[SOLR-12374] Add SolrCore.withSearcher(lambda accepting SolrIndexSearcher)

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 7.4
• Assignee: David Smiley
• Reporter: David Smiley
• Created: 18/May/18 02:55
• Updated: 21/Nov/19 00:41
• Resolved: 29/May/18 21:25

Description

I propose adding the following to SolrCore:

  /**
   * Executes the lambda with the {@link SolrIndexSearcher}.  This is more convenience than using
   * {@link #getSearcher()} since there is no ref-counting business to worry about.
   * Example:
   * <pre class="prettyprint">
   *   IndexReader reader = h.getCore().withSearcher(SolrIndexSearcher::getIndexReader);
   * </pre>
   */
  @SuppressWarnings("unchecked")
  public <R> R withSearcher(Function<SolrIndexSearcher,R> lambda) {
    final RefCounted<SolrIndexSearcher> refCounted = getSearcher();
    try {
      return lambda.apply(refCounted.get());
    } finally {
      refCounted.decref();
    }
  }

This is a nice tight convenience method, avoiding the clumsy RefCounted API which is easy to accidentally incorrectly use – see https://issues.apache.org/jira/browse/SOLR-11616?focusedCommentId=16477719&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16477719

I guess my only (small) concern is if hypothetically you might make the lambda short because it's easy to do that (see the one-liner example above) but the object you return that you're interested in  (say IndexReader) could potentially become invalid if the SolrIndexSearcher closes.  But I think/hope that's impossible normally based on when this getSearcher() used?  I could at least add a warning to the docs.