Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Merge pull request #9840 from vespa-engine/havardpe/remove-error-value | Arne H Juul | 2019-06-20 | 7 | -45/+19 |
|\ | | | | | Havardpe/remove error value | ||||
| * | readjust to a world without ErrorValue | Håvard Pettersen | 2019-06-19 | 7 | -45/+19 |
| | | |||||
* | | Track rename of cheap features to interleaved features. | Tor Egge | 2019-06-18 | 2 | -16/+16 |
| | | |||||
* | | Rename cheap features to interleaved features. | Tor Egge | 2019-06-18 | 2 | -3/+3 |
|/ | |||||
* | Explicit instantiation of BTreeIterator should always be accompanied with | Tor Egge | 2019-06-16 | 2 | -10/+18 |
| | | | | matching explicit instantiation of BTreeConstIterator and BTreeIteratorBase. | ||||
* | test tensor feature extraction for proton | Håvard Pettersen | 2019-06-14 | 2 | -13/+41 |
| | | | | | also make sure we avoid auto-unboxing for both indexed and streaming search | ||||
* | populate feature sets with serialized tensors | Håvard Pettersen | 2019-06-13 | 2 | -15/+29 |
| | |||||
* | Expose absolute values for document frequency in ITermFieldData (as needed ↵ | Geir Storli | 2019-06-12 | 2 | -6/+5 |
| | | | | by bm25 feature). | ||||
* | Add missing includes. | Tor Egge | 2019-06-11 | 2 | -1/+3 |
| | |||||
* | Extend query environment with method to get average field length. | Tor Egge | 2019-06-11 | 9 | -11/+33 |
| | |||||
* | Test propagation of field length info between indexes in index manager. | Geir Storli | 2019-06-07 | 2 | -28/+92 |
| | |||||
* | Rewrite index manager unit tests to gtest. | Geir Storli | 2019-06-07 | 2 | -291/+323 |
| | |||||
* | Save average field length in posting list file header. | Tor Egge | 2019-06-06 | 3 | -2/+3 |
| | |||||
* | Extend MemoryIndex to take an instance of IFieldLengthInspector to setup ↵ | Geir Storli | 2019-06-06 | 7 | -36/+44 |
| | | | | field length calculators per field index. | ||||
* | Change searchcorespi::IndexSearchable to also implement the ↵ | Geir Storli | 2019-06-05 | 5 | -14/+87 |
| | | | | | | IFieldLengthInspector API. Currently the memory and disk index implementations return empty field length info for all fields. | ||||
* | Rewrite indexcollection unit tests to gtest. | Geir Storli | 2019-06-05 | 2 | -76/+67 |
| | |||||
* | Wire field inverter at construction time. | Tor Egge | 2019-06-04 | 1 | -5/+3 |
| | |||||
* | Merge pull request #9633 from vespa-engine/havardpe/remove-tensor-factory | Arne H Juul | 2019-06-01 | 3 | -41/+37 |
|\ | | | | | remove tensor factory | ||||
| * | remove tensor factory | Håvard Pettersen | 2019-05-31 | 3 | -41/+37 |
| | | |||||
* | | Remove non-relevant comment. | Geir Storli | 2019-05-31 | 1 | -1/+0 |
| | | |||||
* | | Fix bug in lid space compaction where partial updates to attributes were ↵ | Geir Storli | 2019-05-31 | 13 | -118/+174 |
|/ | | | | | | | | | lost when moving a document. Instead of using the document store directly (when reading the document to move) the document retriever must be used. The document retriever patches in attribute values and also correctly waits for the thread writing to the document store. The job moving documents between the "ready" and "not ready" sub databases was already doing this correctly. | ||||
* | Move datastore and btree code from searchlib to vespalib | Tor Brede Vekterli | 2019-05-27 | 4 | -10/+10 |
| | | | | | | | | | | Namespace is still `search` and not `vespalib` due to the massive amount of code that would need to be modified for such a change. Other changes: - Move `BufferWriter` from searchlib to vespalib - Move assertion and rand48 utilities from staging_vespalib to vespalib - Move gtest utility code from staging_vespalib to vespalib | ||||
* | Merge pull request #9561 from ↵ | Henning Baldersheim | 2019-05-27 | 4 | -15/+15 |
|\ | | | | | | | | | vespa-engine/vekterli/move-address-space-to-vespalib Move `AddressSpace` from searchlib to vespalib | ||||
| * | Move `AddressSpace` from searchlib to vespalib | Tor Brede Vekterli | 2019-05-24 | 4 | -15/+15 |
| | | |||||
* | | Add unit test for match data tagging. | Tor Egge | 2019-05-27 | 2 | -1/+48 |
| | | |||||
* | | Refactor tagging of TermFieldMatchData regarding which features are | Tor Egge | 2019-05-24 | 3 | -10/+20 |
|/ | | | | needed by feature executors. | ||||
* | Merge pull request #9516 from vespa-engine/vekterli/move-rcuvector-to-vespalib | Tor Brede Vekterli | 2019-05-23 | 13 | -23/+23 |
|\ | | | | | Move RcuVector and relevant support classes to vespalib | ||||
| * | Move RcuVector and relevant support classes to vespalib | Tor Brede Vekterli | 2019-05-23 | 13 | -23/+23 |
| | | | | | | | | | | | | | | | | | | | | | | | | | | Having RCU support available across all our C++ modules open up new opportunities for optimizations. This changes the following: - `RcuVector` moved from `searchlib` to `vespalib` - `MemoryUsage` moved from `searchlib` to `vespalib` - Introduce a simplified, more generic `GrowStrategy` in `vespalib` used by the moved `RcuVector` which does not have any notion of documents. Existing searchlib `GrowStrategy` gets a simple function to convert to this generic strategy. | ||||
* | | Allow enum values to be used directly as mask values. | Tor Egge | 2019-05-23 | 4 | -17/+12 |
| | | |||||
* | | Refactor handle recorder. | Tor Egge | 2019-05-23 | 5 | -36/+44 |
| | | |||||
* | | Merge pull request #9472 from ↵ | Henning Baldersheim | 2019-05-22 | 9 | -32/+136 |
|\ \ | |/ |/| | | | | | vespa-engine/geirst/request-match-data-details-for-features Make it possible for rank features to request which match data that s.. | ||||
| * | Make it possible for rank features to request which match data that should ↵ | Geir Storli | 2019-05-21 | 9 | -32/+136 |
| | | | | | | | | be available in a TermFieldMatchData instance. | ||||
* | | Use std::regex instead of std::basic_regex<char>. | Tor Egge | 2019-05-21 | 1 | -1/+1 |
|/ | |||||
* | Merge pull request #9425 from vespa-engine/balder/multiple-threads-for-fusion | Henning Baldersheim | 2019-05-16 | 4 | -31/+38 |
|\ | | | | | - Provide an executor and use it for fusion. | ||||
| * | - Provide an executor and use it for fusion. | Henning Baldersheim | 2019-05-15 | 4 | -31/+38 |
| | | | | | | | | | | - Use per field temporary directories. - Keep WordNumMapping on the stack to make it thread safe. | ||||
* | | Make optimized path when rank drop limit is not used. | Geir Storli | 2019-05-16 | 2 | -11/+29 |
|/ | | | | This is done as comparing a double with a double that is NaN can be very expensive on some CPUs. | ||||
* | Expose the shared executor as a ThreadExecutor directly | Henning Baldersheim | 2019-05-15 | 4 | -13/+8 |
| | |||||
* | Use the shared executor from the threading service, and remove the loose wire. | Henning Baldersheim | 2019-05-14 | 8 | -67/+52 |
| | |||||
* | Wire in the shared executor in the threading service. | Henning Baldersheim | 2019-05-14 | 16 | -33/+72 |
| | |||||
* | introduce online state in proto rpc adapter | Håvard Pettersen | 2019-05-14 | 3 | -0/+8 |
| | | | | | this is to ensure we do not start server stuff until we are officially online. | ||||
* | Do not follow nullptr until node is ready for it. | Henning Baldersheim | 2019-05-13 | 2 | -2/+3 |
| | |||||
* | Merge branch 'master' into balder/use-2-transport-threads-for-proton-rpc | Henning Baldersheim | 2019-05-11 | 3 | -30/+26 |
|\ | |||||
| * | Simplify the supervisor responsibility | Henning Baldersheim | 2019-05-10 | 4 | -29/+31 |
| | | |||||
* | | Unify on emplace_back and code style. | Henning Baldersheim | 2019-05-08 | 1 | -51/+32 |
| | | |||||
* | | Use 2 thransport threads for proton rpc. | Henning Baldersheim | 2019-05-08 | 2 | -2/+8 |
|/ | |||||
* | Deinline test code. | Henning Baldersheim | 2019-05-06 | 6 | -12/+9 |
| | |||||
* | Deinline test code. | Henning Baldersheim | 2019-05-06 | 1 | -0/+17 |
| | |||||
* | Make final to improve inlining. | Henning Baldersheim | 2019-05-06 | 3 | -26/+19 |
| | |||||
* | Unsigned literal | Olli Virtanen | 2019-04-25 | 1 | -1/+1 |
| | |||||
* | Guard against attempting to create empty thread pool | Olli Virtanen | 2019-04-25 | 1 | -9/+11 |
| |