summaryrefslogtreecommitdiffstats
path: root/searchlib
Commit message (Collapse)AuthorAgeFilesLines
* Add a reorder fence.Henning Baldersheim2018-07-201-0/+1
|
* Merge pull request #6439 from vespa-engine/balder/lift-rankitemHenning Baldersheim2018-07-204-65/+47
|\ | | | | Balder/lift rankitem
| * Use override keywordHenning Baldersheim2018-07-202-3/+3
| |
| * Use = default and C++11 for loops where possible. No semantic changes.Henning Baldersheim2018-07-204-62/+44
| |
* | Merge pull request #6437 from ↵Henning Baldersheim2018-07-201-2/+5
|\ \ | | | | | | | | | | | | vespa-engine/balder/ensure-that-there-is-at-least-1-valid-stopsign-visible-at-all-times There must always be a stopsign visible to any concurrent reader no m…
| * | There must always be a stopsign visible to any concurrent reader no matter ↵Henning Baldersheim2018-07-201-2/+5
| | | | | | | | | | | | | | | | | | | | | | | | | | | where it is. So before removing the old stopsign, the new one must be placed. You can only remove the old one if it was ahead of new one. If not, any readers scanning might have passed the location of the new one, and the old one might be gone before it gets there. And hence it will continue into space or a concrete wall.
* | | Merge pull request #6436 from ↵Henning Baldersheim2018-07-207-138/+215
|\ \ \ | |_|/ |/| | | | | | | | vespa-engine/balder/choose-invalidate-or-update-of-cache Balder/choose invalidate or update of cache
| * | Properly test writethrough on UPDATE strategyHenning Baldersheim2018-07-191-14/+37
| | |
| * | Properly test both INVALIDATE and UPDATE strategy. Fix bug with incorrect ↵Henning Baldersheim2018-07-192-25/+73
| | | | | | | | | | | | size calculations on updating existing elements.
| * | Add control of cache update policy when an item changes value.Henning Baldersheim2018-07-197-109/+115
| |/
* / Use the guava version from container-dep-versions everywhere.gjoranv2018-07-191-1/+0
|/
* Compute bloat after when it is more accurate.Henning Baldersheim2018-07-181-3/+5
|
* test returned value alsoArne Juul2018-07-163-0/+5
|
* use new APIArne Juul2018-07-131-15/+7
| | | | | rewrite weighted set attribute filter to handle all enums that matches the token when folded.
* test new APIArne Juul2018-07-132-0/+5
|
* implement new API in multistringattributeArne Juul2018-07-131-0/+5
|
* implement new API in singlestringattributeArne Juul2018-07-131-0/+3
|
* not-implement new APIArne Juul2018-07-132-0/+8
|
* forward new APIArne Juul2018-07-132-0/+7
|
* require new APIArne Juul2018-07-131-0/+1
|
* implement new API for float and integerArne Juul2018-07-134-0/+28
|
* implement new API in attributevectorArne Juul2018-07-132-0/+8
|
* implement new API in attrvectorArne Juul2018-07-132-0/+14
|
* test new APIArne Juul2018-07-131-0/+37
|
* add findFoldedEnums APIArne Juul2018-07-134-0/+30
|
* add unit test for appendPosition()Arne Juul2018-07-091-0/+44
| | | | * previously this would trigger a bug and crash.
* refactor to avoid code duplicationArne Juul2018-07-062-25/+19
| | | | * also minimize diff from original version
* allocate vector explicitlyArne Juul2018-07-062-19/+43
| | | | | | | | * when appending a position, the old version would call appendToAllocatedVector() even in cases where no vector had already been allocated. That routine then accessed a union to get old size, accessing some random data that would mostly be 0 by accident.
* Fix author tag for SimonBjørn Christian Seime2018-07-05115-115/+115
|
* Merge pull request #6287 from ↵Harald Musum2018-06-271-1/+1
|\ | | | | | | | | vespa-engine/bratseth/dont-fail-allocation-on-bootstrap Bratseth/dont fail allocation on bootstrap
| * Nonfunctional changes onlyJon Bratseth2018-06-261-1/+1
| |
* | Merge pull request #6277 from ↵Geir Storli2018-06-268-88/+154
|\ \ | | | | | | | | | | | | vespa-engine/geirst/ensure-fast-search-struct-field-attributes-work-with-same-element-operator Ensure that all attribute iterators expose the search context creatin…
| * | Ensure that all attribute iterators expose the search context creating the ↵Geir Storli2018-06-258-88/+154
| |/ | | | | | | | | | | iterator. This ensures that also fast-search struct field attributes work together with the sameElement operator.
* / stop using DoubleFormatterArne Juul2018-06-251-1/+1
|/ | | | | * also, update expected results in those places where slight differences in Double formatting was seen.
* some more cleanupArne Juul2018-06-2013-34/+21
|
* move LOG_SETUP after other includesArne Juul2018-06-2066-160/+182
|
* add and use HDR_ABORT insteadArne Juul2018-06-205-9/+14
|
* Merge branch 'master' into arnej/use-log-abortArne Juul2018-06-2024-72/+420
|\
| * Add test for getNamePrefixHenning Baldersheim2018-06-181-2/+19
| |
| * - Remove the execute(string, ...) and force the use of ExecutorId.Henning Baldersheim2018-06-172-10/+9
| | | | | | | | | | - Remove some double bookkeeping in AttributeWriter. - Ensure that we always use attribute.getNamePrefix() to compute executor id.
| * Merge pull request #6190 from vespa-engine/lesters/random-normal-matchGeir Storli2018-06-159-46/+275
| |\ | | | | | | Add randomNormal.match
| | * Use correct seed and don't use spare for randomNormalStableLester Solbakken2018-06-153-13/+16
| | |
| | * Rename feature randomNormalMatch -> randomNormalStableLester Solbakken2018-06-156-37/+37
| | |
| | * Move match version of random normal to it's own featureLester Solbakken2018-06-149-57/+217
| | |
| | * Remove unnecessary logging in unit testLester Solbakken2018-06-141-2/+0
| | |
| | * Fix random normal match testLester Solbakken2018-06-142-8/+31
| | |
| | * Refactor out normal distributed random generatorLester Solbakken2018-06-143-48/+65
| | |
| | * Pass random generator by referenceLester Solbakken2018-06-141-1/+1
| | |
| | * Add match to output in randomNormal testLester Solbakken2018-06-131-1/+1
| | |
| | * Add randomNormal.matchLester Solbakken2018-06-133-13/+41
| | |