summaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* Report transient resource usage for all attributes and document meta store.Geir Storli2022-12-0620-17/+97
|
* Merge pull request #25132 from ↵Geir Storli2022-12-063-2/+6
|\ | | | | | | | | vespa-engine/toregge/wait-for-load-done-before-providing-transient-resource-usage Wait for load done before providing transient resource usage.
| * Wait for load done before providing transient resource usage.Tor Egge2022-12-063-2/+6
| |
* | Turn off FullSearch optimization for active lids blueprint createLeafSearch().Tor Egge2022-12-062-16/+30
|/
* Report transient disk usage for attribute disk snapshots.Geir Storli2022-12-023-15/+153
| | | | All snapshots except the best one count towards transient disk usage.
* Rewrite attribute directory tests to GTest.Geir Storli2022-12-012-85/+84
|
* GC long gone fdispatchrcHenning Baldersheim2022-11-262-7/+0
|
* Update unit test for handling removes in attribute writer.Tor Egge2022-11-251-16/+10
|
* Don't apply remove to attribute if last sync token on attribute is equal to ↵Tor Egge2022-11-251-2/+1
| | | | | | | | | | | serialNum. Several years ago, attribute vector last sync token was updated for each write operation, and lid space compaction generated a move operation that performed two write operations on the same attribute with the same serial number (setting the value for the new lid and clearing the value for the old lid). Current code delays update of attribute vector last sync token until a force commit operation or an attribute vector flush (to disk).
* Allow override of keep-rank-count via the internal ↵Henning Baldersheim2022-11-181-1/+3
| | | | vespa.hitcollector.arraysize rank param.
* Drop support for old gtest.Tor Egge2022-11-181-1/+1
|
* fix typo invokation -> invocationThinh Bui2022-11-141-1/+1
|
* Add a trace if the global filter matches everything.Geir Storli2022-11-011-1/+4
| | | | This case enables an optimization to search the HNSW index without using the filter at all.
* explicit default filterHåvard Pettersen2022-10-311-0/+3
|
* * apply new common bootstrapArne Juul2022-10-241-2/+4
| | | | | * remove now-duplicated code * prefer using ${VESPA_HOME} environment variable
* Use search::test::SchemaBuilder in searchcore unit tests.Tor Egge2022-10-218-79/+59
|
* Return the full search iterator when all lids are active.Geir Storli2022-10-202-2/+48
| | | | | This opens up for optimizations of the global filter iterator tree based on SearchIterator::matches_any().
* Stop testing with url data type, use uri data type instead.Tor Egge2022-10-191-45/+8
|
* Add missing implementations file iflushtarget.cppHenning Baldersheim2022-10-181-0/+19
|
* Use std::make_sharedHenning Baldersheim2022-10-181-17/+16
|
* Deinline IFlushTarget constructor and destructorHenning Baldersheim2022-10-182-14/+7
|
* Revert "Revert "Activate using gid for [docid] sorting.""Henning Baldersheim2022-10-161-1/+2
|
* Revert "Activate using gid for [docid] sorting."Geir Storli2022-10-151-2/+1
|
* Activate using gid for [docid] sorting.Henning Baldersheim2022-10-141-1/+2
|
* Handle that [docid] can use an optional attribute.Henning Baldersheim2022-10-142-6/+6
|
* Add serialisation for sorting to documentmetastore.Henning Baldersheim2022-10-145-28/+78
|
* Use document::AnnotationType::TERM. Remove search::linguistics::TERM.Tor Egge2022-10-131-3/+3
|
* Merge pull request #24430 from ↵Tor Egge2022-10-1321-97/+106
|\ | | | | | | | | vespa-engine/toregge/rename-empty-doc-builder-to-doc-builder-and-move-it-to-search-test-namespace Rename search::index::EmptyDocBuilder to search::test::DocBuilder.
| * Rename search::index::EmptyDocBuilder to search::test::DocBuilder.Tor Egge2022-10-1321-97/+106
| | | | | | | | Rename search::index::StringFieldBuilder to search::test::StringFieldBuilder.
* | Merge pull request #24426 from vespa-engine/balder/cleanup-sorting-code-cppHenning Baldersheim2022-10-138-57/+53
|\ \ | |/ |/| Various cleanup while reading sorting code in backend.
| * Various cleanup while reading sorting code in backend.Henning Baldersheim2022-10-138-57/+53
| |
* | Merge pull request #24427 from ↵Geir Storli2022-10-139-12/+12
|\ \ | |/ |/| | | | | vespa-engine/geirst/generation-hold-list-terminology Align terminology on generation hold lists in attribute vector code
| * Rename removeAllOldGenerations() -> reclaim_unused_memory().Geir Storli2022-10-139-12/+12
| |
* | Add field value builders to EmptyDocBuilder.Tor Egge2022-10-132-15/+15
|/
* Rename removeOldGenerations() -> reclaim_memory().Geir Storli2022-10-122-5/+5
|
* Rename onGenerationChange() -> before_inc_generation() and change parameter ↵Geir Storli2022-10-122-4/+4
| | | | to be the current generation.
* Merge pull request #24407 from ↵Geir Storli2022-10-129-21/+22
|\ | | | | | | | | vespa-engine/geirst/generation-hold-list-terminology Use more general terminology for handling of generation hold lists
| * Rename trimHoldLists() -> reclaim_memory().Geir Storli2022-10-126-13/+13
| |
| * Rename transferHoldLists() -> assign_generation().Geir Storli2022-10-121-3/+3
| |
| * Rename first used generation -> oldest used generation.Geir Storli2022-10-123-5/+6
| |
* | Remove search::index::DocBuilder. Add search::index::StringFieldBuilder.Tor Egge2022-10-1214-90/+124
|/
* Stop using search::index::DocBuilder in DocumentFieldPopulator unitTor Egge2022-10-112-37/+36
| | | | test and document subdbs unit test.
* Implement GenerationHolder in terms of the generic generation hold list.Geir Storli2022-10-103-7/+7
|
* Simplify by using already computed maximun offset.Henning Baldersheim2022-10-031-1/+1
|
* To prevent against damages caused by excessive hits/offset param being sent ↵Henning Baldersheim2022-10-032-16/+38
| | | | down, cap it against numdocs.
* Now we always split phrases.Henning Baldersheim2022-09-304-4/+13
|
* Update tests.Henning Baldersheim2022-09-293-20/+7
|
* Normalize includes in searchcore and searchlib.Tor Egge2022-09-286-7/+7
|
* Normalize includes in searchcore.Tor Egge2022-09-2729-26/+32
|
* Expose more information in content node search reply trace.Geir Storli2022-09-261-0/+2
| | | | | This should make it easier to correlate the backend query blueprints with the actual query terms.