Improve how overlapping quotes are parsed in HTML elements

Blake-Madden · Dec 4, 2023 · 57b6b80 · 57b6b80
1 parent cd5219f
commit 57b6b80
Show file tree

Hide file tree

Showing 2 changed files with 60 additions and 14 deletions.
diff --git a/src/import/html_extract_text.cpp b/src/import/html_extract_text.cpp
@@ -1334,30 +1334,26 @@ namespace lily_of_the_valley
         else if (text[0] == L'<')
             { ++text; }
 
-        bool is_inside_of_quotes = false;
-        bool is_inside_of_single_quotes = false;
+        bool is_inside_of_double_quotes{ false };
+        bool is_inside_of_single_quotes{ false };
         long openTagCount{ 0 };
         while (text)
             {
             if (text[0] == 0)
                 { return nullptr; }
-            else if (text[0] == 0x22) // double quote
+            // flip the state of double or single quote if not inside of
+            // the other type of quotes
+            else if (!is_inside_of_single_quotes && text[0] == L'\"')
                 {
-                is_inside_of_quotes = !is_inside_of_quotes;
-                // whether this double quote ends a quote pair or starts a new one, turn this flag
-                // off. This means that a double quote can close a single quote.
-                is_inside_of_single_quotes = false;
+                is_inside_of_double_quotes = !is_inside_of_double_quotes;
                 }
-            // if a single quote already started a quote pair (and this is closing it) or
-            // we are not inside of a double quote then count single quotes
-            else if ((!is_inside_of_quotes || is_inside_of_single_quotes) && text[0] == 0x27) //single quote
+            else if (!is_inside_of_double_quotes && text[0] == L'\'')
                 {
-                is_inside_of_quotes = !is_inside_of_quotes;
-                is_inside_of_single_quotes = true;
+                is_inside_of_single_quotes = !is_inside_of_single_quotes;
                 }
-            else if (!is_inside_of_quotes && text[0] == L'<')
+            else if (!is_inside_of_double_quotes && !is_inside_of_single_quotes && text[0] == L'<')
                 { ++openTagCount; }
-            else if (!is_inside_of_quotes && text[0] == L'>')
+            else if (!is_inside_of_double_quotes && !is_inside_of_single_quotes && text[0] == L'>')
                 {
                 if (openTagCount == 0)
                     { return text; }

diff --git a/tests/htmlimporttests.cpp b/tests/htmlimporttests.cpp
@@ -1008,6 +1008,56 @@ TEST_CASE("HTML Parser", "[html import]")
         CHECK(std::wcscmp(p, L"List.    (pane)") == 0);
         delete[] text;
         }
+    SECTION("Embedded JS Quotes")
+        {
+        html_extract_text filter_html;
+        const wchar_t* text = LR"(<html>Hello <input type='submit' id='gform_submit_button_12' class='gform_button button' value='Submit'  onclick='if(window["gf_submitting_12"]){return false;}  window["gf_submitting_12"]=true;  ' onkeypress='if( event.keyCode == 13 ){ if(window["gf_submitting_12"]){return false;} window["gf_submitting_12"]=true;  jQuery("#gform_12").trigger("submit",[true]); }' />there</html>)";
+        const wchar_t* p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello \n\nthere") == p);
+        }
+    SECTION("Embedded JS Quotes 2")
+        {
+        html_extract_text filter_html;
+        const wchar_t* text = LR"(<html>Hello <input type="submit" id="gform_submit_button_12" class="gform_button button" value="Submit"  onclick="if(window['gf_submitting_12']){return false;}  window['gf_submitting_12']=true;  " onkeypress="if( event.keyCode == 13 ){ if(window["gf_submitting_12"]){return false;} window["gf_submitting_12"]=true;  jQuery('#gform_12').trigger('submit', [true]); }" />there</html>)";
+        const wchar_t* p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello \n\nthere") == p);
+        }
+    SECTION("Elements with Quotes")
+        {
+        html_extract_text filter_html;
+        const wchar_t* text = LR"(Hello <a hef="submit">there)";
+        const wchar_t* p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello there") == p);
+
+        text = LR"(Hello <a hef='submit'>there)";
+        p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello there") == p);
+
+        text = LR"(Hello <a hef='su"b"mit'>there)";
+        p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello there") == p);
+
+        text = LR"(Hello <a hef="sub'm'it">there)";
+        p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello there") == p);
+
+        text = LR"(Hello <a hef='submit' name="name" value="5">there)";
+        p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello there") == p);
+
+        text = LR"(Hello <a hef='su<>bmit' name="na<>me" value="<5">there)";
+        p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello there") == p);
+
+        // Mismatch, will be trash. Just read what we can.
+        text = LR"(Hello <a hef='submit">there)";
+        p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello ") == p);
+
+        text = LR"(Hello <a hef="submit'>there)";
+        p = filter_html(text, std::wcslen(text), true, false);
+        CHECK(std::wstring(L"Hello ") == p);
+        }
     SECTION("Compare Entities")
         {
         const wchar_t* text = L"<span>List.</span> \r\n (pane)";