summaryrefslogtreecommitdiffstats
path: root/browser/components/urlbar/UrlbarMuxerUnifiedComplete.sys.mjs
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--browser/components/urlbar/UrlbarMuxerUnifiedComplete.sys.mjs1263
1 files changed, 1263 insertions, 0 deletions
diff --git a/browser/components/urlbar/UrlbarMuxerUnifiedComplete.sys.mjs b/browser/components/urlbar/UrlbarMuxerUnifiedComplete.sys.mjs
new file mode 100644
index 0000000000..e736f2c55c
--- /dev/null
+++ b/browser/components/urlbar/UrlbarMuxerUnifiedComplete.sys.mjs
@@ -0,0 +1,1263 @@
+/* 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/. */
+
+/**
+ * This module exports a component used to sort results in a UrlbarQueryContext.
+ */
+
+import { XPCOMUtils } from "resource://gre/modules/XPCOMUtils.sys.mjs";
+
+import {
+ UrlbarMuxer,
+ UrlbarUtils,
+} from "resource:///modules/UrlbarUtils.sys.mjs";
+
+const lazy = {};
+
+ChromeUtils.defineESModuleGetters(lazy, {
+ QuickSuggest: "resource:///modules/QuickSuggest.sys.mjs",
+ UrlbarPrefs: "resource:///modules/UrlbarPrefs.sys.mjs",
+ UrlbarProviderQuickSuggest:
+ "resource:///modules/UrlbarProviderQuickSuggest.sys.mjs",
+ UrlbarProviderTabToSearch:
+ "resource:///modules/UrlbarProviderTabToSearch.sys.mjs",
+ UrlbarSearchUtils: "resource:///modules/UrlbarSearchUtils.sys.mjs",
+});
+
+XPCOMUtils.defineLazyGetter(lazy, "logger", () =>
+ UrlbarUtils.getLogger({ prefix: "MuxerUnifiedComplete" })
+);
+
+/**
+ * Class used to create a muxer.
+ * The muxer receives and sorts results in a UrlbarQueryContext.
+ */
+class MuxerUnifiedComplete extends UrlbarMuxer {
+ constructor() {
+ super();
+ }
+
+ get name() {
+ return "UnifiedComplete";
+ }
+
+ /**
+ * Sorts results in the given UrlbarQueryContext.
+ *
+ * @param {UrlbarQueryContext} context
+ * The query context.
+ */
+ sort(context) {
+ // This method is called multiple times per keystroke, so it should be as
+ // fast and efficient as possible. We do two passes through the results:
+ // one to collect state for the second pass, and then a second to build the
+ // sorted list of results. If you find yourself writing something like
+ // context.results.find(), filter(), sort(), etc., modify one or both passes
+ // instead.
+
+ // Global state we'll use to make decisions during this sort.
+ let state = {
+ context,
+ // RESULT_GROUP => array of results belonging to the group, excluding
+ // group-relative suggestedIndex results
+ resultsByGroup: new Map(),
+ // RESULT_GROUP => array of group-relative suggestedIndex results
+ // belonging to the group
+ suggestedIndexResultsByGroup: new Map(),
+ // This is analogous to `maxResults` except it's the total available
+ // result span instead of the total available result count. We'll add
+ // results until `usedResultSpan` would exceed `availableResultSpan`.
+ availableResultSpan: context.maxResults,
+ // The total span of results that have been added so far.
+ usedResultSpan: 0,
+ strippedUrlToTopPrefixAndTitle: new Map(),
+ urlToTabResultType: new Map(),
+ addedRemoteTabUrls: new Set(),
+ addedSwitchTabUrls: new Set(),
+ canShowPrivateSearch: context.results.length > 1,
+ canShowTailSuggestions: true,
+ // Form history and remote suggestions added so far. Used for deduping
+ // suggestions. Also includes the heuristic query string if the heuristic
+ // is a search result. All strings in the set are lowercased.
+ suggestions: new Set(),
+ canAddTabToSearch: true,
+ hasUnitConversionResult: false,
+ maxHeuristicResultSpan: 0,
+ maxTabToSearchResultSpan: 0,
+ // When you add state, update _copyState() as necessary.
+ };
+
+ // Do the first pass over all results to build some state.
+ for (let result of context.results) {
+ // Add each result to the appropriate `resultsByGroup` map.
+ let group = UrlbarUtils.getResultGroup(result);
+ let resultsByGroup =
+ result.hasSuggestedIndex && result.isSuggestedIndexRelativeToGroup
+ ? state.suggestedIndexResultsByGroup
+ : state.resultsByGroup;
+ let results = resultsByGroup.get(group);
+ if (!results) {
+ results = [];
+ resultsByGroup.set(group, results);
+ }
+ results.push(result);
+
+ // Update pre-add state.
+ this._updateStatePreAdd(result, state);
+ }
+
+ // Now that the first pass is done, adjust the available result span.
+ if (state.maxTabToSearchResultSpan) {
+ // Subtract the max tab-to-search span.
+ state.availableResultSpan = Math.max(
+ state.availableResultSpan - state.maxTabToSearchResultSpan,
+ 0
+ );
+ }
+ if (state.maxHeuristicResultSpan) {
+ if (lazy.UrlbarPrefs.get("experimental.hideHeuristic")) {
+ // The heuristic is hidden. The muxer will include it but the view will
+ // hide it. Increase the available span to compensate so that the total
+ // visible span accurately reflects `context.maxResults`.
+ state.availableResultSpan += state.maxHeuristicResultSpan;
+ } else if (context.maxResults > 0) {
+ // `context.maxResults` is positive. Make sure there's room for the
+ // heuristic even if it means exceeding `context.maxResults`.
+ state.availableResultSpan = Math.max(
+ state.availableResultSpan,
+ state.maxHeuristicResultSpan
+ );
+ }
+ }
+
+ // Determine the result groups to use for this sort. In search mode with
+ // an engine, show search suggestions first.
+ let rootGroup = context.searchMode?.engineName
+ ? lazy.UrlbarPrefs.makeResultGroups({ showSearchSuggestionsFirst: true })
+ : lazy.UrlbarPrefs.get("resultGroups");
+ lazy.logger.debug(`Groups: ${rootGroup}`);
+
+ // Fill the root group.
+ let [sortedResults] = this._fillGroup(
+ rootGroup,
+ { availableSpan: state.availableResultSpan, maxResultCount: Infinity },
+ state
+ );
+
+ // Add global suggestedIndex results unless the max result count is zero,
+ // which isn't really supported but it's easy to honor here. We add them all
+ // even if they exceed the max because we assume they're high-priority
+ // results that should always be shown, and as long as the max is positive
+ // it's not a problem to exceed it sometimes. In practice that will happen
+ // only for small, non-default values of `maxRichResults`.
+ if (context.maxResults > 0) {
+ let suggestedIndexResults = state.resultsByGroup.get(
+ UrlbarUtils.RESULT_GROUP.SUGGESTED_INDEX
+ );
+ if (suggestedIndexResults) {
+ this._addSuggestedIndexResults(
+ suggestedIndexResults,
+ sortedResults,
+ state
+ );
+ }
+ }
+
+ context.results = sortedResults;
+ }
+
+ /**
+ * Returns a *deep* copy of state (except for `state.context`, which is
+ * shallow copied). i.e., any Maps, Sets, and arrays in the state should be
+ * recursively copied so that the original state is not modified when the copy
+ * is modified.
+ *
+ * @param {object} state
+ * The muxer state to copy.
+ * @returns {object}
+ * A deep copy of the state.
+ */
+ _copyState(state) {
+ let copy = Object.assign({}, state, {
+ resultsByGroup: new Map(),
+ suggestedIndexResultsByGroup: new Map(),
+ strippedUrlToTopPrefixAndTitle: new Map(
+ state.strippedUrlToTopPrefixAndTitle
+ ),
+ urlToTabResultType: new Map(state.urlToTabResultType),
+ addedRemoteTabUrls: new Set(state.addedRemoteTabUrls),
+ addedSwitchTabUrls: new Set(state.addedSwitchTabUrls),
+ suggestions: new Set(state.suggestions),
+ });
+
+ // Deep copy the `resultsByGroup` maps.
+ for (let key of ["resultsByGroup", "suggestedIndexResultsByGroup"]) {
+ for (let [group, results] of state[key]) {
+ copy[key].set(group, [...results]);
+ }
+ }
+
+ return copy;
+ }
+
+ /**
+ * Recursively fills a result group and its children.
+ *
+ * There are two ways to limit the number of results in a group:
+ *
+ * (1) By max total result span using the `availableSpan` property. The group
+ * will be filled so that the total span of its results does not exceed this
+ * value.
+ *
+ * (2) By max total result count using the `maxResultCount` property. The
+ * group will be filled so that the total number of its results does not
+ * exceed this value.
+ *
+ * Both `availableSpan` and `maxResultCount` may be defined, and the group's
+ * results will be capped to whichever limit is reached first. If either is
+ * not defined, then the group inherits that limit from its parent group.
+ *
+ * In addition to limiting their total number of results, groups can also
+ * control the composition of their child groups by using flex ratios. A group
+ * can define a `flexChildren: true` property, and in that case each of its
+ * children should have a `flex` property. Each child will be filled according
+ * to the ratio of its flex value and the sum of the flex values of all the
+ * children, similar to HTML flexbox. If some children do not fill up but
+ * others do, the filled-up children will be allowed to grow to use up the
+ * unfilled space.
+ *
+ * @param {object} group
+ * The result group to fill.
+ * @param {object} limits
+ * An object with optional `availableSpan` and `maxResultCount` properties
+ * as described above. They will be used as the limits for the group.
+ * @param {object} state
+ * The muxer state.
+ * @returns {Array}
+ * `[results, usedLimits, hasMoreResults]` -- see `_addResults`.
+ */
+ _fillGroup(group, limits, state) {
+ // Get the group's suggestedIndex results. Reminder: `group.group` is a
+ // `RESULT_GROUP` constant.
+ let suggestedIndexResults;
+ if ("group" in group) {
+ suggestedIndexResults = state.suggestedIndexResultsByGroup.get(
+ group.group
+ );
+ if (suggestedIndexResults) {
+ // Subtract them from the group's limits so there will be room for them
+ // later. Create a new `limits` object so we don't modify the caller's.
+ let span = suggestedIndexResults.reduce((sum, result) => {
+ sum += UrlbarUtils.getSpanForResult(result);
+ return sum;
+ }, 0);
+ limits = { ...limits };
+ limits.availableSpan = Math.max(limits.availableSpan - span, 0);
+ limits.maxResultCount = Math.max(
+ limits.maxResultCount - suggestedIndexResults.length,
+ 0
+ );
+ }
+ }
+
+ // Fill the group. If it has children, fill them recursively. Otherwise fill
+ // the group directly.
+ let [results, usedLimits, hasMoreResults] = group.children
+ ? this._fillGroupChildren(group, limits, state)
+ : this._addResults(group.group, limits, state);
+
+ // Add the group's suggestedIndex results.
+ if (suggestedIndexResults) {
+ let suggestedIndexUsedLimits = this._addSuggestedIndexResults(
+ suggestedIndexResults,
+ results,
+ state
+ );
+ for (let [key, value] of Object.entries(suggestedIndexUsedLimits)) {
+ usedLimits[key] += value;
+ }
+ }
+
+ return [results, usedLimits, hasMoreResults];
+ }
+
+ /**
+ * Helper for `_fillGroup` that fills a group's children.
+ *
+ * @param {object} group
+ * The result group to fill. It's assumed to have a `children` property.
+ * @param {object} limits
+ * An object with optional `availableSpan` and `maxResultCount` properties
+ * as described in `_fillGroup`.
+ * @param {object} state
+ * The muxer state.
+ * @param {Array} flexDataArray
+ * See `_updateFlexData`.
+ * @returns {Array}
+ * `[results, usedLimits, hasMoreResults]` -- see `_addResults`.
+ */
+ _fillGroupChildren(group, limits, state, flexDataArray = null) {
+ // If the group has flexed children, update the data we use during flex
+ // calculations.
+ //
+ // Handling flex is complicated so we discuss it briefly. We may do multiple
+ // passes for a group with flexed children in order to try to optimally fill
+ // them. If after one pass some children do not fill up but others do, we'll
+ // do another pass that tries to overfill the filled-up children while still
+ // respecting their flex ratios. We'll continue to do passes until all
+ // children stop filling up or we reach the parent's limits. The way we
+ // overfill children is by increasing their individual limits to make up for
+ // the unused space in their underfilled siblings. Before starting a new
+ // pass, we discard the results from the current pass so the new pass starts
+ // with a clean slate. That means we need to copy the global sort state
+ // (`state`) before modifying it in the current pass so we can use its
+ // original value in the next pass [1].
+ //
+ // [1] Instead of starting each pass with a clean slate in this way, we
+ // could accumulate results with each pass since we only ever add results to
+ // flexed children and never remove them. However, that would subvert muxer
+ // logic related to the global state (deduping, `_canAddResult`) since we
+ // generally assume the muxer adds results in the order they appear.
+ let stateCopy;
+ if (group.flexChildren) {
+ stateCopy = this._copyState(state);
+ flexDataArray = this._updateFlexData(group, limits, flexDataArray);
+ }
+
+ // Fill each child group, collecting all results in the `results` array.
+ let results = [];
+ let usedLimits = {};
+ for (let key of Object.keys(limits)) {
+ usedLimits[key] = 0;
+ }
+ let anyChildUnderfilled = false;
+ let anyChildHasMoreResults = false;
+ for (let i = 0; i < group.children.length; i++) {
+ let child = group.children[i];
+ let flexData = flexDataArray?.[i];
+
+ // Compute the child's limits.
+ let childLimits = {};
+ for (let key of Object.keys(limits)) {
+ childLimits[key] = flexData
+ ? flexData.limits[key]
+ : Math.min(
+ typeof child[key] == "number" ? child[key] : Infinity,
+ limits[key] - usedLimits[key]
+ );
+ }
+
+ // Recurse and fill the child.
+ let [
+ childResults,
+ childUsedLimits,
+ childHasMoreResults,
+ ] = this._fillGroup(child, childLimits, state);
+ results = results.concat(childResults);
+ for (let key of Object.keys(usedLimits)) {
+ usedLimits[key] += childUsedLimits[key];
+ }
+ anyChildHasMoreResults = anyChildHasMoreResults || childHasMoreResults;
+
+ if (flexData?.hasMoreResults) {
+ // The child is flexed and we possibly added more results to it.
+ flexData.usedLimits = childUsedLimits;
+ flexData.hasMoreResults = childHasMoreResults;
+ anyChildUnderfilled =
+ anyChildUnderfilled ||
+ (!childHasMoreResults &&
+ [...Object.entries(childLimits)].every(
+ ([key, limit]) => flexData.usedLimits[key] < limit
+ ));
+ }
+ }
+
+ // If the children are flexed and some underfilled but others still have
+ // more results, do another pass.
+ if (anyChildUnderfilled && anyChildHasMoreResults) {
+ [results, usedLimits, anyChildHasMoreResults] = this._fillGroupChildren(
+ group,
+ limits,
+ stateCopy,
+ flexDataArray
+ );
+
+ // Update `state` in place so that it's also updated in the caller.
+ for (let [key, value] of Object.entries(stateCopy)) {
+ state[key] = value;
+ }
+ }
+
+ return [results, usedLimits, anyChildHasMoreResults];
+ }
+
+ /**
+ * Updates flex-related state used while filling a group.
+ *
+ * @param {object} group
+ * The result group being filled.
+ * @param {object} limits
+ * An object defining the group's limits as described in `_fillGroup`.
+ * @param {Array} flexDataArray
+ * An array parallel to `group.children`. The object at index i corresponds
+ * to the child in `group.children` at index i. Each object maintains some
+ * flex-related state for its child and is updated during each pass in
+ * `_fillGroup` for `group`. When this method is called in the first pass,
+ * this argument should be null, and the method will create and return a new
+ * `flexDataArray` array that should be used in the remainder of the first
+ * pass and all subsequent passes.
+ * @returns {Array}
+ * A new `flexDataArray` when called in the first pass, and `flexDataArray`
+ * itself when called in subsequent passes.
+ */
+ _updateFlexData(group, limits, flexDataArray) {
+ flexDataArray =
+ flexDataArray ||
+ group.children.map((child, index) => {
+ let data = {
+ // The index of the corresponding child in `group.children`.
+ index,
+ // The child's limits.
+ limits: {},
+ // The fractional parts of the child's unrounded limits; see below.
+ limitFractions: {},
+ // The used-up portions of the child's limits.
+ usedLimits: {},
+ // True if `state.resultsByGroup` has more results of the child's
+ // `RESULT_GROUP`. This is not related to the child's limits.
+ hasMoreResults: true,
+ // The child's flex value.
+ flex: typeof child.flex == "number" ? child.flex : 0,
+ };
+ for (let key of Object.keys(limits)) {
+ data.limits[key] = 0;
+ data.limitFractions[key] = 0;
+ data.usedLimits[key] = 0;
+ }
+ return data;
+ });
+
+ // The data objects for children with more results (i.e., that are still
+ // fillable).
+ let fillableDataArray = [];
+
+ // The sum of the flex values of children with more results.
+ let fillableFlexSum = 0;
+
+ for (let data of flexDataArray) {
+ if (data.hasMoreResults) {
+ fillableFlexSum += data.flex;
+ fillableDataArray.push(data);
+ }
+ }
+
+ // Update each limit.
+ for (let [key, limit] of Object.entries(limits)) {
+ // Calculate the group's limit only including children with more results.
+ let fillableLimit = limit;
+ for (let data of flexDataArray) {
+ if (!data.hasMoreResults) {
+ fillableLimit -= data.usedLimits[key];
+ }
+ }
+
+ // Allow for the possibility that some children may have gone over limit.
+ // `fillableLimit` will be negative in that case.
+ fillableLimit = Math.max(fillableLimit, 0);
+
+ // Next we'll compute the limits of children with more results. This value
+ // is the sum of those limits. It may differ from `fillableLimit` due to
+ // the fact that each individual child limit must be an integer.
+ let summedFillableLimit = 0;
+
+ // Compute the limits of children with more results. If there are also
+ // children that don't have more results, then these new limits will be
+ // larger than they were in the previous pass.
+ for (let data of fillableDataArray) {
+ let unroundedLimit = fillableLimit * (data.flex / fillableFlexSum);
+ // `limitFraction` is the fractional part of the unrounded ideal limit.
+ // e.g., for 5.234 it will be 0.234. We use this to minimize the
+ // mathematical error when tweaking limits below.
+ data.limitFractions[key] = unroundedLimit - Math.floor(unroundedLimit);
+ data.limits[key] = Math.round(unroundedLimit);
+ summedFillableLimit += data.limits[key];
+ }
+
+ // As mentioned above, the sum of the individual child limits may not
+ // equal the group's fillable limit. If the sum is smaller, the group will
+ // end up with too few results. If it's larger, the group will have the
+ // correct number of results (since we stop adding results once limits are
+ // reached) but it may end up with a composition that does not reflect the
+ // child flex ratios as accurately as possible.
+ //
+ // In either case, tweak the individual limits so that (1) their sum
+ // equals the group's fillable limit, and (2) the composition respects the
+ // flex ratios with as little mathematical error as possible.
+ if (summedFillableLimit != fillableLimit) {
+ // Collect the flex datas with a non-zero limit fractions. We'll round
+ // them up or down depending on whether the sum is larger or smaller
+ // than the group's fillable limit.
+ let fractionalDataArray = fillableDataArray.filter(
+ data => data.limitFractions[key]
+ );
+
+ let diff;
+ if (summedFillableLimit < fillableLimit) {
+ // The sum is smaller. We'll increment individual limits until the sum
+ // is equal, starting with the child whose limit fraction is closest
+ // to 1 in order to minimize error.
+ diff = 1;
+ fractionalDataArray.sort((a, b) => {
+ // Sort by fraction descending so larger fractions are first.
+ let cmp = b.limitFractions[key] - a.limitFractions[key];
+ // Secondarily sort by index ascending so that children with the
+ // same fraction are incremented in the order they appear, allowing
+ // earlier children to have larger spans.
+ return cmp || a.index - b.index;
+ });
+ } else if (fillableLimit < summedFillableLimit) {
+ // The sum is larger. We'll decrement individual limits until the sum
+ // is equal, starting with the child whose limit fraction is closest
+ // to 0 in order to minimize error.
+ diff = -1;
+ fractionalDataArray.sort((a, b) => {
+ // Sort by fraction ascending so smaller fractions are first.
+ let cmp = a.limitFractions[key] - b.limitFractions[key];
+ // Secondarily sort by index descending so that children with the
+ // same fraction are decremented in reverse order, allowing earlier
+ // children to retain larger spans.
+ return cmp || b.index - a.index;
+ });
+ }
+
+ // Now increment or decrement individual limits until their sum is equal
+ // to the group's fillable limit.
+ while (summedFillableLimit != fillableLimit) {
+ if (!fractionalDataArray.length) {
+ // This shouldn't happen, but don't let it break us.
+ lazy.logger.error("fractionalDataArray is empty!");
+ break;
+ }
+ let data = flexDataArray[fractionalDataArray.shift().index];
+ data.limits[key] += diff;
+ summedFillableLimit += diff;
+ }
+ }
+ }
+
+ return flexDataArray;
+ }
+
+ /**
+ * Adds results to a group using the results from its `RESULT_GROUP` in
+ * `state.resultsByGroup`.
+ *
+ * @param {UrlbarUtils.RESULT_GROUP} groupConst
+ * The group's `RESULT_GROUP`.
+ * @param {object} limits
+ * An object defining the group's limits as described in `_fillGroup`.
+ * @param {object} state
+ * Global state that we use to make decisions during this sort.
+ * @returns {Array}
+ * `[results, usedLimits, hasMoreResults]` where:
+ * results: A flat array of results in the group, empty if no results
+ * were added.
+ * usedLimits: An object defining the amount of each limit that the
+ * results use. For each possible limit property (see `_fillGroup`),
+ * there will be a corresponding property in this object. For example,
+ * if 3 results are added with a total span of 4, then this object will
+ * be: { maxResultCount: 3, availableSpan: 4 }
+ * hasMoreResults: True if `state.resultsByGroup` has more results of
+ * the same `RESULT_GROUP`. This is not related to the group's limits.
+ */
+ _addResults(groupConst, limits, state) {
+ let usedLimits = {};
+ for (let key of Object.keys(limits)) {
+ usedLimits[key] = 0;
+ }
+
+ // For form history, maxHistoricalSearchSuggestions == 0 implies the user
+ // has opted out of form history completely, so we override the max result
+ // count here in that case. Other values of maxHistoricalSearchSuggestions
+ // are ignored and we use the flex defined on the form history group.
+ if (
+ groupConst == UrlbarUtils.RESULT_GROUP.FORM_HISTORY &&
+ !lazy.UrlbarPrefs.get("maxHistoricalSearchSuggestions")
+ ) {
+ // Create a new `limits` object so we don't modify the caller's.
+ limits = { ...limits };
+ limits.maxResultCount = 0;
+ }
+
+ let addedResults = [];
+ let groupResults = state.resultsByGroup.get(groupConst);
+ while (
+ groupResults?.length &&
+ state.usedResultSpan < state.availableResultSpan &&
+ [...Object.entries(limits)].every(([k, limit]) => usedLimits[k] < limit)
+ ) {
+ let result = groupResults[0];
+ if (this._canAddResult(result, state)) {
+ let span = UrlbarUtils.getSpanForResult(result);
+ let newUsedSpan = usedLimits.availableSpan + span;
+ if (limits.availableSpan < newUsedSpan) {
+ // Adding the result would exceed the group's available span, so stop
+ // adding results to it. Skip the shift() below so the result can be
+ // added to later groups.
+ break;
+ }
+ addedResults.push(result);
+ usedLimits.availableSpan = newUsedSpan;
+ usedLimits.maxResultCount++;
+ state.usedResultSpan += span;
+ this._updateStatePostAdd(result, state);
+ }
+
+ // We either add or discard results in the order they appear in
+ // `groupResults`, so shift() them off. That way later groups with the
+ // same `RESULT_GROUP` won't include results that earlier groups have
+ // added or discarded.
+ groupResults.shift();
+ }
+
+ return [addedResults, usedLimits, !!groupResults?.length];
+ }
+
+ /**
+ * Returns whether a result can be added to its group given the current sort
+ * state.
+ *
+ * @param {UrlbarResult} result
+ * The result.
+ * @param {object} state
+ * Global state that we use to make decisions during this sort.
+ * @returns {boolean}
+ * True if the result can be added and false if it should be discarded.
+ */
+ // TODO (Bug 1741273): Refactor this method to avoid an eslint complexity
+ // error or increase the complexity threshold.
+ // eslint-disable-next-line complexity
+ _canAddResult(result, state) {
+ // QuickSuggest results are shown unless they are best matches
+ // that duplicate heuristic results.
+ if (result.providerName == lazy.UrlbarProviderQuickSuggest.name) {
+ let heuristicUrl = state.context.heuristicResult?.payload.url;
+ if (
+ heuristicUrl &&
+ !lazy.UrlbarPrefs.get("experimental.hideHeuristic") &&
+ result.isBestMatch
+ ) {
+ let opts = {
+ stripHttp: true,
+ stripHttps: true,
+ stripWww: true,
+ trimSlash: true,
+ };
+ return (
+ UrlbarUtils.stripPrefixAndTrim(heuristicUrl, opts)[0] !=
+ UrlbarUtils.stripPrefixAndTrim(result.payload.url, opts)[0]
+ );
+ }
+ return true;
+ }
+
+ // We expect UrlbarProviderPlaces sent us the highest-ranked www. and non-www
+ // origins, if any. Now, compare them to each other and to the heuristic
+ // result.
+ //
+ // 1. If the heuristic result is lower ranked than both, discard the www
+ // origin, unless it has a different page title than the non-www
+ // origin. This is a guard against deduping when www.site.com and
+ // site.com have different content.
+ // 2. If the heuristic result is higher than either the www origin or
+ // non-www origin:
+ // 2a. If the heuristic is a www origin, discard the non-www origin.
+ // 2b. If the heuristic is a non-www origin, discard the www origin.
+ if (
+ !result.heuristic &&
+ result.type == UrlbarUtils.RESULT_TYPE.URL &&
+ result.payload.url
+ ) {
+ let [strippedUrl, prefix] = UrlbarUtils.stripPrefixAndTrim(
+ result.payload.url,
+ {
+ stripHttp: true,
+ stripHttps: true,
+ stripWww: true,
+ trimEmptyQuery: true,
+ }
+ );
+ let topPrefixData = state.strippedUrlToTopPrefixAndTitle.get(strippedUrl);
+ // If the condition below is not met, we are deduping a result against
+ // itself.
+ if (
+ topPrefixData &&
+ (prefix != topPrefixData.prefix ||
+ result.providerName != topPrefixData.providerName)
+ ) {
+ let prefixRank = UrlbarUtils.getPrefixRank(prefix);
+ if (
+ (prefixRank < topPrefixData.rank &&
+ (prefix.endsWith("www.") == topPrefixData.prefix.endsWith("www.") ||
+ result.payload?.title == topPrefixData.title)) ||
+ (prefix == topPrefixData.prefix &&
+ result.providerName != topPrefixData.providerName)
+ ) {
+ return false;
+ }
+ }
+ }
+
+ // Discard results that dupe autofill.
+ if (
+ state.context.heuristicResult &&
+ state.context.heuristicResult.autofill &&
+ !result.autofill &&
+ state.context.heuristicResult.payload?.url == result.payload.url &&
+ state.context.heuristicResult.type == result.type &&
+ !lazy.UrlbarPrefs.get("experimental.hideHeuristic")
+ ) {
+ return false;
+ }
+
+ // HeuristicFallback may add non-heuristic results in some cases, but those
+ // should be retained only if the heuristic result comes from it.
+ if (
+ !result.heuristic &&
+ result.providerName == "HeuristicFallback" &&
+ state.context.heuristicResult?.providerName != "HeuristicFallback"
+ ) {
+ return false;
+ }
+
+ if (result.providerName == lazy.UrlbarProviderTabToSearch.name) {
+ // Discard the result if a tab-to-search result was added already.
+ if (!state.canAddTabToSearch) {
+ return false;
+ }
+
+ // In cases where the heuristic result is not a URL and we have a
+ // tab-to-search result, the tab-to-search provider determined that the
+ // typed string is similar to an engine domain. We can let the
+ // tab-to-search result through.
+ if (state.context.heuristicResult?.type == UrlbarUtils.RESULT_TYPE.URL) {
+ // Discard the result if the heuristic result is not autofill and we are
+ // not making an exception for a fuzzy match.
+ if (
+ !state.context.heuristicResult.autofill &&
+ !result.payload.satisfiesAutofillThreshold
+ ) {
+ return false;
+ }
+
+ let autofillHostname = new URL(
+ state.context.heuristicResult.payload.url
+ ).hostname;
+ let [autofillDomain] = UrlbarUtils.stripPrefixAndTrim(
+ autofillHostname,
+ {
+ stripWww: true,
+ }
+ );
+ // Strip the public suffix because we want to allow matching "domain.it"
+ // with "domain.com".
+ autofillDomain = UrlbarUtils.stripPublicSuffixFromHost(autofillDomain);
+ if (!autofillDomain) {
+ return false;
+ }
+
+ // For tab-to-search results, result.payload.url is the engine's domain
+ // with the public suffix already stripped, for example "www.mozilla.".
+ let [engineDomain] = UrlbarUtils.stripPrefixAndTrim(
+ result.payload.url,
+ {
+ stripWww: true,
+ }
+ );
+ // Discard if the engine domain does not end with the autofilled one.
+ if (!engineDomain.endsWith(autofillDomain)) {
+ return false;
+ }
+ }
+ }
+
+ // Discard "Search in a Private Window" if appropriate.
+ if (
+ result.type == UrlbarUtils.RESULT_TYPE.SEARCH &&
+ result.payload.inPrivateWindow &&
+ !state.canShowPrivateSearch
+ ) {
+ return false;
+ }
+
+ // Discard form history and remote suggestions that dupe previously added
+ // suggestions or the heuristic.
+ if (
+ result.type == UrlbarUtils.RESULT_TYPE.SEARCH &&
+ result.payload.lowerCaseSuggestion
+ ) {
+ let suggestion = result.payload.lowerCaseSuggestion.trim();
+ if (!suggestion || state.suggestions.has(suggestion)) {
+ return false;
+ }
+ }
+
+ // Discard tail suggestions if appropriate.
+ if (
+ result.type == UrlbarUtils.RESULT_TYPE.SEARCH &&
+ result.payload.tail &&
+ !state.canShowTailSuggestions
+ ) {
+ return false;
+ }
+
+ // Discard remote tab results that dupes another remote tab or a
+ // switch-to-tab result.
+ if (result.type == UrlbarUtils.RESULT_TYPE.REMOTE_TAB) {
+ if (state.addedRemoteTabUrls.has(result.payload.url)) {
+ return false;
+ }
+ let maybeDupeType = state.urlToTabResultType.get(result.payload.url);
+ if (maybeDupeType == UrlbarUtils.RESULT_TYPE.TAB_SWITCH) {
+ return false;
+ }
+ }
+
+ // Discard switch-to-tab results that dupes another switch-to-tab result.
+ if (
+ result.type == UrlbarUtils.RESULT_TYPE.TAB_SWITCH &&
+ state.addedSwitchTabUrls.has(result.payload.url)
+ ) {
+ return false;
+ }
+
+ // Discard history results that dupe either remote or switch-to-tab results.
+ if (
+ !result.heuristic &&
+ result.type == UrlbarUtils.RESULT_TYPE.URL &&
+ result.payload.url &&
+ state.urlToTabResultType.has(result.payload.url)
+ ) {
+ return false;
+ }
+
+ // Discard SERPs from browser history that dupe either the heuristic or
+ // previously added suggestions.
+ if (
+ result.source == UrlbarUtils.RESULT_SOURCE.HISTORY &&
+ result.type == UrlbarUtils.RESULT_TYPE.URL
+ ) {
+ let submission = Services.search.parseSubmissionURL(result.payload.url);
+ if (submission) {
+ let resultQuery = submission.terms.trim().toLocaleLowerCase();
+ if (state.suggestions.has(resultQuery)) {
+ // If the result's URL is the same as a brand new SERP URL created
+ // from the query string modulo certain URL params, then treat the
+ // result as a dupe and discard it.
+ let [newSerpURL] = UrlbarUtils.getSearchQueryUrl(
+ submission.engine,
+ submission.terms
+ );
+ if (
+ lazy.UrlbarSearchUtils.serpsAreEquivalent(
+ result.payload.url,
+ newSerpURL
+ )
+ ) {
+ return false;
+ }
+ }
+ }
+ }
+
+ // When in an engine search mode, discard URL results whose hostnames don't
+ // include the root domain of the search mode engine.
+ if (state.context.searchMode?.engineName && result.payload.url) {
+ let engine = Services.search.getEngineByName(
+ state.context.searchMode.engineName
+ );
+ if (engine) {
+ let searchModeRootDomain = lazy.UrlbarSearchUtils.getRootDomainFromEngine(
+ engine
+ );
+ let resultUrl = new URL(result.payload.url);
+ // Add a trailing "." to increase the stringency of the check. This
+ // check covers most general cases. Some edge cases are not covered,
+ // like `resultUrl` being ebay.mydomain.com, which would escape this
+ // check if `searchModeRootDomain` was "ebay".
+ if (!resultUrl.hostname.includes(`${searchModeRootDomain}.`)) {
+ return false;
+ }
+ }
+ }
+
+ // Discard history results that dupe the quick suggest result.
+ if (
+ state.quickSuggestResult &&
+ !result.heuristic &&
+ result.type == UrlbarUtils.RESULT_TYPE.URL &&
+ lazy.QuickSuggest.isURLEquivalentToResultURL(
+ result.payload.url,
+ state.quickSuggestResult
+ )
+ ) {
+ return false;
+ }
+
+ // Discard history results whose URLs were originally sponsored. We use the
+ // presence of a partner's URL search param to detect these. The param is
+ // defined in the pref below, which is also used for the newtab page.
+ if (
+ result.source == UrlbarUtils.RESULT_SOURCE.HISTORY &&
+ result.type == UrlbarUtils.RESULT_TYPE.URL
+ ) {
+ let param = Services.prefs.getCharPref(
+ "browser.newtabpage.activity-stream.hideTopSitesWithSearchParam"
+ );
+ if (param) {
+ let [key, value] = param.split("=");
+ let searchParams;
+ try {
+ ({ searchParams } = new URL(result.payload.url));
+ } catch (error) {}
+ if (
+ (value === undefined && searchParams?.has(key)) ||
+ (value !== undefined && searchParams?.getAll(key).includes(value))
+ ) {
+ return false;
+ }
+ }
+ }
+
+ // Heuristic results must always be the first result. If this result is a
+ // heuristic but we've already added results, discard it. Normally this
+ // should never happen because the standard result groups are set up so
+ // that there's always at most one heuristic and it's always first, but
+ // since result groups are stored in a pref and can therefore be modified
+ // by the user, we perform this check.
+ if (result.heuristic && state.usedResultSpan) {
+ return false;
+ }
+
+ // Google search engine might suggest a result for unit conversion with
+ // format that starts with "= ". If our UnitConversion can provide the
+ // result, we discard the suggestion of Google in order to deduplicate.
+ if (
+ result.type == UrlbarUtils.RESULT_TYPE.SEARCH &&
+ result.payload.engine == "Google" &&
+ result.payload.suggestion?.startsWith("= ") &&
+ state.hasUnitConversionResult
+ ) {
+ return false;
+ }
+
+ // Include the result.
+ return true;
+ }
+
+ /**
+ * Updates the global state that we use to make decisions during sort. This
+ * should be called for results before we've decided whether to add or discard
+ * them.
+ *
+ * @param {UrlbarResult} result
+ * The result.
+ * @param {object} state
+ * Global state that we use to make decisions during this sort.
+ */
+ _updateStatePreAdd(result, state) {
+ // Keep track of the largest heuristic result span.
+ if (result.heuristic && this._canAddResult(result, state)) {
+ state.maxHeuristicResultSpan = Math.max(
+ state.maxHeuristicResultSpan,
+ UrlbarUtils.getSpanForResult(result)
+ );
+ }
+
+ // Subtract from `availableResultSpan` the span of global suggestedIndex
+ // results so there will be room for them at the end of the sort. Except
+ // when `maxRichResults` is zero and other special cases, we assume
+ // suggestedIndex results will always be shown regardless of the total
+ // available result span, `context.maxResults`, and `maxRichResults`.
+ if (
+ result.hasSuggestedIndex &&
+ !result.isSuggestedIndexRelativeToGroup &&
+ this._canAddResult(result, state)
+ ) {
+ let span = UrlbarUtils.getSpanForResult(result);
+ if (result.providerName == lazy.UrlbarProviderTabToSearch.name) {
+ state.maxTabToSearchResultSpan = Math.max(
+ state.maxTabToSearchResultSpan,
+ span
+ );
+ } else {
+ state.availableResultSpan = Math.max(
+ state.availableResultSpan - span,
+ 0
+ );
+ }
+ }
+
+ // Save some state we'll use later to dedupe URL results.
+ if (
+ (result.type == UrlbarUtils.RESULT_TYPE.URL ||
+ result.type == UrlbarUtils.RESULT_TYPE.KEYWORD) &&
+ result.payload.url &&
+ (!result.heuristic || !lazy.UrlbarPrefs.get("experimental.hideHeuristic"))
+ ) {
+ let [strippedUrl, prefix] = UrlbarUtils.stripPrefixAndTrim(
+ result.payload.url,
+ {
+ stripHttp: true,
+ stripHttps: true,
+ stripWww: true,
+ trimEmptyQuery: true,
+ }
+ );
+ let prefixRank = UrlbarUtils.getPrefixRank(prefix);
+ let topPrefixData = state.strippedUrlToTopPrefixAndTitle.get(strippedUrl);
+ let topPrefixRank = topPrefixData ? topPrefixData.rank : -1;
+ if (
+ topPrefixRank < prefixRank ||
+ // If a quick suggest result has the same stripped URL and prefix rank
+ // as another result, store the quick suggest as the top rank so we
+ // discard the other during deduping. That happens after the user picks
+ // the quick suggest: The URL is added to history and later both a
+ // history result and the quick suggest may match a query.
+ (topPrefixRank == prefixRank &&
+ result.providerName == lazy.UrlbarProviderQuickSuggest.name)
+ ) {
+ // strippedUrl => { prefix, title, rank, providerName }
+ state.strippedUrlToTopPrefixAndTitle.set(strippedUrl, {
+ prefix,
+ title: result.payload.title,
+ rank: prefixRank,
+ providerName: result.providerName,
+ });
+ }
+ }
+
+ // Save some state we'll use later to dedupe results from open/remote tabs.
+ if (
+ result.payload.url &&
+ (result.type == UrlbarUtils.RESULT_TYPE.TAB_SWITCH ||
+ (result.type == UrlbarUtils.RESULT_TYPE.REMOTE_TAB &&
+ !state.urlToTabResultType.has(result.payload.url)))
+ ) {
+ // url => result type
+ state.urlToTabResultType.set(result.payload.url, result.type);
+ }
+
+ // If we find results other than the heuristic, "Search in Private
+ // Window," or tail suggestions, then we should hide tail suggestions
+ // since they're a last resort.
+ if (
+ state.canShowTailSuggestions &&
+ !result.heuristic &&
+ (result.type != UrlbarUtils.RESULT_TYPE.SEARCH ||
+ (!result.payload.inPrivateWindow && !result.payload.tail))
+ ) {
+ state.canShowTailSuggestions = false;
+ }
+
+ if (result.providerName == lazy.UrlbarProviderQuickSuggest.name) {
+ state.quickSuggestResult = result;
+ }
+
+ state.hasUnitConversionResult =
+ state.hasUnitConversionResult || result.providerName == "UnitConversion";
+ }
+
+ /**
+ * Updates the global state that we use to make decisions during sort. This
+ * should be called for results after they've been added. It should not be
+ * called for discarded results.
+ *
+ * @param {UrlbarResult} result
+ * The result.
+ * @param {object} state
+ * Global state that we use to make decisions during this sort.
+ */
+ _updateStatePostAdd(result, state) {
+ // Update heuristic state.
+ if (result.heuristic) {
+ state.context.heuristicResult = result;
+ if (
+ result.type == UrlbarUtils.RESULT_TYPE.SEARCH &&
+ result.payload.query &&
+ !lazy.UrlbarPrefs.get("experimental.hideHeuristic")
+ ) {
+ let query = result.payload.query.trim().toLocaleLowerCase();
+ if (query) {
+ state.suggestions.add(query);
+ }
+ }
+ }
+
+ // The "Search in a Private Window" result should only be shown when there
+ // are other results and all of them are searches. It should not be shown
+ // if the user typed an alias because that's an explicit engine choice.
+ if (
+ !Services.search.separatePrivateDefaultUrlbarResultEnabled ||
+ (state.canShowPrivateSearch &&
+ (result.type != UrlbarUtils.RESULT_TYPE.SEARCH ||
+ result.payload.providesSearchMode ||
+ (result.heuristic && result.payload.keyword)))
+ ) {
+ state.canShowPrivateSearch = false;
+ }
+
+ // Update suggestions.
+ if (
+ result.type == UrlbarUtils.RESULT_TYPE.SEARCH &&
+ result.payload.lowerCaseSuggestion
+ ) {
+ let suggestion = result.payload.lowerCaseSuggestion.trim();
+ if (suggestion) {
+ state.suggestions.add(suggestion);
+ }
+ }
+
+ // Avoid multiple tab-to-search results.
+ // TODO (Bug 1670185): figure out better strategies to manage this case.
+ if (result.providerName == lazy.UrlbarProviderTabToSearch.name) {
+ state.canAddTabToSearch = false;
+ // We want to record in urlbar.tips once per engagement per engine. Since
+ // whether these results are shown is dependent on the Muxer, we must
+ // add to `enginesShown` here.
+ if (result.payload.dynamicType) {
+ lazy.UrlbarProviderTabToSearch.enginesShown.onboarding.add(
+ result.payload.engine
+ );
+ } else {
+ lazy.UrlbarProviderTabToSearch.enginesShown.regular.add(
+ result.payload.engine
+ );
+ }
+ }
+
+ // Sync will send us duplicate remote tabs if multiple copies of a tab are
+ // open on a synced client. Keep track of which remote tabs we've added to
+ // dedupe these.
+ if (result.type == UrlbarUtils.RESULT_TYPE.REMOTE_TAB) {
+ state.addedRemoteTabUrls.add(result.payload.url);
+ }
+
+ // Keep track of which switch tabs we've added to dedupe switch tabs.
+ if (result.type == UrlbarUtils.RESULT_TYPE.TAB_SWITCH) {
+ state.addedSwitchTabUrls.add(result.payload.url);
+ }
+ }
+
+ /**
+ * Inserts results with suggested indexes. This can be called for either
+ * global or group-relative suggestedIndex results. It should be called after
+ * `sortedResults` has been filled in.
+ *
+ * @param {Array} suggestedIndexResults
+ * Results with a `suggestedIndex` property.
+ * @param {Array} sortedResults
+ * The sorted results. For global suggestedIndex results, this should be the
+ * final list of all results before suggestedIndex results are inserted. For
+ * group-relative suggestedIndex results, this should be the final list of
+ * results in the group before group-relative suggestedIndex results are
+ * inserted.
+ * @param {object} state
+ * Global state that we use to make decisions during this sort.
+ * @returns {object}
+ * A `usedLimits` object that describes the total span and count of all the
+ * added results. See `_addResults`.
+ */
+ _addSuggestedIndexResults(suggestedIndexResults, sortedResults, state) {
+ let usedLimits = {
+ availableSpan: 0,
+ maxResultCount: 0,
+ };
+
+ if (!suggestedIndexResults?.length) {
+ // This is just a slight optimization; no need to continue.
+ return usedLimits;
+ }
+
+ // Partition the results into positive- and negative-index arrays. Positive
+ // indexes are relative to the start of the list and negative indexes are
+ // relative to the end.
+ let positive = [];
+ let negative = [];
+ for (let result of suggestedIndexResults) {
+ let results = result.suggestedIndex < 0 ? negative : positive;
+ results.push(result);
+ }
+
+ // Sort the positive results ascending so that results at the end of the
+ // array don't end up offset by later insertions at the front.
+ positive.sort((a, b) => {
+ if (a.suggestedIndex !== b.suggestedIndex) {
+ return a.suggestedIndex - b.suggestedIndex;
+ }
+
+ if (a.providerName === b.providerName) {
+ return 0;
+ }
+
+ // If same suggestedIndex, change the displaying order along to following
+ // provider priority.
+ // TabToSearch > QuickSuggest > Other providers
+ if (a.providerName === lazy.UrlbarProviderTabToSearch.name) {
+ return 1;
+ }
+ if (b.providerName === lazy.UrlbarProviderTabToSearch.name) {
+ return -1;
+ }
+ if (a.providerName === lazy.UrlbarProviderQuickSuggest.name) {
+ return 1;
+ }
+ if (b.providerName === lazy.UrlbarProviderQuickSuggest.name) {
+ return -1;
+ }
+
+ return 0;
+ });
+
+ // Conversely, sort the negative results descending so that results at the
+ // front of the array don't end up offset by later insertions at the end.
+ negative.sort((a, b) => b.suggestedIndex - a.suggestedIndex);
+
+ // Insert the results. We start with the positive results because we have
+ // tests that assume they're inserted first. In practice it shouldn't matter
+ // because there's no good reason we would ever have a negative result come
+ // before a positive result in the same query. Even if we did, we have to
+ // insert one before the other, and there's no right or wrong order.
+ for (let results of [positive, negative]) {
+ let prevResult;
+ let prevIndex;
+ for (let result of results) {
+ if (this._canAddResult(result, state)) {
+ let index;
+ if (
+ prevResult &&
+ prevResult.suggestedIndex == result.suggestedIndex
+ ) {
+ index = prevIndex;
+ } else {
+ index =
+ result.suggestedIndex >= 0
+ ? Math.min(result.suggestedIndex, sortedResults.length)
+ : Math.max(result.suggestedIndex + sortedResults.length + 1, 0);
+ }
+ prevResult = result;
+ prevIndex = index;
+ sortedResults.splice(index, 0, result);
+ usedLimits.availableSpan += UrlbarUtils.getSpanForResult(result);
+ usedLimits.maxResultCount++;
+ this._updateStatePostAdd(result, state);
+ }
+ }
+ }
+
+ return usedLimits;
+ }
+}
+
+export var UrlbarMuxerUnifiedComplete = new MuxerUnifiedComplete();