Issue 3584 - Work around WebKit uppercase ID matching bug

abp2blocklist.js

Index: abp2blocklist.js
diff --git a/abp2blocklist.js b/abp2blocklist.js
index 71d227f3e66f9a9f9c044409840967523982d923..01f21bf132717872dc7c71ea7485ce711e694641 100644
--- a/abp2blocklist.js
+++ b/abp2blocklist.js
@@ -254,6 +254,62 @@ function hasNonASCI(obj)
   return false;
 }
 
+function attributeSyntaxForIDs(selector)

[Sebastian Noack, 2016/02/17 14:40:59]

Perhaps a better name for this function: convertIDSelectorsToAttributeSelectors

[kzar, 2016/02/17 15:13:52]

Done.

+{
+  if (selector.indexOf("#") == -1)

[Sebastian Noack, 2016/02/17 14:22:19]

Does this optimization has any measurable impact on the overall performance?

[kzar, 2016/02/17 15:13:52]

No, removed.

+    return selector;
+
+  // First we figure out where all the IDs are
+  let sep = "";
+  let IDstart = null;

[Sebastian Noack, 2016/02/17 14:40:59]

Nit: Since we only have one kind of start value and positions, omit the ID
prefix? It looks kinda ugly as it breaks the camel case notation.

[kzar, 2016/02/17 15:13:52]

Done.

+  let IDpositions = [];
+  for (let i = 0; i < selector.length; i++)
+  {
+    let chr = selector[i];
+
+    if (chr == "\\")        // ignore escaped characters
+      i++;
+    else if (chr == sep)    // don't split within quoted text
+      sep = "";             // e.g. [attr=","]
+    else if (sep == "")
+    {
+      if (chr == '"' || chr == "'")
+        sep = chr;
+      else if (IDstart == null)  // look for the start of an ID
+      {
+        if (chr == "#")
+          IDstart = i;
+      }
+      else if (chr < "0" && chr != "-" ||
+               chr > "9" && chr < "A" ||
+               chr > "Z" && chr != "_" && chr < "a" ||
+               chr > "z" && chr < "\x80") // look for the end of the ID
+      {
+        IDpositions.push({start: IDstart, end: i});

[Sebastian Noack, 2016/02/17 14:22:19]

I wonder we should do the replace inline:

selector = selector.substring(0, IDstart) +
           "[id=" + selector.substring(IDstart + 1, i) + "]" +
           selector.substring(i);
i += 4;

This logic would have to move to a function, to avoid duplication. Also we might
want to calculate the offset (4) rather than hardcoding it. But I could imagine
that it will still be simpler than the dance we do below, plus we would need
less state.

[Sebastian Noack, 2016/02/17 14:41:00]

Here is an example how that could look like:

  function replaceIDSubSelector(selector, start , end)
  {
    return selector.substring(0, start) +
           "[id=" + selector.substring(start + 1, end) + "]" +
           selector.substring(end);
  }

  ...

  let newSelector = replaceIDSubSelector(selector, IDStart, i);
  i += newSelector.length - selector.length;
  selector = newSelector;

[kzar, 2016/02/17 15:13:52]

Mind if I leave this as is? It works fine and I find it easier to understand how
it is.

+        IDstart = null;
+      }
+    }
+  }
+  if (IDstart != null)
+    IDpositions.push({start: IDstart, end: selector.length});
+
+  if (IDpositions.length == 0)

[Sebastian Noack, 2016/02/17 14:22:20]

Does this optimization has any measurable impact on the overall performance?

[kzar, 2016/02/17 15:13:52]

No, removed.

+    return selector;
+
+  // Now replace them all with the [id="someID"] form
+  let newSelector = "";
+  let position = 0;
+  for (let ID of IDpositions)
+  {
+    newSelector += selector.substring(position, ID.start);
+    newSelector += '[id="' + selector.substring(ID.start + 1, ID.end) + '"]';

[Sebastian Noack, 2016/02/17 14:22:19]

The quotes are actually redundant, as we copy an unquoted token here, which
cannot possibly have any spaces.

[kzar, 2016/02/17 15:13:52]

Acknowledged.

+    position = ID.end;
+  }
+  newSelector += selector.substring(position);
+
+  return newSelector;
+}
+
 function logRules()
 {
   let rules = [];
@@ -286,10 +342,16 @@ function logRules()
   {
     while (selectors.length)
     {
+      let selector = selectors.splice(0, selectorLimit).join(", ");
+
+      // As of Safari 9.0 element IDs are matched as lowercase. We work around
+      // this by converting to the attribute format [id="elementID"]
+      selector = attributeSyntaxForIDs(selector);
+
       addRule({
         trigger: {"url-filter": matchDomain},
         action: {type: "css-display-none",
-                 selector: selectors.splice(0, selectorLimit).join(", ")}
+                 selector: selector}
       });
     }
   });