magibney commented on a change in pull request #592:
URL: https://github.com/apache/solr/pull/592#discussion_r801238955
##########
File path: solr/core/src/java/org/apache/solr/search/SolrIndexSearcher.java
##########
@@ -865,9 +929,31 @@ public BitDocSet getLiveDocSet() throws IOException {
// Going through the filter cache will provide thread safety here if we
only had getLiveDocs,
// but the addition of setLiveDocs means we needed to add volatile to
"liveDocs".
BitDocSet docs = liveDocs;
- if (docs == null) {
- //note: maybe should instead calc manually by segment, using
FixedBitSet.copyOf(segLiveDocs); avoid filter cache?
- liveDocs = docs = getDocSetBits(matchAllDocsQuery);
+ if (docs != null) {
+ matchAllDocsCacheHitCount.incrementAndGet();
+ } else {
+ if (matchAllDocsCacheComputationTracker.compareAndSet(Long.MIN_VALUE,
0)) {
+ // run the initial/only/final future inline
+ // This thread will block execution here and `liveDocsFuture.get()`
(below) should then return immediately
+ liveDocsFuture.run();
+ } else {
+ // another thread has already called `computeLiveDocs.run()`; this
thread will block on
+ // `liveDocsFuture.get()` (below)
+ if (matchAllDocsCacheComputationTracker.getAndIncrement() < 0) {
Review comment:
:+1: ... any reasons I might have had for doing it this way are obviated
by following your suggestion below, which just leans on a single-element
CaffeineCache to handle the concurrency issues.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
