add Recovery.Attributes module

lib/feed_reader/recovery/attributes.ex

@@ -0,0 +1,98 @@
+defmodule FeedReader.Recovery.Attributes do
+  @moduledoc """
+  Repairs malformed quotes in XML attributes.
+
+  A small number of feeds place literal quotes inside a quoted attribute value.
+  This module repairs those quotes only within one complete markup element, so
+  valid quotes in neighboring elements cannot be rewritten accidentally.
+  Incomplete elements are retained as state for the next input chunk.
+  """
+
+  @doc "Repairs literal quotes in streamed XML attributes."
+  def normalize(input) do
+    Stream.transform(
+      input,
+      fn -> <<>> end,
+      fn chunk, pending ->
+        {output, pending} = scan(pending <> chunk, [])
+        {if(output == [], do: [], else: [IO.iodata_to_binary(Enum.reverse(output))]), pending}
+      end,
+      fn
+        <<>> -> []
+        pending -> [pending]
+      end
+    )
+  end
+
+  defp scan(<<>>, output), do: {output, <<>>}
+
+  defp scan(data, output) do
+    case :binary.match(data, "<") do
+      :nomatch ->
+        {[data | output], <<>>}
+
+      {position, 1} when position > 0 ->
+        {prefix, rest} = :erlang.split_binary(data, position)
+        scan(rest, [prefix | output])
+
+      {0, 1} ->
+        case markup_end(data) do
+          :incomplete -> {output, data}
+          {markup, rest} -> scan(rest, [repair(markup) | output])
+        end
+    end
+  end
+
+  defp markup_end(<<"<![CDATA[", _rest::binary>> = data), do: special_end(data, "]]>")
+  defp markup_end(<<"<!--", _rest::binary>> = data), do: special_end(data, "-->")
+
+  defp markup_end(data) do
+    case :binary.match(data, ">") do
+      :nomatch -> :incomplete
+      {position, 1} -> :erlang.split_binary(data, position + 1)
+    end
+  end
+
+  defp special_end(data, closing) do
+    case :binary.match(data, closing) do
+      :nomatch -> :incomplete
+      {position, length} -> :erlang.split_binary(data, position + length)
+    end
+  end
+
+  defp repair(<<"<![CDATA[", _rest::binary>> = markup), do: markup
+  defp repair(<<"<!--", _rest::binary>> = markup), do: markup
+  defp repair(<<"<", rest::binary>>), do: "<" <> repair_tag(rest, nil, false, [])
+  defp repair_tag(<<>>, _quote, _expect, output), do: IO.iodata_to_binary(Enum.reverse(output))
+
+  defp repair_tag(<<char, rest::binary>>, nil, expect, output) when char in [32, 9, 10, 13],
+    do: repair_tag(rest, nil, expect, [<<char>> | output])
+
+  defp repair_tag(<<char, rest::binary>>, nil, true, output) when char in [34, 39],
+    do: repair_tag(rest, {char, false}, false, [<<char>> | output])
+
+  defp repair_tag(<<char, rest::binary>>, nil, _expect, output),
+    do: repair_tag(rest, nil, char == ?=, [<<char>> | output])
+
+  defp repair_tag(<<quote, rest::binary>>, {quote, has_content}, _expect, output) do
+    cond do
+      delimiter?(rest) ->
+        repair_tag(rest, nil, false, [<<quote>> | output])
+
+      empty_quote?(rest, quote, has_content) ->
+        <<_quote, after_quote::binary>> = rest
+        repair_tag(after_quote, nil, false, [<<quote>> | output])
+
+      true ->
+        repair_tag(rest, {quote, true}, false, ["&quot;" | output])
+    end
+  end
+
+  defp repair_tag(<<char, rest::binary>>, {quote, _has_content}, _expect, output),
+    do: repair_tag(rest, {quote, true}, false, [<<char>> | output])
+
+  defp delimiter?(<<char, _rest::binary>>) when char in [32, 9, 10, 13, 47, 62, 63], do: true
+  defp delimiter?(_rest), do: false
+  defp empty_quote?(<<quote, rest::binary>>, quote, false), do: delimiter?(rest)
+  defp empty_quote?(_rest, _quote, _has_content), do: false
+end

test/feed_reader/recovery/attributes_test.exs

@@ -0,0 +1,48 @@
+defmodule FeedReader.Recovery.AttributesTest do
+  use ExUnit.Case, async: true
+
+  alias FeedReader.Recovery.Attributes
+
+  test "preserves valid attributes" do
+    xml = ~s(<link href="https://example.com" rel="alternate" />)
+
+    assert normalize([xml]) == xml
+  end
+
+  test "repairs literal double quotes inside an attribute value" do
+    xml = ~s(<link title="How to "Sustain Heroku"" />)
+
+    assert normalize([xml]) == ~s(<link title="How to &quot;Sustain Heroku&quot;" />)
+  end
+
+  test "repairs malformed attributes split across chunks" do
+    chunks = [~s(<link title="How to "Sustain), ~s( Heroku"" />)]
+
+    assert normalize(chunks) == ~s(<link title="How to &quot;Sustain Heroku&quot;" />)
+  end
+
+  test "preserves empty attributes and single-quoted attributes" do
+    xml = ~s(<source url="" label='A "quoted" source' />)
+
+    assert normalize([xml]) == xml
+  end
+
+  test "does not repair quotes in CDATA or comments" do
+    xml = ~s(<![CDATA[<p title="A "quoted" value">Text</p>]]><!-- A "quoted" comment -->)
+
+    assert normalize([xml]) == xml
+  end
+
+  test "does not repair across adjacent elements" do
+    xml = ~s(<source url="">Source</source><category domain="region">News</category>)
+
+    assert normalize([xml]) == xml
+  end
+
+  test "preserves text outside markup" do
+    assert normalize(["before ", ~s(<item title="Story" />), " after"]) ==
+             ~s(before <item title="Story" /> after)
+  end
+
+  defp normalize(chunks), do: chunks |> Attributes.normalize() |> Enum.join()
+end