Extra unit test

Have a test with attributes where text was removed from the main body, and the attributes are still in the xML tree.
#977

Author: teusbenschop

editor/html2usfm.cpp

@@ -288,9 +288,11 @@ void Editor_Html2Usfm::close_element_node (const pugi::xml_node& node)
     if (!wla_class.empty()) {
       const std::string contents = m_word_level_attributes[wla_class];
       m_word_level_attributes.erase(wla_class);
-      // The vertical bar separates the canonical word from the attribute(s) following it.
-      current_line.append("|");
-      current_line.append(contents);
+      if (!contents.empty()) {
+        // The vertical bar separates the canonical word from the attribute(s) following it.
+        current_line.append("|");
+        current_line.append(contents);
+      }
     }
     // Add closing USFM, optionally closing embedded tags in reverse order.
     character_styles = filter::strings::array_diff (character_styles, classes);

unittests/usfm-html.cpp

@@ -1783,4 +1783,18 @@ TEST_F (usfm_html, word_level_attributes_linking)
 }
 
 
+TEST_F (usfm_html, word_level_attributes_removed)
+{
+  // This tests a situation where the \w ... is removed from the text body,
+  // but the word-level attributes themselves are still in the XML tree.
+  const std::string html = R"(<p class="b-p"><span>A</span><span class="i-w0wla2">B</span></p><p class="b-wordlevelattributes"> </p><p class="b-wla1">C="D"</p>)";
+  const std::string usfm = R"(\p A\w B\w*)";
+  Editor_Html2Usfm editor_html2usfm;
+  editor_html2usfm.load (html);
+  editor_html2usfm.stylesheet (styles_logic_standard_sheet ());
+  editor_html2usfm.run ();
+  EXPECT_EQ (usfm, editor_html2usfm.get());
+}
+
+
 #endif