Issue 5870 - Implement the new ElemHideEmulation filter type

compiled/filter/ElemHideBase.cpp

Index: compiled/filter/ElemHideBase.cpp
===================================================================
--- a/compiled/filter/ElemHideBase.cpp
+++ b/compiled/filter/ElemHideBase.cpp
@@ -44,20 +44,112 @@
         delta++;
       else
         text[pos - delta] = text[pos];
     }
     selectorStart -= delta;
 
     text.reset(text, 0, len - delta);
   }
+
+  // Convert filter from the old syntax to the new.
+  OwnedString ConvertFilter(const String& text, String::size_type at)

[sergei, 2018/01/30 19:02:40]

I'm not sure that we need to build OwnedString for each filter, perhaps the
contructor of base classes should just support OwnedString&&. What do you think
about it? If we decide to add support of OwnedString&& in the Filter constructor
then IMO it should be in a separate commit and with the issue.

[hub, 2018/01/30 22:53:13]

Since we copy the string at the end in the Filter constructor we could have
chain of constructors that take an OwnedString&& so that ConvertFilter does the
copy or the conversion.

+  {
+    static const auto propsSelector = u"[-abp-properties="_str;
+    static const auto newPropsSelector = u":-abp-properties("_str;
+    auto selectorPos = text.find(propsSelector, at);
+    if (selectorPos != text.npos)
+    {
+      auto length = text.length();
+      auto properties = selectorPos + propsSelector.length();
+      String::value_type quote = 0;
+      bool escape = false;
+      String::size_type removed = 0; // how many chars we remove
+      String::size_type end = properties;
+      String::size_type quote_start = 0;
+      String::size_type quote_end = 0;
+      for (auto index = properties;
+           index < length && end == properties; index++)
+      {
+        if (escape)
+        {
+          escape = false;
+          continue;
+        }
+
+        auto c = text[index];
+        switch (c)
+        {
+        case '\\':
+          escape = true;
+          break;
+        case '"':
+        case '\'':
+          if (quote == 0)
+          {
+            quote = c;
+            quote_start = index + 1;
+          }
+          else if (quote == c)
+          {
+            // end of quoted.
+            quote = 0;
+            removed += 2;
+            quote_end = index;
+          }
+          break;
+        case ']':
+          if (quote == 0)
+            end = index + 1; // end of properties (after ])
+          break;
+        default:
+          break;
+        }
+      }
+
+      if (quote != 0)
+        quote_end = end - 1;
+      else if (quote_end <= quote_start)
+      {
+        // we likely didn't find a quoted content so we just take it as is.
+        quote_start = properties;
+        quote_end = end - 1;
+      }
+
+      OwnedString converted(length - removed);
+      String::size_type offset = 0;
+      std::memcpy(converted.data(), text.data(),
+                  selectorPos * sizeof(String::value_type));
+      offset += selectorPos;
+
+      std::memcpy(converted.data() + offset, newPropsSelector.data(),
+                  newPropsSelector.length() * sizeof(String::value_type));
+      offset += newPropsSelector.length();
+
+      std::memcpy(converted.data() + offset, text.data() + quote_start,
+                  (quote_end - quote_start) * sizeof(String::value_type));
+      offset += quote_end - quote_start;
+
+      std::memcpy(converted.data() + offset, u")", sizeof(String::value_type));
+      offset++;
+
+      std::memcpy(converted.data() + offset, text.data() + end,
+                  (length - end) * sizeof(String::value_type));
+      offset += (length - end) * sizeof(String::value_type);
+
+      return converted;
+    }
+
+    return OwnedString(text);
+  }
 }
 
 ElemHideBase::ElemHideBase(Type type, const String& text, const ElemHideData& data)
-    : ActiveFilter(type, text, false), mData(data)
+    : ActiveFilter(type, ConvertFilter(text, data.mSelectorStart), false),
+      mData(data)
 {
   if (mData.HasDomains())
     ParseDomains(mData.GetDomainsSource(mText), u',');
 }
 
 Filter::Type ElemHideBase::Parse(DependentString& text, ElemHideData& data)
 {
   StringScanner scanner(text);
@@ -84,19 +176,22 @@
         return Type::UNKNOWN;
       case u' ':
         seenSpaces = true;
         break;
     }
   }
 
   seenSpaces |= scanner.skip(u' ');
+  bool emulation = false;
   bool exception = scanner.skipOne(u'@');
   if (exception)
     seenSpaces |= scanner.skip(u' ');
+  else
+    emulation = scanner.skipOne(u'?');

[sergei, 2018/01/30 19:02:40]

It's actually different from "Filter.elemhideRegExp =
/^([^/*|@"!]*?)#([@?])?#(.+)$/;"
https://github.com/adblockplus/adblockpluscore/blob/ec65d0a80470fc2eacd9ef83e...
where it allows only one character @ or ? but here we allow "#@?#". I think it
should be fixed.

[hub, 2018/01/30 22:53:13]

#@?# is "rejected". actually according to issue 6234 it is made to a blocking
filter. But this is correct. Same as the regexp.

Adding a test.

 
   String::value_type next = scanner.next();
   if (next != u'#')
     return Type::UNKNOWN;
 
   // Selector part
 
   // Selector shouldn't be empty
@@ -109,17 +204,17 @@
   // We are done validating, now we can normalize whitespace and the domain part
   if (seenSpaces)
     NormalizeWhitespace(text, data.mDomainsEnd, data.mSelectorStart);
   DependentString(text, 0, data.mDomainsEnd).toLower();
 
   if (exception)
     return Type::ELEMHIDEEXCEPTION;
 
-  if (text.find(u"[-abp-properties="_str, data.mSelectorStart) != text.npos)
+  if (emulation)
     return Type::ELEMHIDEEMULATION;
 
   return Type::ELEMHIDE;
 }
 
 namespace
 {
   static constexpr String::value_type OPENING_CURLY_REPLACEMENT[] = u"\\7B ";