* remove commented-out/obselete AwaitsFix
All of these issues are fixed, but the AwaitsFix annotation is still there, just commented out. This causes confusion and makes it harder to keep an eye/review the AwaitsFix tests, e.g. false positives when running 'git grep AwaitsFix'
* Remove @AwaitsFix from TestMatchRegionRetriever. The problem has been fixed in LUCENE-10229.
Co-authored-by: Dawid Weiss <dawid.weiss@carrotsearch.com>
This is very noisy, can contain gradle status updates, various other tests.verbose prints from other threads, you name it.
It causes the check to be flaky, and randomly "miss" seeing a test that executed.
Instead, let's look at the zip files. We can still preserve the essence of what the test wants to do, but without any flakiness.
- MergeOnFlushMergePolicy doesn't try to avoid O(n^2) merges, so I'm disabling
the test on it for now.
- TestUpgradeIndexMergePolicy would sometimes wrap with a non-standard merge
policy like the alcoholic merge policy, I forced it to wrap a
TieredMergePolicy.
LogDocMergePolicy would previously always force-merge an index that has 10
segments of size 1 to 10, due to the min doc count. This is not the case
anymore, but the test was assuming that such an index would get merged, so I
fixed the test's expectations.
Also changed the merge policy to keep working when RAM buffers are flushed in
such a way that segments do not appear in decreasing size order using the same
logic as LogMergePolicy.
If a LogByteSizeMergePolicy is used, then it might decide to not merge the two
one-document segments if their on-disk sizes are too different. Using a
LogDocMergePolicy addresses the issue as both segments are always considered
the same size.
This updates TieredMergePolicy and Log(Doc|Size)MergePolicy to only ever
consider merges where the resulting segment would be at least 50% bigger than
the biggest input segment. While a merge that only grows the biggest segment by
50% is still quite inefficient, this constraint is good enough to prevent
pathological O(N^2) merging.
- Removed dependency on LineFileDocs to improve reproducibility.
- Relaxed the expected exception type: any exception is ok.
- Ignore rare cases when a file still appears to have a well-formed footer
after truncation.
The original HNSW paper (https://arxiv.org/pdf/1603.09320.pdf) suggests
to use a different maxConn for the upper layers vs. the bottom one
(which contains the full neighborhood graph). Specifically, they
suggest using maxConn=M for upper layers and maxConn=2*M for the bottom.
This patch ensures that we follow this recommendation and use
maxConn=2*M for the bottom layer.
Doc values terms dictionaries keep the first term of each block uncompressed so
that they can somewhat efficiently perform binary searches across blocks.
Suffixes of the other 63 terms are compressed together using LZ4 to leverage
redundancy across suffixes. This change improves compression a bit by using the
first (uncompressed) term of each block as a dictionary when compressing
suffixes of the 63 other terms. This helps with compressing the first few
suffixes when there's not much context yet that can be leveraged to find
duplicates.
Lu Xugang
Tomoko Uchida
Michael Sokolov
tang donghai
Yannick Welsch
Adrien Grand
Greg Miller
András Salamon
Alan Woodward
Robert Muir
Dawid Weiss
Shai Erera
Rushabh Shah
sunwq
Uwe Schindler
Yuting Gan
Julie Tibshirani
Mayya Sharipova