Search results should be easy to navigate, sort, and filter. Give your users an optimal search experience by understanding the default search experience and the many configuration options at your disposal.
Multiple factors affect search results:
How assets are indexed
The search widgets used
Whether results are post-processed
Developers of content types (assets in Liferay DXP) control much about how the asset’s information is indexed and how its information is searched and returned in the search results. For further control, an Indexer Post Processor can modify an asset’s indexing behavior and how search queries are constructed to look up the assets.
The concepts below are essential to understand before you begin changing any settings.
Filtering Results with Facets¶
Results are filtered using facets. Most users have encountered similar filtering capabilities in other applications, particularly during commerce activities. Users enter a search term, are presented with a list of results, and selecting search facets refines the results. You can think of a facet as a bucket of results with a shared characteristics.
Facets are configurable. Read about configuring facets to learn more.
Search Results Relevance¶
The search engine processes and orders results by relevance. Relevance is a score calculated by the search engine. The score is calculated by algorithms provided by the search engine.
Results relevance is configurable:
Permissions and Search Results¶
A search result doesn’t appear for a User lacking permission to view the asset. A logged in User with the Site Administrator Role likely sees more search results than an anonymous guest.
There are two rounds of permissions checks:
Pre-filtering happens in the search engine’s index. It’s faster than checking database permissions information, but occasionally the search index can have stale permissions information.
Post-filtering happens on the results prior to display, to ensure the search engine’s index has correct, up-to-date permissions information.
Pre-filtering adds filter clauses to the search query, so searches contain results the current User can view.
You can configure pre-filtering at Control Panel → Configuration → System Settings → Search → Permission Checker by controlling the number of search clauses added to queries:
Permissions Term Limit: Limits the number of permission search clauses added to the search query before this level of permission checking is aborted. Permission checking then relies solely on the final permission filtering described below.
The only reason to limit permissions terms is performance. Users with administrative access to lots of Sites and Organizations generate many permissions terms added to the query. Too many terms in a query can make the search engine time out.
Post-filtering happens prior to presenting results in the UI. For example, if a User searches for liferay, the search engine returns all relevant forum posts. As the Search Results iterates through this list, it performs one last permission check of each post to ensure the User can view the post and its categories. If a User doesn’t have permission to view the post, it isn’t displayed in the list of search results.
Post-filtering is configurable at Control Panel → Configuration → System Settings → Search → Default Search Result Permission Filter. It includes two settings:
Permission Filtered Search Result Accurate Count Threshold: Specify the maximum number of search results to permissions-filter before results are counted. A higher threshold increases count accuracy, but decreases performance. Since results in the currently displayed page are always checked, any value below the search results pagination delta effectively disables this behavior.
Search Query Result Window Limit: Set the maximum batch size for each permission checking request. This is impacted by pagination. If there are 100 results per page and you jump all the way to page 200 of the search results, all results between page one and 200 must be checked to ensure you have permission. That’s 20,000 results to permissions check. Doing this in one trip to and from the search engine can result in performance issues. Set the maximum batch size for each permission checking request.
Search and Staging¶
With staging, content is placed first in a preview and testing environment before being published to the live Site. Indexed content is marked so the search API knows if an item is live or not. In the live version of the Site, only live indexed content is searchable.
In the staged version of the Site, all content—live or staged—is searchable.
A result summary condenses information from the original asset into an abstract. Asset developers choose what fields are included in the summary. A common summary includes a title and some of the content, with title displayed first. The asset type always appears on the second line, followed by a snippet of content matching the search term. Assets without content fields, like Documents and Media documents, display the description instead.
Searching for Users: When you click an asset in the search results, it’s displayed in an Asset Publisher (unless the View in Context option is selected in the Search Results widget). Users are different, though. Think of them as invisible assets, not intended for display in the Asset Publisher application. While Users appear as search results with other indexed assets, when you click one you’re taken to the User’s profile page. If public personal pages are disabled, clicking on a User from the list of search results shows you a blank page. Only the User’s full name and the asset type (User) appear in User result summaries:
For assets containing other assets (Web Content and Documents & Media folders) or whose content is not amenable to display (Dynamic Data List Records and Calendar Events), it makes more sense to display the title, asset type, and description in results summaries:
Asset developers determine which fields are summary-enabled, but logic invoked at search time determines precisely the part of the summary fields to display. For example, a
content field can have a lot of text, but the summary only shows and highlights the relevant snippet of the field’s text containing the keyword.
Search terms appearing in the summary are highlighted by default. If this is undesirable, disable it in the widget configuration screen.
Highlighting is a helpful visual cue that hints at why the result is returned, but beware: high scoring hits can appear at the top of results without having any highlights in the summary. That’s because not all indexed fields appear in the summary. Consider a User named Arthur C. Clarke. He has a searchable email address of firstname.lastname@example.org. Because Users result summaries only contain their full names, searching for Mr. Clarke by his email address returns the User, but no term is highlighted.
There may be additional cases where search results don’t have highlighting.