gecko-dev/browser/actors/SearchSERPTelemetryChild.sy...

1614 строки
51 KiB
JavaScript

/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
import { XPCOMUtils } from "resource://gre/modules/XPCOMUtils.sys.mjs";
const lazy = {};
ChromeUtils.defineESModuleGetters(lazy, {
clearTimeout: "resource://gre/modules/Timer.sys.mjs",
setTimeout: "resource://gre/modules/Timer.sys.mjs",
});
XPCOMUtils.defineLazyPreferenceGetter(
lazy,
"serpEventTelemetryCategorization",
"browser.search.serpEventTelemetryCategorization.enabled",
false
);
export const CATEGORIZATION_SETTINGS = {
MAX_DOMAINS_TO_CATEGORIZE: 10,
};
// Duplicated from SearchSERPTelemetry to avoid loading the module on content
// startup.
const SEARCH_TELEMETRY_SHARED = {
PROVIDER_INFO: "SearchTelemetry:ProviderInfo",
LOAD_TIMEOUT: "SearchTelemetry:LoadTimeout",
SPA_LOAD_TIMEOUT: "SearchTelemetry:SPALoadTimeout",
};
/**
* Standard events mapped to the telemetry action.
*/
const EVENT_TYPE_TO_ACTION = {
click: "clicked",
};
/**
* A map of object conditions mapped to the condition that should be run when
* an event is triggered. The condition name is referenced in Remote Settings
* under the optional `condition` string for an event listener.
*/
const CONDITIONS = {
keydownEnter: event => event.key == "Enter",
};
/**
* SearchProviders looks after keeping track of the search provider information
* received from the main process.
*
* It is separate to SearchTelemetryChild so that it is not constructed for each
* tab, but once per process.
*/
class SearchProviders {
constructor() {
this._searchProviderInfo = null;
Services.cpmm.sharedData.addEventListener("change", this);
}
/**
* Gets the search provider information for any provider with advert information.
* If there is nothing in the cache, it will obtain it from shared data.
*
* @returns {object} Returns the search provider information.
* @see SearchTelemetry.sys.mjs
*/
get info() {
if (this._searchProviderInfo) {
return this._searchProviderInfo;
}
this._searchProviderInfo = Services.cpmm.sharedData.get(
SEARCH_TELEMETRY_SHARED.PROVIDER_INFO
);
if (!this._searchProviderInfo) {
return null;
}
this._searchProviderInfo = this._searchProviderInfo
// Filter-out non-ad providers so that we're not trying to match against
// those unnecessarily.
.filter(p => "extraAdServersRegexps" in p)
// Pre-build the regular expressions.
.map(p => {
p.adServerAttributes = p.adServerAttributes ?? [];
if (p.shoppingTab?.inspectRegexpInSERP) {
p.shoppingTab.regexp = new RegExp(p.shoppingTab.regexp);
}
return {
...p,
searchPageRegexp: new RegExp(p.searchPageRegexp),
extraAdServersRegexps: p.extraAdServersRegexps.map(
r => new RegExp(r)
),
};
});
return this._searchProviderInfo;
}
/**
* Handles events received from sharedData notifications.
*
* @param {object} event The event details.
*/
handleEvent(event) {
switch (event.type) {
case "change": {
if (event.changedKeys.includes(SEARCH_TELEMETRY_SHARED.PROVIDER_INFO)) {
// Just null out the provider information for now, we'll fetch it next
// time we need it.
this._searchProviderInfo = null;
}
break;
}
}
}
}
/**
* @typedef {object} EventListenerParam
* @property {string} eventType
* The type of event the listener should listen for. If the event type is
* is non-standard, it should correspond to a definition in
* CUSTOM_EVENT_TYPE_TO_DATA that will re-map it to a standard type. TODO
* @property {string} target
* The type of component that was the source of the event.
* @property {string | null} action
* The action that should be reported in telemetry.
*/
/**
* Provides a way to add listeners to elements, as well as unload them.
*/
class ListenerHelper {
/**
* Adds each event listener in an array of event listeners to each element
* in an array of elements, and sets their unloading.
*
* @param {Array<Element>} elements
* DOM elements to add event listeners to.
* @param {Array<EventListenerParam>} eventListenerParams
* The type of event to add the listener to.
* @param {string} target
*/
static addListeners(elements, eventListenerParams, target) {
if (!elements?.length || !eventListenerParams?.length) {
return;
}
let document = elements[0].ownerGlobal.document;
let callback = documentToEventCallbackMap.get(document);
if (!callback) {
return;
}
// The map might have entries from previous callers, so we must ensure
// we don't discard existing event listener callbacks.
let removeListenerCallbacks = [];
if (documentToRemoveEventListenersMap.has(document)) {
removeListenerCallbacks = documentToRemoveEventListenersMap.get(document);
}
for (let params of eventListenerParams) {
let removeListeners = ListenerHelper.addListener(
elements,
params,
target,
callback
);
removeListenerCallbacks = removeListenerCallbacks.concat(removeListeners);
}
documentToRemoveEventListenersMap.set(document, removeListenerCallbacks);
}
/**
* Add an event listener to each element in an array of elements.
*
* @param {Array<Element>} elements
* DOM elements to add event listeners to.
* @param {EventListenerParam} eventListenerParam
* @param {string} target
* @param {Function} callback
* @returns {Array<function>} Array of remove event listener functions.
*/
static addListener(elements, eventListenerParam, target, callback) {
let { action, eventType, target: customTarget } = eventListenerParam;
if (customTarget) {
target = customTarget;
}
if (!action) {
action = EVENT_TYPE_TO_ACTION[eventType];
if (!action) {
return [];
}
}
// Some events might have specific conditions we want to check before
// registering an engagement event.
let eventCallback;
if (eventListenerParam.condition) {
if (CONDITIONS[eventListenerParam.condition]) {
let condition = CONDITIONS[eventListenerParam.condition];
eventCallback = async event => {
let start = Cu.now();
if (condition(event)) {
callback({ action, target });
}
ChromeUtils.addProfilerMarker(
"SearchSERPTelemetryChild._eventCallback",
start,
"Call cached function before callback."
);
};
} else {
// If a component included a condition, but it wasn't found it is
// due to the fact that it was added in a more recent Firefox version
// than what is provided via search-telemetry-v2. Since the version of
// Firefox the user is using doesn't include this condition,
// we shouldn't add the event.
return [];
}
} else {
eventCallback = () => {
callback({ action, target });
};
}
let removeListenerCallbacks = [];
for (let element of elements) {
element.addEventListener(eventType, eventCallback);
removeListenerCallbacks.push(() => {
element.removeEventListener(eventType, eventCallback);
});
}
return removeListenerCallbacks;
}
}
/**
* Scans SERPs for ad components.
*/
class SearchAdImpression {
/**
* A reference to ad component information that is used if an anchor
* element could not be categorized to a specific ad component.
*
* @type {object}
*/
#defaultComponent = null;
/**
* Maps DOM elements to AdData.
*
* @type {Map<Element, AdData>}
*
* @typedef AdData
* @type {object}
* @property {string} type
* The type of ad component.
* @property {number} adsLoaded
* The number of ads counted as loaded for the component.
* @property {boolean} countChildren
* Whether all the children were counted for the component.
*/
#elementToAdDataMap = new Map();
/**
* An array of components to do a top-down search.
*/
#topDownComponents = [];
/**
* A reference the providerInfo for this SERP.
*
* @type {object}
*/
#providerInfo = null;
set providerInfo(providerInfo) {
if (this.#providerInfo?.telemetryId == providerInfo.telemetryId) {
return;
}
this.#providerInfo = providerInfo;
// Reset values.
this.#topDownComponents = [];
for (let component of this.#providerInfo.components) {
if (component.default) {
this.#defaultComponent = component;
continue;
}
if (component.topDown) {
this.#topDownComponents.push(component);
}
}
}
/**
* Check if the page has a shopping tab.
*
* @param {Document} document
* @return {boolean}
* Whether the page has a shopping tab. Defaults to false.
*/
hasShoppingTab(document) {
if (!this.#providerInfo?.shoppingTab) {
return false;
}
// If a provider has the inspectRegexpInSERP, we assume there must be an
// associated regexp that must be used on any hrefs matched by the elements
// found using the selector. If inspectRegexpInSERP is false, then check if
// the number of items found using the selector matches exactly one element
// to ensure we've used a fine-grained search.
let elements = document.querySelectorAll(
this.#providerInfo.shoppingTab.selector
);
if (this.#providerInfo.shoppingTab.inspectRegexpInSERP) {
let regexp = this.#providerInfo.shoppingTab.regexp;
for (let element of elements) {
let href = element.getAttribute("href");
if (href && regexp.test(href)) {
this.#recordElementData(element, {
type: "shopping_tab",
count: 1,
});
return true;
}
}
} else if (elements.length == 1) {
this.#recordElementData(elements[0], {
type: "shopping_tab",
count: 1,
});
return true;
}
return false;
}
/**
* Examine the list of anchors and the document object and find components
* on the page.
*
* With the list of anchors, go through each and find the component it
* belongs to and save it in elementToAdDataMap.
*
* Then, with the document object find components and save the results to
* elementToAdDataMap.
*
* Lastly, combine the results together in a new Map that contains the number
* of loaded, visible, and blocked results for the component.
*
* @param {HTMLCollectionOf<HTMLAnchorElement>} anchors
* @param {Document} document
*
* @returns {Map<string, object>}
* A map where the key is a string containing the type of ad component
* and the value is an object containing the number of adsLoaded,
* adsVisible, and adsHidden within the component.
*/
categorize(anchors, document) {
// Used for various functions to make relative URLs absolute.
let origin = new URL(document.documentURI).origin;
// Bottom up approach.
this.#categorizeAnchors(anchors, origin);
// Top down approach.
this.#categorizeDocument(document);
let componentToVisibilityMap = new Map();
let hrefToComponentMap = new Map();
let innerWindowHeight = document.ownerGlobal.innerHeight;
let scrollY = document.ownerGlobal.scrollY;
// Iterate over the results:
// - If it's searchbox add event listeners.
// - If it is a non_ads_link, map its href to component type.
// - For others, map its component type and check visibility.
for (let [element, data] of this.#elementToAdDataMap.entries()) {
if (data.type == "incontent_searchbox") {
// Bug 1880413: Deprecate hard coding the incontent search box.
// If searchbox has child elements, observe those, otherwise
// fallback to its parent element.
let searchElements = data.childElements.length
? data.childElements
: [element];
ListenerHelper.addListeners(
searchElements,
[
{ eventType: "click", target: data.type },
{
eventType: "keydown",
target: data.type,
action: "submitted",
condition: "keydownEnter",
},
],
data.type
);
continue;
}
if (data.childElements.length) {
for (let child of data.childElements) {
let href = this.#extractHref(child, origin);
if (href) {
hrefToComponentMap.set(href, data.type);
}
}
} else {
let href = this.#extractHref(element, origin);
if (href) {
hrefToComponentMap.set(href, data.type);
}
}
// If the component is a non_ads_link, skip visibility checks.
if (data.type == "non_ads_link") {
continue;
}
// If proxy children were found, check the visibility of all of them
// otherwise just check the visiblity of the first child.
let childElements;
if (data.proxyChildElements.length) {
childElements = data.proxyChildElements;
} else if (data.childElements.length) {
childElements = [data.childElements[0]];
}
let count = this.#countVisibleAndHiddenAds(
element,
data.adsLoaded,
childElements,
innerWindowHeight,
scrollY
);
if (componentToVisibilityMap.has(data.type)) {
let componentInfo = componentToVisibilityMap.get(data.type);
componentInfo.adsLoaded += data.adsLoaded;
componentInfo.adsVisible += count.adsVisible;
componentInfo.adsHidden += count.adsHidden;
} else {
componentToVisibilityMap.set(data.type, {
adsLoaded: data.adsLoaded,
adsVisible: count.adsVisible,
adsHidden: count.adsHidden,
});
}
}
// Release the DOM elements from the Map.
this.#elementToAdDataMap.clear();
return { componentToVisibilityMap, hrefToComponentMap };
}
/**
* Given an element, find the href that is most likely to make the request if
* the element is clicked. If the element contains a specific data attribute
* known to contain the url used to make the initial request, use it,
* otherwise use its href. Specific character conversions are done to mimic
* conversions likely to take place when urls are observed in network
* activity.
*
* @param {Element} element
* The element to inspect.
* @param {string} origin
* The origin for relative urls.
* @returns {string}
* The href of the element.
*/
#extractHref(element, origin) {
let href;
// Prioritize the href from a known data attribute value instead of
// its href property, as the former is the initial url the page will
// navigate to before being re-directed to the href.
for (let name of this.#providerInfo.adServerAttributes) {
if (
element.dataset[name] &&
this.#providerInfo.extraAdServersRegexps.some(regexp =>
regexp.test(element.dataset[name])
)
) {
href = element.dataset[name];
break;
}
}
// If a data attribute value was not found, fallback to the href.
href = href ?? element.getAttribute("href");
if (!href) {
return "";
}
// Avoid extracting or fixing up Javascript URLs.
if (href.startsWith("javascript")) {
return "";
}
// Hrefs can be relative.
if (!href.startsWith("https://") && !href.startsWith("http://")) {
href = origin + href;
}
// Per Bug 376844, apostrophes in query params are escaped, and thus, are
// percent-encoded by the time they are observed in the network. Even
// though it's more comprehensive, we avoid using newURI because its more
// expensive and conversions should be the exception.
// e.g. /path'?q=Mozilla's -> /path'?q=Mozilla%27s
let arr = href.split("?");
if (arr.length == 2 && arr[1].includes("'")) {
href = arr[0] + "?" + arr[1].replaceAll("'", "%27");
}
return href;
}
/**
* Given a list of anchor elements, group them into ad components.
*
* The first step in the process is to check if the anchor should be
* inspected. This is based on whether it contains an href or a
* data-attribute values that matches an ad link, or if it contains a
* pattern caught by a components included regular expression.
*
* Determine which component it belongs to and the number of matches for
* the component. The heuristic is described in findDataForAnchor.
* If there was a result and we haven't seen it before, save it in
* elementToAdDataMap.
*
* @param {HTMLCollectionOf<HTMLAnchorElement>} anchors
* The list of anchors to inspect.
* @param {string} origin
* The origin of the document the anchors belong to.
*/
#categorizeAnchors(anchors, origin) {
for (let anchor of anchors) {
if (this.#shouldInspectAnchor(anchor, origin)) {
let result = this.#findDataForAnchor(anchor);
if (result) {
this.#recordElementData(result.element, {
type: result.type,
count: result.count,
proxyChildElements: result.proxyChildElements,
childElements: result.childElements,
});
}
if (result.relatedElements?.length) {
// Bug 1880413: Deprecate related elements.
// Bottom-up approach with related elements are only used for
// non-link elements related to ads, like carousel arrows.
ListenerHelper.addListeners(
result.relatedElements,
[
{
action: "expanded",
eventType: "click",
},
],
result.type
);
}
}
}
}
/**
* Find components from the document object. This is mostly relevant for
* components that are non-ads and don't have an obvious regular expression
* that could match the pattern of the href.
*
* @param {Document} document
*/
#categorizeDocument(document) {
// using the subset of components that are top down,
// go through each one.
for (let component of this.#topDownComponents) {
// Top-down searches must have the topDown attribute.
if (!component.topDown) {
continue;
}
// Top down searches must include a parent.
if (!component.included?.parent) {
continue;
}
let parents = document.querySelectorAll(
component.included.parent.selector
);
if (parents.length) {
let eventListeners = component.included.parent.eventListeners;
if (eventListeners?.length) {
ListenerHelper.addListeners(parents, eventListeners, component.type);
}
for (let parent of parents) {
// Bug 1880413: Deprecate related elements.
// Top-down related elements are either used for auto-suggested
// elements of a searchbox, or elements on a page which we can't
// find through a bottom up approach but we want an add a listener,
// like carousels with arrows.
if (component.included.related?.selector) {
let relatedElements = parent.querySelectorAll(
component.included.related.selector
);
if (relatedElements.length) {
// For the search box, related elements with event listeners are
// auto-suggested terms. For everything else (e.g. carousels)
// they are expanded.
ListenerHelper.addListeners(
relatedElements,
[
{
action:
component.type == "incontent_searchbox"
? "submitted"
: "expanded",
eventType: "click",
},
],
component.type
);
}
}
if (component.included.children) {
for (let child of component.included.children) {
let childElements = parent.querySelectorAll(child.selector);
if (childElements.length) {
if (child.eventListeners) {
childElements = Array.from(childElements);
ListenerHelper.addListeners(
childElements,
child.eventListeners,
child.type ?? component.type
);
}
if (!child.skipCount) {
this.#recordElementData(parent, {
type: component.type,
childElements: Array.from(childElements),
});
}
}
}
} else if (!component.included.parent.skipCount) {
this.#recordElementData(parent, {
type: component.type,
});
}
}
}
}
}
/**
* Evaluates whether an anchor should be inspected based on matching
* regular expressions on either its href or specified data-attribute values.
*
* @param {HTMLAnchorElement} anchor
* @param {string} origin
* @returns {boolean}
*/
#shouldInspectAnchor(anchor, origin) {
let href = anchor.getAttribute("href");
if (!href) {
return false;
}
// Some hrefs might be relative.
if (!href.startsWith("https://") && !href.startsWith("http://")) {
href = origin + href;
}
let regexps = this.#providerInfo.extraAdServersRegexps;
// Anchors can contain ad links in a data-attribute.
for (let name of this.#providerInfo.adServerAttributes) {
let attributeValue = anchor.dataset[name];
if (
attributeValue &&
regexps.some(regexp => regexp.test(attributeValue))
) {
return true;
}
}
// Anchors can contain ad links in a specific href.
if (regexps.some(regexp => regexp.test(href))) {
return true;
}
return false;
}
/**
* Find the component data for an anchor.
*
* To categorize the anchor, we iterate over the list of possible components
* the anchor could be categorized. If the component is default, we skip
* checking because the fallback option for all anchor links is the default.
*
* First, get the "parent" of the anchor which best represents the DOM element
* that contains the anchor links for the component and no other component.
* This parent will be cached so that other anchors that share the same
* parent can be counted together.
*
* The check for a parent is a loop because we can define more than one best
* parent since on certain SERPs, it's possible for a "better" DOM element
* parent to appear occassionally.
*
* If no parent is found, skip this component.
*
* If a parent was found, check for specific child elements.
*
* Finding child DOM elements of a parent is optional. One reason to do so is
* to use child elements instead of anchor links to count the number of ads for
* a component via the `countChildren` property. This is provided because some ads
* (i.e. carousels) have multiple ad links in a single child element that go to the
* same location. In this scenario, all instances of the child are recorded as ads.
* Subsequent anchor elements that map to the same parent are ignored.
*
* Whether or not a child was found, return the information that was found,
* including whether or not all child elements were counted instead of anchors.
*
* If another anchor belonging to a parent that was previously recorded is the input
* for this function, we either increment the ad count by 1 or don't increment the ad
* count because the parent used `countChildren` completed the calculation in a
* previous step.
*
*
* @param {HTMLAnchorElement} anchor
* The anchor to be inspected.
* @returns {object}
* An object containing the element representing the root DOM element for
* the component, the type of component, how many ads were counted,
* and whether or not the count was of all the children.
*/
#findDataForAnchor(anchor) {
for (let component of this.#providerInfo.components) {
// First, check various conditions for skipping a component.
// A component should always have at least one included statement.
if (!component.included) {
continue;
}
// Top down searches are done after the bottom up search.
if (component.topDown) {
continue;
}
// The default component doesn't need to be checked,
// as it will be the fallback option.
if (component.default) {
continue;
}
// The anchor shouldn't belong to an excluded parent component if one
// is provided.
if (
component.excluded?.parent?.selector &&
anchor.closest(component.excluded.parent.selector)
) {
continue;
}
// All components with included should have a parent entry.
if (!component.included.parent) {
continue;
}
// Find the parent of the anchor.
let parent = anchor.closest(component.included.parent.selector);
if (!parent) {
continue;
}
// If we've already inspected the parent, add the child element to the
// list of anchors. Don't increment the ads loaded count, as we only care
// about grouping the anchor with the correct parent.
if (this.#elementToAdDataMap.has(parent)) {
return {
element: parent,
childElements: [anchor],
};
}
let relatedElements = [];
if (component.included.related?.selector) {
relatedElements = parent.querySelectorAll(
component.included.related.selector
);
}
// If the component has no defined children, return the parent element.
if (component.included.children) {
// Look for the first instance of a matching child selector.
for (let child of component.included.children) {
// If counting by child, get all of them at once.
if (child.countChildren) {
let proxyChildElements = parent.querySelectorAll(child.selector);
if (proxyChildElements.length) {
return {
element: parent,
type: child.type ?? component.type,
proxyChildElements: Array.from(proxyChildElements),
count: proxyChildElements.length,
childElements: [anchor],
relatedElements,
};
}
} else if (parent.querySelector(child.selector)) {
return {
element: parent,
type: child.type ?? component.type,
childElements: [anchor],
relatedElements,
};
}
}
}
// If no children were defined for this component, or none were found
// in the DOM, use the default definition.
return {
element: parent,
type: component.type,
childElements: [anchor],
relatedElements,
};
}
// If no component was found, use default values.
return {
element: anchor,
type: this.#defaultComponent.type,
};
}
/**
* Determines whether or not an ad was visible or hidden.
*
* An ad is considered visible if the parent element containing the
* component has non-zero dimensions, and all child element in the
* component have non-zero dimensions and fits within the window
* at the time when the impression was takent.
*
* For some components, like text ads, we don't send every child
* element for visibility, just the first text ad. For other components
* like carousels, we send all child elements because we do care about
* counting how many elements of the carousel were visible.
*
* @param {Element} element
* Element to be inspected
* @param {number} adsLoaded
* Number of ads initially determined to be loaded for this element.
* @param {Array<Element>} childElements
* List of children belonging to element.
* @param {number} innerWindowHeight
* Current height of the window containing the elements.
* @param {number} scrollY
* Current distance the window has been scrolled.
* @returns {object}
* Contains adsVisible which is the number of ads shown for the element
* and adsHidden, the number of ads not visible to the user.
*/
#countVisibleAndHiddenAds(
element,
adsLoaded,
childElements,
innerWindowHeight,
scrollY
) {
let elementRect =
element.ownerGlobal.windowUtils.getBoundsWithoutFlushing(element);
// If the element lacks a dimension, assume all ads that
// were contained within it are hidden.
if (elementRect.width == 0 || elementRect.height == 0) {
return {
adsVisible: 0,
adsHidden: adsLoaded,
};
}
// If an ad is far above the possible visible area of a window, an
// adblocker might be doing it as a workaround for blocking the ad.
if (
elementRect.bottom < 0 &&
innerWindowHeight + scrollY + elementRect.bottom < 0
) {
return {
adsVisible: 0,
adsHidden: adsLoaded,
};
}
// Since the parent element has dimensions but no child elements we want
// to inspect, check the parent itself is within the viewable area.
if (!childElements || !childElements.length) {
if (innerWindowHeight < elementRect.y + elementRect.height) {
return {
adsVisible: 0,
adsHidden: 0,
};
}
return {
adsVisible: 1,
adsHidden: 0,
};
}
let adsVisible = 0;
let adsHidden = 0;
for (let child of childElements) {
let itemRect =
child.ownerGlobal.windowUtils.getBoundsWithoutFlushing(child);
// If the child element we're inspecting has no dimension, it is hidden.
if (itemRect.height == 0 || itemRect.width == 0) {
adsHidden += 1;
continue;
}
// If the child element is to the left of the containing element, or to
// the right of the containing element, skip it.
if (
itemRect.x < elementRect.x ||
itemRect.x + itemRect.width > elementRect.x + elementRect.width
) {
continue;
}
// If the child element is too far down, skip it.
if (innerWindowHeight < itemRect.y + itemRect.height) {
continue;
}
++adsVisible;
}
return {
adsVisible,
adsHidden,
};
}
/**
* Caches ad data for a DOM element. The key of the map is by Element rather
* than Component for fast lookup on whether an Element has been already been
* categorized as a component. Subsequent calls to this passing the same
* element will update the list of child elements.
*
* @param {Element} element
* The element considered to be the root for the component.
* @param {object} params
* Various parameters that can be recorded. Whether the input values exist
* or not depends on which component was found, which heuristic should be used
* to determine whether an ad was visible, and whether we've already seen this
* element.
* @param {string | null} params.type
* The type of component.
* @param {number} params.count
* The number of ads found for a component. The number represents either
* the number of elements that match an ad expression or the number of DOM
* elements containing an ad link.
* @param {Array<Element>} params.proxyChildElements
* An array of DOM elements that should be inspected for visibility instead
* of the actual child elements, possibly because they are grouped.
* @param {Array<Element>} params.childElements
* An array of DOM elements to inspect.
*/
#recordElementData(
element,
{ type, count = 1, proxyChildElements = [], childElements = [] } = {}
) {
if (this.#elementToAdDataMap.has(element)) {
let recordedValues = this.#elementToAdDataMap.get(element);
if (childElements.length) {
recordedValues.childElements =
recordedValues.childElements.concat(childElements);
}
} else {
this.#elementToAdDataMap.set(element, {
type,
adsLoaded: count,
proxyChildElements,
childElements,
});
}
}
}
/**
* An object indicating which elements to examine for domains to extract and
* which heuristic technique to use to extract that element's domain.
*
* @typedef {object} ExtractorInfo
* @property {string} selectors
* A string representing the CSS selector that targets the elements on the
* page that contain domains we want to extract.
* @property {string} method
* A string representing which domain extraction heuristic to use.
* One of: "href", "dataAttribute" or "textContent".
* @property {object | null} options
* Options related to the domain extraction heuristic used.
* @property {string | null} options.dataAttributeKey
* The key name of the data attribute to lookup.
* @property {string | null} options.queryParamKey
* The key name of the query param value to lookup.
* @property {boolean | null} options.queryParamValueIsHref
* Whether the query param value is expected to contain an href.
*/
/**
* DomainExtractor examines elements on a page to retrieve the domains.
*/
class DomainExtractor {
/**
* Extract domains from the page using an array of information pertaining to
* the SERP.
*
* @param {Document} document
* The document for the SERP we are extracting domains from.
* @param {Array<ExtractorInfo>} extractorInfos
* Information used to target the domains we need to extract.
* @param {string} providerName
* Name of the search provider.
* @return {Set<string>}
* A set of the domains extracted from the page.
*/
extractDomainsFromDocument(document, extractorInfos, providerName) {
let extractedDomains = new Set();
if (!extractorInfos?.length) {
return extractedDomains;
}
for (let extractorInfo of extractorInfos) {
if (!extractorInfo.selectors) {
continue;
}
let elements = document.querySelectorAll(extractorInfo.selectors);
if (!elements) {
continue;
}
switch (extractorInfo.method) {
case "href": {
// Origin is used in case a URL needs to be made absolute.
let origin = new URL(document.documentURI).origin;
this.#fromElementsConvertHrefsIntoDomains(
elements,
origin,
providerName,
extractedDomains,
extractorInfo.options?.queryParamKey,
extractorInfo.options?.queryParamValueIsHref
);
break;
}
case "dataAttribute": {
this.#fromElementsRetrieveDataAttributeValues(
elements,
providerName,
extractorInfo.options?.dataAttributeKey,
extractedDomains
);
break;
}
case "textContent": {
this.#fromElementsRetrieveTextContent(
elements,
extractedDomains,
providerName
);
break;
}
}
}
return extractedDomains;
}
/**
* Given a list of elements, extract domains using href attributes. If the
* URL in the href includes the specified query param, the domain will be
* that query param's value. Otherwise it will be the hostname of the href
* attribute's URL.
*
* @param {NodeList<Element>} elements
* A list of elements from the page whose href attributes we want to
* inspect.
* @param {string} origin
* Origin of the current page.
* @param {string} providerName
* The name of the search provider.
* @param {Set<string>} extractedDomains
* The result set of domains extracted from the page.
* @param {string | null} queryParam
* An optional query param to search for in an element's href attribute.
* @param {boolean | null} queryParamValueIsHref
* Whether the query param value is expected to contain an href.
*/
#fromElementsConvertHrefsIntoDomains(
elements,
origin,
providerName,
extractedDomains,
queryParam,
queryParamValueIsHref
) {
for (let element of elements) {
if (this.#exceedsThreshold(extractedDomains.size)) {
return;
}
let href = element.getAttribute("href");
let url;
try {
url = new URL(href, origin);
} catch (ex) {
continue;
}
// Ignore non-standard protocols.
if (url.protocol != "https:" && url.protocol != "http:") {
continue;
}
if (queryParam) {
let paramValue = url.searchParams.get(queryParam);
if (queryParamValueIsHref) {
try {
paramValue = new URL(paramValue).hostname;
} catch (e) {
continue;
}
paramValue = this.#processDomain(paramValue, providerName);
}
if (paramValue && !extractedDomains.has(paramValue)) {
extractedDomains.add(paramValue);
}
} else if (url.hostname) {
let processedHostname = this.#processDomain(url.hostname, providerName);
if (processedHostname && !extractedDomains.has(processedHostname)) {
extractedDomains.add(processedHostname);
}
}
}
}
/**
* Given a list of elements, examine each for the specified data attribute.
* If found, add that data attribute's value to the result set of extracted
* domains as is.
*
* @param {NodeList<Element>} elements
* A list of elements from the page whose data attributes we want to
* inspect.
* @param {string} providerName
* The name of the search provider.
* @param {string} attribute
* The name of a data attribute to search for within an element.
* @param {Set<string>} extractedDomains
* The result set of domains extracted from the page.
*/
#fromElementsRetrieveDataAttributeValues(
elements,
providerName,
attribute,
extractedDomains
) {
for (let element of elements) {
if (this.#exceedsThreshold(extractedDomains.size)) {
return;
}
let value = element.dataset[attribute];
value = this.#processDomain(value, providerName);
if (value && !extractedDomains.has(value)) {
extractedDomains.add(value);
}
}
}
/* Given a list of elements, examine the text content for each element, which
* may be 1) a URL from which we can extract a domain or 2) text we can fix
* up to create a best guess as to a URL. If either condition is met, we add
* the domain to the result set.
*
* @param {NodeList<Element>} elements
* A list of elements from the page whose text content we want to inspect.
* @param {Set<string>} extractedDomains
* The result set of domains extracted from the page.
* @param {string} providerName
* The name of the search provider.
*/
#fromElementsRetrieveTextContent(elements, extractedDomains, providerName) {
// Not an exhaustive regex, but it fits our purpose for this method.
const LOOSE_URL_REGEX =
/^(?:https?:\/\/)?(?:www\.)?(?:[\w\-]+\.)+(?:[\w\-]{2,})/i;
// Known but acceptable limitations to this function, where the return
// value won't be correctly fixed up:
// 1) A url is embedded within other text. Ex: "xkcd.com is cool."
// 2) The url contains legal but unusual characters. Ex: $ ! * '
function fixup(textContent) {
return textContent
.toLowerCase()
.replaceAll(" ", "")
.replace(/\.$/, "")
.concat(".com");
}
for (let element of elements) {
if (this.#exceedsThreshold(extractedDomains.size)) {
return;
}
let textContent = element.textContent;
if (!textContent) {
continue;
}
let domain;
if (LOOSE_URL_REGEX.test(textContent)) {
// Creating a new URL object will throw if the protocol is missing.
if (!/^https?:\/\//.test(textContent)) {
textContent = "https://" + textContent;
}
try {
domain = new URL(textContent).hostname;
} catch (e) {
domain = fixup(textContent);
}
} else {
domain = fixup(textContent);
}
let processedDomain = this.#processDomain(domain, providerName);
if (processedDomain && !extractedDomains.has(processedDomain)) {
extractedDomains.add(processedDomain);
}
}
}
/**
* Processes a raw domain extracted from the SERP into its final form before
* categorization.
*
* @param {string} domain
* The domain extracted from the page.
* @param {string} providerName
* The provider associated with the page.
* @returns {string}
* The domain without any subdomains.
*/
#processDomain(domain, providerName) {
if (
domain.startsWith(`${providerName}.`) ||
domain.includes(`.${providerName}.`)
) {
return "";
}
return this.#stripDomainOfSubdomains(domain);
}
/**
* Helper to strip domains of any subdomains.
*
* @param {string} domain
* The domain to strip of any subdomains.
* @returns {object} browser
* The given domain with any subdomains removed.
*/
#stripDomainOfSubdomains(domain) {
let tld;
// Can throw an exception if the input has too few domain levels.
try {
tld = Services.eTLD.getKnownPublicSuffixFromHost(domain);
} catch (ex) {
return "";
}
let domainWithoutTLD = domain.substring(0, domain.length - tld.length);
let secondLevelDomain = domainWithoutTLD.split(".").at(-2);
return secondLevelDomain ? `${secondLevelDomain}.${tld}` : "";
}
/**
* Per a request from Data Science, we need to limit the number of domains
* categorized to 10 non-ad domains and 10 ad domains.
*
* @param {number} nDomains The number of domains processed.
* @returns {boolean} Whether or not the threshold was exceeded.
*/
#exceedsThreshold(nDomains) {
return nDomains >= CATEGORIZATION_SETTINGS.MAX_DOMAINS_TO_CATEGORIZE;
}
}
export const domainExtractor = new DomainExtractor();
const searchProviders = new SearchProviders();
const searchAdImpression = new SearchAdImpression();
const documentToEventCallbackMap = new WeakMap();
const documentToRemoveEventListenersMap = new WeakMap();
const documentToSubmitMap = new WeakMap();
/**
* SearchTelemetryChild monitors for pages that are partner searches, and
* looks through them to find links which looks like adverts and sends back
* a notification to SearchTelemetry for possible telemetry reporting.
*
* Only the partner details and the fact that at least one ad was found on the
* page are returned to SearchTelemetry. If no ads are found, no notification is
* given.
*/
export class SearchSERPTelemetryChild extends JSWindowActorChild {
/**
* Amount of time to wait after a page event before examining the page
* for ads.
*
* @type {number | null}
*/
#adTimeout;
/**
* Determines if there is a provider that matches the supplied URL and returns
* the information associated with that provider.
*
* @param {string} url The url to check
* @returns {array|null} Returns null if there's no match, otherwise an array
* of provider name and the provider information.
*/
_getProviderInfoForUrl(url) {
return searchProviders.info?.find(info => info.searchPageRegexp.test(url));
}
/**
* Checks to see if the page is a partner and has an ad link within it. If so,
* it will notify SearchTelemetry.
*/
_checkForAdLink(eventType) {
try {
if (!this.contentWindow) {
return;
}
} catch (ex) {
// unload occurred before the timer expired
return;
}
let doc = this.document;
let url = doc.documentURI;
let providerInfo = this._getProviderInfoForUrl(url);
if (!providerInfo) {
return;
}
let regexps = providerInfo.extraAdServersRegexps;
let anchors = doc.getElementsByTagName("a");
let hasAds = false;
for (let anchor of anchors) {
if (!anchor.href) {
continue;
}
for (let name of providerInfo.adServerAttributes) {
hasAds = regexps.some(regexp => regexp.test(anchor.dataset[name]));
if (hasAds) {
break;
}
}
if (!hasAds) {
hasAds = regexps.some(regexp => regexp.test(anchor.href));
}
if (hasAds) {
break;
}
}
if (hasAds) {
this.sendAsyncMessage("SearchTelemetry:PageInfo", {
hasAds,
url,
});
}
if (
providerInfo.components?.length &&
(eventType == "load" || eventType == "pageshow")
) {
// Start performance measurements.
let start = Cu.now();
let timerId = Glean.serp.categorizationDuration.start();
let pageActionCallback = info => {
if (info.action == "submitted") {
documentToSubmitMap.set(doc, true);
}
this.sendAsyncMessage("SearchTelemetry:Action", {
target: info.target,
url: info.url,
action: info.action,
});
};
documentToEventCallbackMap.set(this.document, pageActionCallback);
let componentToVisibilityMap, hrefToComponentMap;
try {
let result = searchAdImpression.categorize(anchors, doc);
componentToVisibilityMap = result.componentToVisibilityMap;
hrefToComponentMap = result.hrefToComponentMap;
} catch (e) {
// Cancel the timer if an error encountered.
Glean.serp.categorizationDuration.cancel(timerId);
}
if (componentToVisibilityMap && hrefToComponentMap) {
// End measurements.
ChromeUtils.addProfilerMarker(
"SearchSERPTelemetryChild._checkForAdLink",
start,
"Checked anchors for visibility"
);
Glean.serp.categorizationDuration.stopAndAccumulate(timerId);
this.sendAsyncMessage("SearchTelemetry:AdImpressions", {
adImpressions: componentToVisibilityMap,
hrefToComponentMap,
url,
});
}
}
if (
lazy.serpEventTelemetryCategorization &&
providerInfo.domainExtraction &&
(eventType == "load" || eventType == "pageshow")
) {
let start = Cu.now();
let nonAdDomains = domainExtractor.extractDomainsFromDocument(
doc,
providerInfo.domainExtraction.nonAds,
providerInfo.telemetryId
);
let adDomains = domainExtractor.extractDomainsFromDocument(
doc,
providerInfo.domainExtraction.ads,
providerInfo.telemetryId
);
this.sendAsyncMessage("SearchTelemetry:Domains", {
url,
nonAdDomains,
adDomains,
});
ChromeUtils.addProfilerMarker(
"SearchSERPTelemetryChild._checkForAdLink",
start,
"Extract domains from elements"
);
}
}
/**
* Checks for the presence of certain components on the page that are
* required for recording the page impression.
*/
#checkForPageImpressionComponents() {
let url = this.document.documentURI;
let providerInfo = this._getProviderInfoForUrl(url);
if (providerInfo.components?.length) {
searchAdImpression.providerInfo = providerInfo;
let start = Cu.now();
let shoppingTabDisplayed = searchAdImpression.hasShoppingTab(
this.document
);
ChromeUtils.addProfilerMarker(
"SearchSERPTelemetryChild.#recordImpression",
start,
"Checked for shopping tab"
);
this.sendAsyncMessage("SearchTelemetry:PageImpression", {
url,
shoppingTabDisplayed,
});
}
}
#removeEventListeners() {
let callbacks = documentToRemoveEventListenersMap.get(this.document);
if (callbacks) {
for (let callback of callbacks) {
callback();
}
documentToRemoveEventListenersMap.delete(this.document);
}
}
/**
* Handles events received from the actor child notifications.
*
* @param {object} event The event details.
*/
handleEvent(event) {
if (!this.#urlIsSERP(this.document.documentURI)) {
return;
}
switch (event.type) {
case "pageshow": {
// If a page is loaded from the bfcache, we won't get a "DOMContentLoaded"
// event, so we need to rely on "pageshow" in this case. Note: we do this
// so that we remain consistent with the *.in-content:sap* count for the
// SEARCH_COUNTS histogram.
if (event.persisted) {
this.#checkForPageImpressionComponents();
this.#check(event.type);
}
break;
}
case "DOMContentLoaded": {
this.#checkForPageImpressionComponents();
this.#check(event.type);
break;
}
case "load": {
// We check both DOMContentLoaded and load in case the page has
// taken a long time to load and the ad is only detected on load.
// We still check at DOMContentLoaded because if the page hasn't
// finished loading and the user navigates away, we still want to know
// if there were ads on the page or not at that time.
this.#check(event.type);
break;
}
case "pagehide": {
let callbacks = documentToRemoveEventListenersMap.get(this.document);
if (callbacks) {
for (let removeEventListenerCallback of callbacks) {
removeEventListenerCallback();
}
documentToRemoveEventListenersMap.delete(this.document);
}
this.#cancelCheck();
break;
}
}
}
async receiveMessage(message) {
switch (message.name) {
case "SearchSERPTelemetry:WaitForSPAPageLoad":
lazy.setTimeout(() => {
this.#checkForPageImpressionComponents();
this._checkForAdLink("load");
}, Services.cpmm.sharedData.get(SEARCH_TELEMETRY_SHARED.SPA_LOAD_TIMEOUT));
break;
case "SearchSERPTelemetry:StopTrackingDocument":
this.#removeDocumentFromSubmitMap();
this.#removeEventListeners();
break;
case "SearchSERPTelemetry:DidSubmit":
return this.#didSubmit();
}
return null;
}
#didSubmit() {
return documentToSubmitMap.get(this.document);
}
#removeDocumentFromSubmitMap() {
documentToSubmitMap.delete(this.document);
}
#urlIsSERP(url) {
let provider = this._getProviderInfoForUrl(this.document.documentURI);
if (provider) {
// Some URLs can match provider info but also be the provider's homepage
// instead of a SERP.
// e.g. https://example.com/ vs. https://example.com/?foo=bar
// To check this, we look for the presence of the query parameter
// that contains a search term.
let queries = new URLSearchParams(url.split("#")[0].split("?")[1]);
for (let queryParamName of provider.queryParamNames) {
if (queries.get(queryParamName)) {
return true;
}
}
}
return false;
}
#cancelCheck() {
if (this._waitForContentTimeout) {
lazy.clearTimeout(this._waitForContentTimeout);
}
}
#check(eventType) {
if (!this.#adTimeout) {
this.#adTimeout = Services.cpmm.sharedData.get(
SEARCH_TELEMETRY_SHARED.LOAD_TIMEOUT
);
}
this.#cancelCheck();
this._waitForContentTimeout = lazy.setTimeout(() => {
this._checkForAdLink(eventType);
}, this.#adTimeout);
}
}