, _, _} | rest], acc) do
    collect_spaces(rest, acc <> space)
  end

  defp collect_spaces(tokens, acc), do: {acc, tokens}

  defp skip_spaces([{:space, _, _} | rest]), do: skip_spaces(rest)
  defp skip_spaces([{:space, _, _, _} | rest]), do: skip_spaces(rest)
  defp skip_spaces(tokens), do: tokens

  defp parse_declaration_value(tokens) do
    case collect_value_tokens(tokens, []) do
      {value_tokens, remaining} ->
        {value, important} = process_value_tokens(value_tokens)
        {value, important, remaining}

      :error ->
        :error
    end
  end

  defp collect_value_tokens([], acc), do: {Enum.reverse(acc), []}

  defp collect_value_tokens([{:semicolon, _, _} | _] = tokens, acc) do
    {Enum.reverse(acc), tokens}
  end

  defp collect_value_tokens([{:close_brace, _, _} | _] = tokens, acc) do
    {Enum.reverse(acc), tokens}
  end

  defp collect_value_tokens([token | rest], acc) do
    collect_value_tokens(rest, [token | acc])
  end

  defp process_value_tokens(tokens) do
    # Check for !important
    case Enum.reverse(tokens) do
      [{:word, "!important", _, _} | rest_tokens] ->
        value = extract_value_from_tokens(Enum.reverse(rest_tokens))
        {value, true}

      _ ->
        value = extract_value_from_tokens(tokens)
        {value, false}
    end
  end

  defp extract_value_from_tokens(tokens) do
    tokens
    |> Enum.map(fn
      {:word, value, _, _} -> value
      {:string, value, _, _} -> value
      {:colon, value, _} -> value
      {:at_word, value, _, _} -> value
      {_, value, _} -> value
      {_, value, _, _} -> value
    end)
    |> Enum.join("")
    |> String.trim()
  end

  # Skip semicolon but preserve whitespace (for use in parse_declarations_with_whitespace)
  defp skip_semicolon_preserve_whitespace([{:semicolon, _, _} | rest]), do: rest
  defp skip_semicolon_preserve_whitespace(tokens), do: tokens

  # Determine if an at-rule should contain declarations vs rules
  # Based on common CSS at-rules that typically contain declarations
  defp should_contain_declarations?(name) do
    name in [
      "font-face",
      "page",
      "counter-style",
      "font-feature-values",
      "property"
    ]
  end

  defp parse_at_rule(at_rule_text, tokens) do
    # Extract the rule name (remove the @ symbol)
    name = String.slice(at_rule_text, 1..-1//1)

    # Collect parameters until we hit a semicolon or opening brace
    {param_tokens, remaining} = collect_at_rule_params(tokens, [])

    # Convert parameter tokens to a string
    params =
      if Enum.empty?(param_tokens) do
        nil
      else
        extract_params_from_tokens(param_tokens)
      end

    case remaining do
      [{:semicolon, _, _} | rest] ->
        # Simple at-rule like @import url(...);
        at_rule = %AtRule{name: name, params: params, nodes: []}
        {at_rule, rest}

      [{:open_brace, _, _} | rest] ->
        # Block at-rule like @media (...) { ... }
        # Determine if this at-rule should contain declarations or rules
        {nodes, remaining} =
          if should_contain_declarations?(name) do
            parse_declarations_with_whitespace(rest, [], "")
          else
            parse_nodes_with_whitespace(rest, [], "")
          end

        case remaining do
          [{:close_brace, _, _} | rest_tokens] ->
            at_rule = %AtRule{name: name, params: params, nodes: nodes}
            {at_rule, rest_tokens}

          [] ->
            raise CssSyntaxError.new("Unclosed at-rule: missing '}'")

          _ ->
            raise CssSyntaxError.new("Expected '}' after at-rule content")
        end

      [] ->
        # At-rule at end of input
        at_rule = %AtRule{name: name, params: params, nodes: []}
        {at_rule, []}

      _ ->
        raise CssSyntaxError.new("Expected ';' or '{' after at-rule")
    end
  end

  defp collect_at_rule_params([], acc), do: {Enum.reverse(acc), []}

  defp collect_at_rule_params([{:semicolon, _, _} | _] = tokens, acc),
    do: {Enum.reverse(acc), tokens}

  defp collect_at_rule_params([{:open_brace, _, _} | _] = tokens, acc),
    do: {Enum.reverse(acc), tokens}

  defp collect_at_rule_params([token | rest], acc),
    do: collect_at_rule_params(rest, [token | acc])

  defp extract_params_from_tokens(tokens) do
    tokens
    |> Enum.map(fn
      {:word, value, _, _} -> value
      {:string, value, _, _} -> value
      {:open_paren, value, _} -> value
      {:close_paren, value, _} -> value
      {:space, value, _} -> value
      {:space, value, _, _} -> value
      {:comma, value, _} -> value
      {:colon, value, _} -> value
      {_, value, _} -> value
      {_, value, _, _} -> value
    end)
    |> Enum.join("")
    |> String.trim()
  end

  defp parse_comment(comment_text, remaining_tokens) do
    # Extract comment content (remove /* and */)
    # comment_text is like "/* comment */"
    text =
      comment_text
      # Remove /* and */
      |> String.slice(2..-3//1)

    # Parse whitespace like PostCSS does
    if String.trim(text) == "" do
      # Empty comment, preserve all whitespace as left
      comment = %Comment{text: "", raws: %{left: text, right: ""}}
      {comment, remaining_tokens}
    else
      # Parse leading and trailing whitespace
      case Regex.run(~r/^(\s*)(.*?\S)(\s*)$/s, text) do
        [_, left, content, right] ->
          comment = %Comment{text: content, raws: %{left: left, right: right}}
          {comment, remaining_tokens}

        nil ->
          # Fallback if regex doesn't match
          comment = %Comment{text: text, raws: %{}}
          {comment, remaining_tokens}
      end
    end
  end
end