used in parallel() streaming expression. Hash algorithm is different.
* Simpler
* Don't use Filter (to be removed)
* Do use TwoPhaseIterator, not PostFilter
* Don't pre-compute matching docs (wasteful)
* Support more fields, and more field types
* Faster hash on Strings (avoid Char conversion)
* Stronger hash when using multiple fields
For/PFor code has BLOCK_SIZE=128 as a static final constant, with a lot
of assumptions and optimizations for that case. For example it will
encode 3 exceptions at most and optimizes the exception encoding with a
single byte.
This would not work at all if you changed the constant in the code to
something like 512, but an assertion at an early stage helps make
experimentation less painful, and better "documents" the assumption of how
the exception encoding currently works.
LZ4-HC hashtable is heavy (128kb int[] + 128kb short[]) and must be
filled with special values on initialization. This is a lot of overhead
for fields that might not use the compression at all.
Don't initialize this for a field until we see hints that the data might
be compressible and need to use the table in order to test it out.
* Fix JSON Faceting on EnumFieldType if allBuckets, numBuckets or missing is set.
* Enhance hash method of JSON faceting to support EnumFieldType and perhaps some other/custom field types
Co-authored-by: Thomas Wöckinger <two@silbergrau.com>
Co-authored-by: David Smiley <dsmiley@apache.org>
The 7.0 backwards tests added to master must have come from an older
branch before they were fixed: they've added minutes to my test times.
These tests have already been fixed in master, so that the crazy
corner-case stress tests are only running slowly in jekins and we dont
have 15-30s long tests locally.
Re-applying same fixes to 7.0 tests removes minutes from my test times.
Partial (AKA Atomic) updates could encounter "LazyField" instances in the document
cache and not know hot to deal with them when writing the updated doc to the update log.
* LUCENE-9810: Hunspell: when generating suggestions, skip too deep word FST subtrees
we skip roots longer than misspelled+4 anyway, so there's no need to read their arcs
* check more in TestPerformance.de_suggest
* LUCENE-9806: Hunspell: speed up affix condition checking
check only stem beginning/end without strip/condition, not the whole candidate
avoid regexp if possible
* hunspell: simplify AffixCondition, add more tests
* add a license to the test
Peter Gromov
Joel Bernstein
David Smiley
Robert Muir
Bruno Roustant
noblepaul
Thomas Wöckinger
Eric Pugh
Christine Poerschke
zacharymorn
百岁
Michael Sokolov
Dawid Weiss
Andrzej Bialecki
Ilan Ginzburg
Thomas Wöckinger
Ignacio Vera
Timothy Potter