Issue 7450 - Implement hide-if-contains-visible-text snippet

lib/content/snippets.js

 /*
  * 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
@@ skipping 390 matching lines @@
     }
   });
 }
 
 exports["hide-if-shadow-contains"] = makeInjector(hideIfShadowContains,
                                                   toRegExp, regexEscape,
                                                   hideElement);
 
 /**
  * Hides any HTML element or one of its ancestors matching a CSS selector if
+ * it matches the provided condition.
+ *
+ * @param {function} match The function that provides the matching condition.
+ * @param {string} selector The CSS selector that an HTML element must match
+ *   for it to be hidden.
+ * @param {?string} [searchSelector] The CSS selector that an HTML element
+ *   containing the given string must match. Defaults to the value of the
+ *   <code>selector</code> argument.
+ */
+function hideIfMatches(match, selector, searchSelector)
+{
+  if (searchSelector == null)
+    searchSelector = selector;
+
+  let callback = () =>
+  {
+    for (let element of document.querySelectorAll(searchSelector))
+    {
+      let closest = element.closest(selector);
+      if (closest && match(element, closest))
+        hideElement(closest);
+    }
+  };
+  new MutationObserver(callback)
+    .observe(document, {childList: true, characterData: true, subtree: true});
+  callback();
+}
+
+/**
+ * Hides any HTML element or one of its ancestors matching a CSS selector if
  * the text content of the element contains a given string.
  *
  * @param {string} search The string to look for in HTML elements. If the
  *   string begins and ends with a slash (<code>/</code>), the text in between
  *   is treated as a regular expression.
  * @param {string} selector The CSS selector that an HTML element must match
  *   for it to be hidden.
- * @param {string?} [searchSelector] The CSS selector that an HTML element
+ * @param {?string} [searchSelector] The CSS selector that an HTML element
  *   containing the given string must match. Defaults to the value of the
  *   <code>selector</code> argument.
  */
 function hideIfContains(search, selector = "*", searchSelector = null)
 {
-  if (searchSelector == null)
-    searchSelector = selector;
+  let re = toRegExp(search);
+
+  hideIfMatches(element => re.test(element.textContent),
+                selector, searchSelector);
+}
+
+exports["hide-if-contains"] = hideIfContains;
+
+/**
+ * Hides any HTML element matching a CSS selector if the visible text content
+ * of the element contains a given string.
+ *
+ * @param {string} search The string to match to the visible text. Is considered
+ *   visible text that isn't hidden by CSS properties or other means.
+ *   If the string begins and ends with a slash (<code>/</code>), the
+ *   text in between is treated as a regular expression.
+ * @param {string} selector The CSS selector that an HTML element must match
+ *   for it to be hidden.
+ * @param {?string} [searchSelector] The CSS selector that an HTML element
+ *   containing the given string must match. Defaults to the value of the
+ *   <code>selector</code> argument.
+ */
+function hideIfContainsVisibleText(search, selector, searchSelector = null)
+{
+  /**
+   * Determines if the text inside the element is visible.
+   * @param {Element} element The element we are checking.
+   * @param {?CSSStyleDeclaration} style The computed style of element. If
+   *   falsey it will be queried.
+   * @returns {bool} Whether the text is visible.
+   */
+  function isTextVisible(element, style)
+  {
+    if (!style)
+      style = window.getComputedStyle(element);
+
+    if (style.getPropertyValue("opacity") == "0")
+      return false;
+    if (style.getPropertyValue("font-size") == "0px")
+      return false;
+
+    let color = style.getPropertyValue("color");
+    // if color is transparent...
+    if (color == "rgba(0, 0, 0, 0)")
+      return false;
+    if (style.getPropertyValue("background-color") == color)
+      return false;
+
+    return true;
+  }
+
+  /**
+   * Check if an element is visible
+   * @param {Element} element The element to check visibility of/
+   * @param {?CSSStyleDeclaration} style The computed style of element. If
+   *   falsey it will be queried.
+   * @param {?Element} closest The closest parent to reach.
+   * @return {bool} Whether the element is visible.
+   */
+  function isVisible(element, style, closest)
+  {
+    if (!style)
+      style = window.getComputedStyle(element);
+
+    if (style.getPropertyValue("display") == "none")
+      return false;
+    let visibility = style.getPropertyValue("visibility");
+    if (visibility == "hidden" || visibility == "collapse")
+      return false;
+
+    if (!closest || element == closest)
+      return true;
+
+    return isVisible(element.parentNode, null, closest);

[a.giammarchi, 2019/05/08 09:43:47]

I have bad feelings about this for at least two reasons:

 1. it breaks if it reaches the `document`, since no `parentElement` would be
there
 2. it's very expensive in complex DOM

I have a coupe of suggestions here:

 1. how about we cut the mustard and use element.getBoundingClientRect()
directly, so that if either width or height are 0 we can call it a day without
crawling to the root each node that is visible?
 2. if gBCR() is not a solution, mind checking that `parentNode` is a `nodeType`
=== `Node.ELEMENT_NODE` so that stuff in fragments, or document itself, wouldn't
break?

[hub, 2019/05/08 17:03:42]

I'll stop if parent is null.
How come? it is already limited to closest. I realise I should have used
`parentElement` instead.
To "closest"
Using `parentElement` now and stop when it returns null

[a.giammarchi, 2019/05/09 08:58:59]

where does this happen? 'cause all I see is that you grab the computed style
right away from an element without checking that element is live/not null

I guess logically, all I am saying is that if !closest || element == closest ||
!element.parentElement you should return and avoid recursion.

Please note I am talking about the signature as is, where one could pass closest
as plain `document` simply to know if a generic element is visible.

In that case:

  * closest is document
  * element is anything
  * the moment element crawl up to documentElement, closest is there but no
parentElement is returned so that the following breaks:

window.getComputedStyle(document.documentElement.parentElement)

[hub, 2019/05/09 12:06:26]

It is actually better if I actually properly submit the changes for review.
Sorry about that.

+  }
+
+  /**
+   * Returns the visible text content from an element and its descendants.
+   * @param {Element} element The element whose visible text we want.
+   * @param {Element} closest The closest parent to reach while checking
+   *   for visibility.
+   * @returns {string} The text that is visible.
+   */
+  function getVisibleContent(element, closest)
+  {
+    let style = window.getComputedStyle(element);
+    if (!isVisible(element, style, closest))
+      return "";
+
+    let text = "";
+    for (let node of element.childNodes)
+    {
+      switch (node.nodeType)
+      {
+        case Node.ELEMENT_NODE:
+          text += getVisibleContent(node, element);
+          break;
+        case Node.TEXT_NODE:
+          if (isTextVisible(element, style))
+            text += node.nodeValue;
+          break;
+      }
+    }
+    return text;
+  }
 
   let re = toRegExp(search);
 
-  new MutationObserver(() =>
-  {
-    for (let element of document.querySelectorAll(searchSelector))
-    {
-      if (re.test(element.textContent))
-      {
-        let closest = element.closest(selector);
-        if (closest)
-          hideElement(closest);
-      }
-    }
-  })
-  .observe(document, {childList: true, characterData: true, subtree: true});
+  hideIfMatches(
+    (element, closest) => re.test(getVisibleContent(element, closest)),
+    selector, searchSelector);
 }
 
-exports["hide-if-contains"] = hideIfContains;
+exports["hide-if-contains-visible-text"] = hideIfContainsVisibleText;
 
 /**
  * Hides any HTML element or one of its ancestors matching a CSS selector if
  * the text content of the element contains a given string and, optionally, if
  * the element's computed style contains a given string.
  *
  * @param {string} search The string to look for in HTML elements. If the
  *   string begins and ends with a slash (<code>/</code>), the text in between
  *   is treated as a regular expression.
  * @param {string} selector The CSS selector that an HTML element must match
@@ skipping 438 matching lines @@
 
       args[0] = url.href;
     }
 
     return fetch_.apply(this, args);
   };
 }
 
 exports["strip-fetch-query-parameter"] = makeInjector(stripFetchQueryParameter,
                                                       toRegExp, regexEscape);