Catalog search #

This page is for library and archive managers deciding how readers will search the catalog, and for the website creator who sets it up. It explains what the built-in search does, how to learn from what readers search for, and when the extra work of a Solr server through ezfind pays off.

Contents #

1. Two engines
2. The built-in search
3. What gets indexed
4. Searching inside files
5. Learning from what readers search
6. When Solr is worth it
7. Switching to Solr, in outline
8. Browsing complements searching
9. Search checklist

Two engines #

Sources: Search with ezfind and Solr and the suggest block of composer.json.

The built-in search #

• site.ini [SearchSettings] SearchEngine=eZSearchEngine indexes the attributes marked Searchable when content is published. "Visitors only find what they may read", so restricted records stay hidden from the public.
• content/search takes a search box; content/advancedsearch offers more fields (phrase, class, date), useful for a catalog: "books only, published after 2000".
• Shipped limits include MaximumSearchLimit=30 results per page and wildcards off (EnableWildcard=false), because term* searches are "costly on large sites".

Source: Site search.

What gets indexed #

Mark only useful fields Searchable: "titles, intros, bodies; not internal codes" (same page). For a catalog, also mark the identifiers readers type, such as an ISBN, and the creators' names on Person objects.

Subject headings from eztags are indexed with their synonyms by default (IndexSynonyms=enabled), so a reader who types a non-preferred term still finds the record (Tagging and subject headings).

After a large import, rebuild the whole index:

php bin/php/updatesearchindex.php --clean --allow-root-user

With DelayedIndexing enabled, content is indexed by cron instead of on publish, which keeps publishing fast during imports (Site search, Indexing).

Searching inside files #

settings/binaryfile.ini has handlers that extract text from PDF and Word files for the search index ([PDFHandlerSettings], [WordHandlerSettings]), "using external programs you install". With Solr, file contents can be indexed too (Files and downloads, Searching inside files). Ask your hosting partner to install the extraction programs if digitized documents should be found by their text.

Learning from what readers search #

With LogSearchStats=enabled on the public siteaccess, Setup > Search statistics (/search/stats) shows every phrase, how often it was searched and the average number of results; "phrases that found nothing show what content is missing". Phrases can be personal data, which is why logging is off by default (Site search, Search statistics).

Use it monthly: searches with no results point to missing records, missing synonyms, or spelling variants to add as tag synonyms. Decide with your data protection contact whether to switch it on and how long to keep the phrases.

When Solr is worth it #

Consider Solr when one of these is true:

• To do: Readers need to narrow results by facets (format, subject, date, language).
• To do: The catalog has grown to the point where searches slow down.
• To do: Relevance ranking matters (best matches first) and the built-in ranking is not good enough.
• To do: The full text of many files must be searchable.

And weigh the cost: "Solr is a separate server (Java) to install, run, monitor and back up. For a small site the built-in engine is fine" (Search with ezfind and Solr). For hosting, see Search backends.

Switching to Solr, in outline #

1. Install a Solr server (locally, or a remote one).
2. Set [SolrSettings] Enabled=true and its address in settings/override/solr.ini.append.php.
3. Check with ./console exp:solr status, then ping.
4. Install and activate ezfind; its settings switch [SearchSettings] SearchEngine to ezsolr.
5. Clear the INI cache and rebuild the index.

Source: Search with ezfind and Solr, Switching it on and Server control commands. content/search keeps working; facets need templates that use ezfind's own fetch functions.

Browsing complements searching #

Many readers browse rather than search. Give them:

• subject pages from tags (Tagging and subject headings);
• collections as containers in the menu (Modeling catalog records);
• a "new acquisitions" list and feed (Feeds for new acquisitions).

Search checklist #

• To do: Searchable set on purpose for each field of each record class.
• To do: A search results template that shows the number of results, paging and what to try when nothing is found.
• To do: Advanced search offered if readers need class or date filters.
• To do: Index rebuilt after every large import.
• To do: Text extraction programs installed if file contents must be found.
• To do: Search statistics decided (on or off, retention) with the data protection contact.
• To do: A dated decision on Solr: not yet, or planned with a hosting partner.

Related pages #

• Libraries and catalogs
• Site search
• Search with ezfind and Solr
• Searching (for editors)
• Free documentation: this page is free to copy, translate and improve under the GNU FDL 1.3.