Issue 5870 - Implement the new ElemHideEmulation filter type

compiled/filter/ElemHideBase.cpp

 /*
  * This file is part of Adblock Plus <https://adblockplus.org/>,
  * Copyright (C) 2006-present eyeo GmbH
  *
  * Adblock Plus is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License version 3 as
  * published by the Free Software Foundation.
  *
  * Adblock Plus is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  *
  * You should have received a copy of the GNU General Public License
  * along with Adblock Plus.  If not, see <http://www.gnu.org/licenses/>.
  */
 
 #include <cstring>
 
 #include "ElemHideBase.h"
 #include "../StringScanner.h"
+#include "../Utils.h"
 
 ABP_NS_USING
 
 namespace
 {
   void NormalizeWhitespace(DependentString& text, String::size_type& domainsEnd,
       String::size_type& selectorStart)
   {
     // For element hiding filters we only want to remove spaces preceding the
     // selector part. The positions we've determined already have to be adjusted
@@ skipping 12 matching lines @@
       // Only spaces before selectorStart position should be removed.
       if (pos < selectorStart && text[pos] == ' ')
         delta++;
       else
         text[pos - delta] = text[pos];
     }
     selectorStart -= delta;
 
     text.reset(text, 0, len - delta);
   }
+
+  static constexpr String::value_type ELEM_HIDE_DELIMITER[] = u"##";
+  static constexpr String::size_type ELEM_HIDE_DELIMITER_LEN = LENGTH_OF(ELEM_HIDE_DELIMITER);
+
+  static constexpr String::value_type ELEM_HIDE_EMULATION_DELIMITER[] = u"#?#";
+  static constexpr String::size_type ELEM_HIDE_EMULATION_DELIMITER_LEN = LENGTH_OF(ELEM_HIDE_EMULATION_DELIMITER);
+
+  static constexpr String::value_type OLD_PROPS_SELECTOR[] = u"[-abp-properties=";
+  static constexpr String::size_type OLD_PROPS_SELECTOR_LEN = LENGTH_OF(OLD_PROPS_SELECTOR);
+
+  static constexpr String::value_type PROPS_SELECTOR[] = u":-abp-properties(";
+  static constexpr String::size_type PROPS_SELECTOR_LEN = LENGTH_OF(PROPS_SELECTOR);
 }
 
 ElemHideBase::ElemHideBase(Type type, const String& text, const ElemHideData& data)
     : ActiveFilter(type, text, false), mData(data)
 {
   if (mData.HasDomains())
     ParseDomains(mData.GetDomainsSource(mText), u',');
 }
 
-Filter::Type ElemHideBase::Parse(DependentString& text, ElemHideData& data)
+Filter::Type ElemHideBase::Parse(DependentString& text, ElemHideData& data, bool& needConversion)
 {
   StringScanner scanner(text);
 
   // Domains part
   bool seenSpaces = false;
   while (!scanner.done())
   {
     String::value_type next = scanner.next();
     if (next == u'#')
     {
@@ skipping 10 matching lines @@
       case u'"':
       case u'!':
         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'?');
 
   String::value_type next = scanner.next();
   if (next != u'#')
     return Type::UNKNOWN;
 
   // Selector part
 
   // Selector shouldn't be empty
   seenSpaces |= scanner.skip(u' ');
   if (scanner.done())
     return Type::UNKNOWN;
 
   data.mSelectorStart = scanner.position() + 1;
 
   // 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();
 
+  // We still need to check the old syntax. It will be converted when
+  // we instantiate the filter.
+  if (!emulation &&
+      text.find(OLD_PROPS_SELECTOR, data.mSelectorStart, OLD_PROPS_SELECTOR_LEN) != text.npos)
+  {
+    needConversion = true;
+    emulation = !exception;
+  }
+
   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
 {
+  struct Range
+  {
+    String::size_type start;
+    String::size_type end;
+    String::size_type len() const
+    {
+        return end - start;
+    }
+    String::size_type byte_len() const
+    {
+      return len() * sizeof(String::value_type);
+    }
+  };
+}
+
+// Convert filter from the old syntax to the new.
+DependentString ElemHideBase::ConvertFilter(String& text, String::size_type& at)
+{
+  auto length = text.length();
+  Range prefix = {at, 0};
+  Range suffix = {at, length};
+  prefix.end = text.find(OLD_PROPS_SELECTOR, at, OLD_PROPS_SELECTOR_LEN);

[sergei, 2018/02/12 12:53:18]

not important but perhaps it would be better to say
Range prefix = {at, text.find(OLD_PROPS_SELECTOR, at, OLD_PROPS_SELECTOR_LEN)};

[hub, 2018/02/12 18:14:39]

Done.

+  if (prefix.end != text.npos)

[sergei, 2018/02/12 12:53:19]

What about returning DependentString(text); here in order to reduce the
indentation below?Subjectively it also improves readability, BTW `suffix` and
`length` will be declared after that `if`.

[hub, 2018/02/12 18:14:39]

Done.

+  {
+    Range properties = { prefix.end + OLD_PROPS_SELECTOR_LEN, 0 };
+    String::value_type quote = 0;
+    bool closing = false;
+    for (auto index = properties.start;
+         index < length && !closing; index++)
+    {
+      auto c = text[index];
+      switch (c)
+      {
+      case u'"':
+      case u'\'':
+        if (quote == 0)
+        {
+          // syntax error: we already have a quoted section.
+          if (properties.end)
+            return DependentString();
+
+          quote = c;
+          properties.start = index + 1;
+        }
+        else if (quote == c)
+        {
+          // end of quoted.
+          quote = 0;
+          properties.end = index;
+        }
+        break;
+      case u']':
+        if (quote == 0)
+        {
+          if (properties.end == 0)
+            return DependentString();
+          suffix.start = index + 1;
+          closing = true;

[sergei, 2018/02/12 12:53:19]

What about using of suffix.start instead of `closing`? We could initialize it to
zero. In addition to that if suffix.start is still zero after finishing the loop
then the filter is malformed because there is no closing ']'. Continuing without
taking it into account can result in bad sizes below.

[hub, 2018/02/12 18:14:39]

suffix.start is initialised to `at`. So I'll use that. And I'll return an
invalid string.

+        }
+        break;
+      default:
+        break;
+      }
+    }
+
+    String::size_type delimiter = text.find(ELEM_HIDE_DELIMITER, 0,
+                                            ELEM_HIDE_DELIMITER_LEN);
+    // +1 for the replacement of "##" by "#?#"
+    if (delimiter != text.npos)
+      at++;

[sergei, 2018/02/12 12:53:20]

BTW, if there is no such delimiter then is it a malformed filter string?

[hub, 2018/02/12 18:14:40]

no. #@# is valid too. We could check if it is #@# eventually but I don't think
it is part of this.

+    auto new_len = prefix.len() + suffix.len() + properties.len()
+      + PROPS_SELECTOR_LEN + 1 + at;

[sergei, 2018/02/12 12:53:19]

What about
at + prefix.len() + PROPS_SELECTOR_LEN + properties.len() + 1 /* ')' */ +
suffix.len()
?

[hub, 2018/02/12 18:14:39]

Done.

+

[sergei, 2018/02/12 12:53:18]

Could you please add the assert here that
new_len + 1 == length || (delimiter == text.npos && new_len + 2 == length)
just in case? Despite there are tests ensuring that nothing is lost, it just
looks scaring :)).

[hub, 2018/02/12 18:14:40]

Done.

+    DependentString converted(text, 0, new_len);
+
+    if (suffix.len())
+    {
+      new_len -= suffix.len();
+      std::memmove(converted.data() + new_len,
+                   text.data() + suffix.start,
+                   suffix.byte_len());
+    }
+    new_len--;
+    // here we need to move the properties before inserting the ')'
+    auto parens = new_len;
+    if (properties.len())
+    {
+      new_len -= properties.len();
+      std::memmove(converted.data() + new_len,
+                   text.data() + properties.start, properties.byte_len());
+    }
+    converted[parens] = u')';
+
+    new_len -= PROPS_SELECTOR_LEN;
+    std::memmove(converted.data() + new_len,

[sergei, 2018/02/12 12:53:18]

logically it should be memcpy but since memmove does not modify the memory
accessed by `src` argument it's fine.

The same for ELEM_HIDE_EMULATION_DELIMITER below.

[hub, 2018/02/12 18:14:39]

You are right. Changing it to memcpy().

+                 PROPS_SELECTOR,
+                 PROPS_SELECTOR_LEN * sizeof(String::value_type));
+    if (prefix.len())
+    {
+      new_len -= prefix.len();
+      std::memmove(converted.data() + new_len,
+                   text.data() + prefix.start, prefix.byte_len());
+    }
+
+    if (delimiter != String::npos)
+    {
+      std::memmove(converted.data() + delimiter, ELEM_HIDE_EMULATION_DELIMITER,
+                   ELEM_HIDE_EMULATION_DELIMITER_LEN * sizeof(String::value_type));
+    }
+
+    return converted;
+  }
+
+  return DependentString(text);
+}
+
+namespace
+{
   static constexpr String::value_type OPENING_CURLY_REPLACEMENT[] = u"\\7B ";
   static constexpr String::value_type CLOSING_CURLY_REPLACEMENT[] = u"\\7D ";
-  static constexpr String::size_type CURLY_REPLACEMENT_SIZE = sizeof(OPENING_CURLY_REPLACEMENT) / sizeof(OPENING_CURLY_REPLACEMENT[0]) - 1;
+  static constexpr String::size_type CURLY_REPLACEMENT_SIZE = LENGTH_OF(OPENING_CURLY_REPLACEMENT);
 
   OwnedString EscapeCurlies(String::size_type replacementCount,
                             const DependentString& str)
   {
     OwnedString result(str.length() + replacementCount * (CURLY_REPLACEMENT_SIZE - 1));
 
     String::value_type* current = result.data();
     for (String::size_type i = 0; i < str.length(); i++)
     {
       switch(str[i])
@@ skipping 43 matching lines @@
       if (item.second && !item.first.empty())
       {
         if (!result.empty())
           result.append(u',');
         result.append(item.first);
       }
     }
   }
   return result;
 }