See documents that fail to index in the CMS. With version 4.1.0 of the Search SDK, a document that fails to index is recorded against the record it came from and listed in the Search Indexing admin, with the reason and, where one is available, a stack trace. You can retry or clear a failure one row at a time or in bulk. Retrying all failures resumes an interrupted index job where there is one, and queues the rest as one job per index. You can also choose to record documents that were skipped because they are not publicly viewable. To use this, update the SDK with composer require silverstripe/silverstripe-search-sdk:^4.1, then run a dev/build, which adds the table the failures are stored in. See the silverstripe-forager documentation for how to use it.
Dashboard refinements. The help menu lists the service desk and the search documentation. The user menu links to your account settings. Tables scroll horizontally on narrow screens instead of overflowing, and a flyout panel no longer dims the page header.
Documents the search service rejects are no longer reported as indexed. The search service accepts a batch of documents as a whole and reports a rejection against each document it refuses. The SDK reported every document in the batch as indexed, so a rejected document looked indexed and the reason was not recorded. With SDK 4.1.0 a rejected document is recorded as a failure with the reason the service gave, and appears in the Search Indexing admin.
The Search Indexing admin opens when the document count cannot be read. If the SDK could not read the number of documents in an index from the search service, the admin failed with an error. It now opens and shows the problem in the admin.
Top queries reports load for long date ranges. The top queries reports could fail with an error for a date range near the 91-day maximum. They now load for any date range up to the maximum.
Clicks are counted for each analytics tag. In the top queries with clicks and top queries with no clicks reports, clicks on a query were counted by the query's text alone. Where the same query was searched with different tags, each tag received the click total for every tag. Clicks are now counted against the tags of the search they came from. If your site sends tags, the click counts you see per tag will be lower than before, and they add up to the total. If your site does not send tags, the counts do not change.