[experiment] Issue 7045 - Optimize V8 memory layout of filter text

lib/filterText.js

Index: lib/filterText.js
===================================================================
new file mode 100644
--- /dev/null
+++ b/lib/filterText.js
@@ -0,0 +1,158 @@
+/*
+ * 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/>.
+ */
+
+"use strict";
+
+/**
+ * @fileOverview Filter text operations.
+ */
+
+// The memory optimizations in this file work as of V8 7.1.314.
+
+/**
+ * The minimum length of a domain string for it to be considered long enough to
+ * be eligible for optimization.
+ * @type {number}
+ */
+const LONG_DOMAINS_THRESHOLD = 1000;
+
+/**
+ * Cached strings.
+ * @type {Map.<string,string>}
+ */
+let strings = new Map();
+
+/**
+ * Slices a string out of its internal parent string, thus allowing the memory
+ * occupied by the parent string to be freed up.
+ *
+ * JavaScript engines like V8 tend to hold on to the parent of a sliced string
+ * even if there are no references to the parent string. While this
+ * optimization is enormously beneficial in general, in certain cases it has
+ * the opposite effect of needlessly holding on to large amounts of unused
+ * memory (V8 issue #2869). This operation creates an entirely new internal
+ * string with its own copy of the original string's memory, thus allowing the
+ * original string and any of its parents to be freed up.
+ *
+ * Note: This is a relatively expensive operation and should be used only when
+ * the benefit outweighs the cost.
+ *
+ * @param {string} string The string to slice.
+ *
+ * @returns {string} An entirely new copy of the original string.
+ */
+function trueSlice(string)
+{
+  return JSON.parse(JSON.stringify(string));
+}
+
+/**
+ * Optimizes content filter text.
+ *
+ * @param {string} text The filter text.
+ * @param {?string} [domains] The domains part of the filter text.
+ * @param {?string} [type] The type part of the filter text, either
+ *   <code>null</code> or <code>undefined</code>, or one of <code>@</code>,
+ *   <code>?</code>, and <code>$</code>.
+ * @param {string} body The body part of the filter text.
+ *
+ * @returns {Array.<string>} An array containing the optimized filter text and
+ *   its optimized parts.
+ */
+function optimizeContentFilterText(text, domains, type, body)
+{
+  // In EasyList+AA there are a handful of filters with very long domain
+  // strings of 1,000 characters and even up to 100,000 characters. These tend
+  // to take up a lot of memory. We can restructure the text here to optimize
+  // for better memory usage on V8.
+  if (domains && domains.length >= LONG_DOMAINS_THRESHOLD)
+  {
+    let copy = strings.get(domains);
+    if (copy)
+    {
+      // Point to the cached copy.
+      domains = copy;
+
+      // V8 tends to hold on to the parent of a sliced string even if there are
+      // no references to it. We must "slice" out these strings properly so the
+      // original filter text is freed up.
+      // https://bugs.chromium.org/p/v8/issues/detail?id=2869
+      if (typeof type == "string")
+        type = trueSlice(type);
+      body = trueSlice(body);
+
+      // Reconstruct the text with an optimized layout.
+      text = domains + "#" + (type || "") + "#" + body;
+    }
+    else
+    {
+      strings.set(domains, domains);
+    }
+  }
+
+  return [text, domains, type, body];
+}
+
+exports.optimizeContentFilterText = optimizeContentFilterText;
+
+/**
+ * Optimizes blocking and whitelist filter text.
+ *
+ * @param {string} text The filter text.
+ * @param {string} pattern The pattern part of the filter text.
+ * @param {?string} [domains] The domains part of the filter text.
+ * @param {?string} [sitekeys] The sitekeys part of the filter text.
+ * @param {?string} [csp] The CSP part of the filter text.
+ * @param {?string} [rewrite] The rewrite pattern part of the filter text.
+ *
+ * @returns {Array.<string>} An array containing the optimized filter text and
+ *   its optimized parts.
+ */
+function optimizeRegExpFilterText(text, pattern, domains, sitekeys, csp,
+                                  rewrite)
+{
+  if (!sitekeys && !csp && !rewrite &&
+      domains && domains.length >= LONG_DOMAINS_THRESHOLD &&
+      text.endsWith(domains))
+  {
+    let copy = strings.get(domains);
+    if (copy)
+    {
+      domains = copy;
+
+      text = trueSlice(text.substring(0, text.length - domains.length));
+
+      if (text[0] == "@" && text[1] == "@")
+        pattern = text.substring(2, 2 + pattern.length);
+      else
+        pattern = text.substring(0, pattern.length);
+
+      // Note: This must be the last operation on the text in order for this
+      // optimization to work. Any further operations on the text may undo the
+      // optimization.
+      text += domains;
+    }
+    else
+    {
+      strings.set(domains, domains);
+    }
+  }
+
+  return [text, pattern, domains, sitekeys, csp, rewrite];
+}
+
+exports.optimizeRegExpFilterText = optimizeRegExpFilterText;