aboutsummaryrefslogtreecommitdiffstats
path: root/searchlib
Commit message (Collapse)AuthorAgeFilesLines
* Compact lid space on source selector.Tor Egge2019-01-034-6/+28
|
* clion recommended cleanup while browsing test.Henning Baldersheim2018-12-192-89/+57
|
* Use a bitvector for the single bit numeric attribute.Henning Baldersheim2018-12-1912-76/+392
|
* Also test inverted bitvector iterators.Henning Baldersheim2018-12-191-60/+68
|
* Follow api changeHenning Baldersheim2018-12-192-9/+3
|
* Remove test that will soon be invalid.Henning Baldersheim2018-12-191-1/+0
|
* Pass invert as a parameterHenning Baldersheim2018-12-195-88/+76
|
* NULL -> nullptrHenning Baldersheim2018-12-191-40/+40
|
* Add test for inverted bitvector iterator.Henning Baldersheim2018-12-182-11/+28
|
* Add inverted BitvectorIteratorHenning Baldersheim2018-12-185-83/+148
|
* std::make_uniqueHenning Baldersheim2018-12-181-29/+19
|
* Return information if a buffer is held.Henning Baldersheim2018-12-187-36/+33
| | | | | Use std::make_unique. Minor unifying cleanup.
* std::make_uniqueHenning Baldersheim2018-12-181-55/+49
|
* UINT1 -> BOOLHenning Baldersheim2018-12-1613-25/+21
|
* UINT1 -> BOOLHenning Baldersheim2018-12-161-1/+1
|
* Merge pull request #7896 from ↵Henning Baldersheim2018-12-064-18/+81
|\ | | | | | | | | vespa-engine/geirst/fix-bug-in-datastore-memory-allocation Fix bug where allocation of too large buffers in data store caused of…
| * Fix bug where allocation of too large buffers in data store caused offset in ↵Geir Storli2018-12-064-18/+81
| | | | | | | | | | | | | | EntryRefT to be out of bounds. This happened in cases where wanted number of bytes was not a power of 2 and less than huge page size. After rounding up to nearest power of 2 (to match the underlying memory allocator) the allocated buffer became too large for the offset in EntryRefT to handle.
* | Merge pull request #7889 from ↵Henning Baldersheim2018-12-066-37/+45
|\ \ | | | | | | | | | | | | vespa-engine/balder/test-and-fix-postinglist-fallback-iterator Balder/test and fix postinglist fallback iterator
| * | Return correct type.Henning Baldersheim2018-12-061-1/+1
| | |
| * | Add test for range search and also fetch postings with correct strictness.Henning Baldersheim2018-12-061-25/+35
| | |
| * | Allow repeated fetchPostings callsHenning Baldersheim2018-12-061-4/+4
| | |
| * | = defaultHenning Baldersheim2018-12-063-7/+5
| |/
* / Manage ph-javacc-maven-plugin config in parent pom.gjoranv2018-12-061-9/+0
|/ | | | + Remove unused management for old javacc plugin.
* Revert "Add assert to check sanity of input."Henning Baldersheim2018-12-051-6/+1
|
* Add assert to check sanity of input.Henning Baldersheim2018-12-051-1/+6
|
* Flip the bits when hashing to get good hashing.Henning Baldersheim2018-12-041-1/+1
|
* Revert "Swap offset and bufferid for better and more compact hashing."Henning Baldersheim2018-12-042-9/+5
|
* Merge pull request #7817 from ↵Jon Bratseth2018-11-291-1/+1
|\ | | | | | | | | vespa-engine/toregge/allow-zero-arg-parenthesis-fixup Don't allow function argument list to start with comma.
| * Don't allow function argument list to start with comma.Tor Egge2018-11-291-1/+1
| |
* | Merge pull request #7804 from vespa-engine/balder/randomize-keysHenning Baldersheim2018-11-293-3/+6
|\ \ | |/ |/| Balder/randomize keys
| * Follow template instantiation.Henning Baldersheim2018-11-281-1/+3
| |
| * Use a simple and instead of div for modulo operation.Henning Baldersheim2018-11-281-1/+2
| |
| * Reserve a large hashmap to reduce number of collisions.Henning Baldersheim2018-11-281-1/+1
| |
* | Merge pull request #7805 from vespa-engine/bratseth/allow-zero-arg-parenthesisLester Solbakken2018-11-291-1/+1
|\ \ | |/ |/| Allow no-arg function invocations to use parentheses
| * Allow no-arg function invocations to use parenthesesJon Bratseth2018-11-281-1/+1
| |
* | Filter higher bits if not all are given.Henning Baldersheim2018-11-281-1/+5
| |
* | Bits are flipped.Henning Baldersheim2018-11-281-1/+1
| |
* | Swap offset and bufferid for better and more compact hashing.Henning Baldersheim2018-11-281-4/+4
| |
* | Reduce locked section by using a hashmap to reduce cachemisses during ↵Henning Baldersheim2018-11-2811-76/+56
| | | | | | | | | | | | | | | | | | | | repopulation. Create the old2New mapping during compaction No more precompute phase. Conflicts: searchlib/src/vespa/searchlib/attribute/multienumattribute.hpp
* | Add hash method.Henning Baldersheim2018-11-281-0/+1
| |
* | Avoid implicit conversion bug.Henning Baldersheim2018-11-281-1/+1
| |
* | Use explicit EntryRef constructor to avoid unintended conversionHenning Baldersheim2018-11-2857-442/+317
| | | | | | | | | | Conflicts: searchlib/src/vespa/searchlib/attribute/multienumattribute.hpp
* | =default while reading code.Henning Baldersheim2018-11-287-45/+19
| |
* | = deafult and NULL -> nullptrHenning Baldersheim2018-11-283-17/+11
|/
* Use provided scope to avoid class cast issuesHenning Baldersheim2018-11-241-0/+1
|
* Move all importing to model-integrationJon Bratseth2018-11-2229-2713/+6
|
* Restore old package nameJon Bratseth2018-11-2228-67/+65
|
* Separate model integration into a separate moduleJon Bratseth2018-11-2155-66015/+35
| | | | | | | This allows us to access model importers (such as TensorFlow) in config models without loading one instance per config model instance, which is not possible with TensorFlow because it depends on JNI code.
* Thaw dictionary in unique store when changing reference during compaction.Tor Egge2018-11-191-1/+1
|
* Use TF 1.11.0 everywhereJon Bratseth2018-11-151-2/+0
|