salvage code actions from experimental server

Author: sheldak

apps/language_server/lib/language_server/code_unit.ex

@@ -0,0 +1,180 @@
+defmodule ElixirLS.LanguageServer.CodeUnit do
+  @moduledoc """
+  Code unit and offset conversions
+
+  The LSP protocol speaks in positions, which defines where something happens in a document.
+  Positions have a start and an end, which are defined as code unit _offsets_ from the beginning
+  of a line. this module helps to convert between utf8, which most of the world speaks
+  natively, and utf16, which has been forced upon us by microsoft.
+
+  Converting between offsets and code units is 0(n), and allocations only happen if a
+  multi-byte character is detected, at which point, only that character is allocated.
+  This exploits the fact that most source code consists of ascii characters, with at best,
+  sporadic multi-byte characters in it. Thus, the vast majority of documents will not require
+  any allocations at all.
+  """
+  @type utf8_code_unit :: non_neg_integer()
+  @type utf16_code_unit :: non_neg_integer()
+  @type utf8_offset :: non_neg_integer()
+  @type utf16_offset :: non_neg_integer()
+
+  @type error :: {:error, :misaligned} | {:error, :out_of_bounds}
+
+  # public
+
+  @doc """
+  Converts a utf8 character offset into a utf16 character offset. This implementation
+  clamps the maximum size of an offset so that any initial character position can be
+  passed in and the offset returned will reflect the end of the line.
+  """
+  @spec utf16_offset(String.t(), utf8_offset()) :: utf16_offset()
+  def utf16_offset(binary, character_position) do
+    do_utf16_offset(binary, character_position, 0)
+  end
+
+  @doc """
+  Converts a utf16 character offset into a utf8 character offset. This implementation
+  clamps the maximum size of an offset so that any initial character position can be
+  passed in and the offset returned will reflect the end of the line.
+  """
+  @spec utf8_offset(String.t(), utf16_offset()) :: utf8_offset()
+  def utf8_offset(binary, character_position) do
+    do_utf8_offset(binary, character_position, 0)
+  end
+
+  @spec to_utf8(String.t(), utf16_code_unit()) :: {:ok, utf8_code_unit()} | error
+  def to_utf8(binary, utf16_unit) do
+    do_to_utf8(binary, utf16_unit, 0)
+  end
+
+  @spec to_utf16(String.t(), utf8_code_unit()) :: {:ok, utf16_code_unit()} | error
+  def to_utf16(binary, utf16_unit) do
+    do_to_utf16(binary, utf16_unit, 0)
+  end
+
+  def count(:utf16, binary) do
+    do_count_utf16(binary, 0)
+  end
+
+  # Private
+
+  # UTF-16
+
+  def do_count_utf16(<<>>, count) do
+    count
+  end
+
+  def do_count_utf16(<<c, rest::binary>>, count) when c < 128 do
+    do_count_utf16(rest, count + 1)
+  end
+
+  def do_count_utf16(<<c::utf8, rest::binary>>, count) do
+    increment =
+      <<c::utf16>>
+      |> byte_size()
+      |> div(2)
+
+    do_count_utf16(rest, count + increment)
+  end
+
+  defp do_utf16_offset(_, 0, offset) do
+    offset
+  end
+
+  defp do_utf16_offset(<<>>, _, offset) do
+    # this clause pegs the offset at the end of the string
+    # no matter the character index
+    offset
+  end
+
+  defp do_utf16_offset(<<c, rest::binary>>, remaining, offset) when c < 128 do
+    do_utf16_offset(rest, remaining - 1, offset + 1)
+  end
+
+  defp do_utf16_offset(<<c::utf8, rest::binary>>, remaining, offset) do
+    s = <<c::utf8>>
+    increment = utf16_size(s)
+    do_utf16_offset(rest, remaining - 1, offset + increment)
+  end
+
+  defp do_to_utf16(_, 0, utf16_unit) do
+    {:ok, utf16_unit}
+  end
+
+  defp do_to_utf16(_, utf8_unit, _) when utf8_unit < 0 do
+    {:error, :misaligned}
+  end
+
+  defp do_to_utf16(<<>>, _remaining, _utf16_unit) do
+    {:error, :out_of_bounds}
+  end
+
+  defp do_to_utf16(<<c, rest::binary>>, utf8_unit, utf16_unit) when c < 128 do
+    do_to_utf16(rest, utf8_unit - 1, utf16_unit + 1)
+  end
+
+  defp do_to_utf16(<<c::utf8, rest::binary>>, utf8_unit, utf16_unit) do
+    utf8_string = <<c::utf8>>
+    increment = utf16_size(utf8_string)
+    decrement = byte_size(utf8_string)
+
+    do_to_utf16(rest, utf8_unit - decrement, utf16_unit + increment)
+  end
+
+  defp utf16_size(binary) when is_binary(binary) do
+    binary
+    |> :unicode.characters_to_binary(:utf8, :utf16)
+    |> byte_size()
+    |> div(2)
+  end
+
+  # UTF-8
+
+  defp do_utf8_offset(_, 0, offset) do
+    offset
+  end
+
+  defp do_utf8_offset(<<>>, _, offset) do
+    # this clause pegs the offset at the end of the string
+    # no matter the character index
+    offset
+  end
+
+  defp do_utf8_offset(<<c, rest::binary>>, remaining, offset) when c < 128 do
+    do_utf8_offset(rest, remaining - 1, offset + 1)
+  end
+
+  defp do_utf8_offset(<<c::utf8, rest::binary>>, remaining, offset) do
+    s = <<c::utf8>>
+    increment = utf8_size(s)
+    decrement = utf16_size(s)
+    do_utf8_offset(rest, remaining - decrement, offset + increment)
+  end
+
+  defp do_to_utf8(_, 0, utf8_unit) do
+    {:ok, utf8_unit}
+  end
+
+  defp do_to_utf8(_, utf_16_units, _) when utf_16_units < 0 do
+    {:error, :misaligned}
+  end
+
+  defp do_to_utf8(<<>>, _remaining, _utf8_unit) do
+    {:error, :out_of_bounds}
+  end
+
+  defp do_to_utf8(<<c, rest::binary>>, utf16_unit, utf8_unit) when c < 128 do
+    do_to_utf8(rest, utf16_unit - 1, utf8_unit + 1)
+  end
+
+  defp do_to_utf8(<<c::utf8, rest::binary>>, utf16_unit, utf8_unit) do
+    utf8_code_units = byte_size(<<c::utf8>>)
+    utf16_code_units = utf16_size(<<c::utf8>>)
+
+    do_to_utf8(rest, utf16_unit - utf16_code_units, utf8_unit + utf8_code_units)
+  end
+
+  defp utf8_size(binary) when is_binary(binary) do
+    byte_size(binary)
+  end
+end

apps/language_server/lib/language_server/protocol.ex

@@ -209,6 +209,17 @@ defmodule ElixirLS.LanguageServer.Protocol do
     end
   end
 
+  defmacro code_action_req(id, uri, diagnostics) do
+    quote do
+      request(unquote(id), "textDocument/codeAction", %{
+        "context" => %{"diagnostics" => unquote(diagnostics)},
+        "textDocument" => %{
+          "uri" => unquote(uri)
+        }
+      })
+    end
+  end
+
   # Other utilities
 
   defmacro range(start_line, start_character, end_line, end_character) do

apps/language_server/lib/language_server/protocol/text_edit.ex

@@ -4,6 +4,9 @@ defmodule ElixirLS.LanguageServer.Protocol.TextEdit do
 
   For details see https://microsoft.github.io/language-server-protocol/specification#textEdit
   """
+
+  @type t :: %__MODULE__{}
+
   @derive JasonV.Encoder
   defstruct [:range, :newText]
 end

apps/language_server/lib/language_server/providers/code_action.ex

@@ -0,0 +1,12 @@
+defmodule ElixirLS.LanguageServer.Providers.CodeAction do
+  alias ElixirLS.LanguageServer.Providers.CodeAction.ReplaceRemoteFunction
+  alias ElixirLS.LanguageServer.Providers.CodeAction.ReplaceWithUnderscore
+
+  @code_actions [ReplaceRemoteFunction, ReplaceWithUnderscore]
+
+  def code_actions(source_file, uri, diagnostic) do
+    code_actions = Enum.flat_map(@code_actions, & &1.apply(source_file, uri, diagnostic))
+
+    {:ok, code_actions}
+  end
+end

apps/language_server/lib/language_server/providers/code_action/code_action_result.ex

@@ -0,0 +1,24 @@
+defmodule ElixirLS.LanguageServer.Providers.CodeAction.CodeActionResult do
+  alias ElixirLS.LanguageServer.Protocol.TextEdit
+
+  @type t :: %{
+          title: String.t(),
+          kind: String.t(),
+          edit: %{
+            changes: %{String.t() => TextEdit.t()}
+          }
+        }
+
+  @spec new(String.t(), String.t(), [TextEdit.t()], String.t()) :: t()
+  def new(title, kind, text_edits, uri) do
+    %{
+      :title => title,
+      :kind => kind,
+      :edit => %{
+        :changes => %{
+          uri => text_edits
+        }
+      }
+    }
+  end
+end

apps/language_server/lib/language_server/providers/code_action/helpers.ex

@@ -0,0 +1,53 @@
+defmodule ElixirLS.LanguageServer.Providers.CodeAction.Helpers do
+  alias ElixirLS.LanguageServer.Protocol.TextEdit
+  alias ElixirLS.LanguageServer.Providers.CodeMod.Ast
+  alias ElixirLS.LanguageServer.Providers.CodeMod.Text
+
+  @spec update_line(TextEdit.t(), non_neg_integer()) :: TextEdit.t()
+  def update_line(
+        %TextEdit{range: %{"start" => start_line, "end" => end_line}} = text_edit,
+        line_number
+      ) do
+    %TextEdit{
+      text_edit
+      | range: %{
+          "start" => %{start_line | "line" => line_number},
+          "end" => %{end_line | "line" => line_number}
+        }
+    }
+  end
+
+  @spec to_one_line_string(Ast.t()) :: {:ok, String.t()} | :error
+  def to_one_line_string(updated_ast) do
+    updated_ast
+    |> Ast.to_string()
+    # We're dealing with a single error on a single line.
+    # If the line doesn't compile (like it has a do with no end), ElixirSense
+    # adds additional lines to documents with errors. Also, in case of a one-line do,
+    # ElixirSense creates do with end from the AST.
+    |> maybe_recover_one_line_do(updated_ast)
+    |> Text.fetch_line(0)
+  end
+
+  @do_regex ~r/\s*do\s*/
+  defp maybe_recover_one_line_do(updated_text, {_name, context, _children} = _updated_ast) do
+    wrong_do_end_conditions = [
+      not Keyword.has_key?(context, :do),
+      not Keyword.has_key?(context, :end),
+      Regex.match?(@do_regex, updated_text),
+      String.ends_with?(updated_text, "\nend")
+    ]
+
+    if Enum.all?(wrong_do_end_conditions) do
+      updated_text
+      |> String.replace(@do_regex, ", do: ")
+      |> String.trim_trailing("\nend")
+    else
+      updated_text
+    end
+  end
+
+  defp maybe_recover_one_line_do(updated_text, _updated_ast) do
+    updated_text
+  end
+end