Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Do not read past the end. | Henning Baldersheim | 2019-10-14 | 1 | -1/+1 |
| | |||||
* | Recognize if-inverted decision trees in GBDT optimizer | Lester Solbakken | 2019-10-11 | 4 | -11/+30 |
| | |||||
* | Support missing values in expression evaluation in Java | Lester Solbakken | 2019-10-11 | 10 | -34/+103 |
| | |||||
* | improve fake searchable attributes | Håvard Pettersen | 2019-10-08 | 8 | -184/+236 |
| | | | | | | | | | | | | | blueprint/search iterator now exposes a more functional attribute search context fake attribute searches will now unpack a single entry containing the sum of all matched weights improve matcher test for same element matching by also using the attribute element iterator added tests for identifying matching elements in docsum request | ||||
* | enable use of fast forest gbdt evaluation for ranking | Håvard Pettersen | 2019-10-04 | 7 | -13/+110 |
| | |||||
* | Merge pull request #10875 from ↵ | Geir Storli | 2019-10-04 | 2 | -0/+6 |
|\ | | | | | | | | | vespa-engine/havardpe/find-matching-elements-for-stand-alone-attribute-terms find matching elements for attribute terms searching in struct subfields | ||||
| * | find matching elements for attribute terms searching in struct subfields | Håvard Pettersen | 2019-10-04 | 2 | -0/+6 |
| | | |||||
* | | Don't ignore truncate result. | Tor Egge | 2019-10-03 | 1 | -1/+2 |
|/ | |||||
* | collect matching elements by using SameElement blueprints | Håvard Pettersen | 2019-10-02 | 4 | -10/+22 |
| | |||||
* | Merge pull request #10840 from vespa-engine/havardpe/faster-boosted-models | Geir Storli | 2019-10-02 | 1 | -2/+15 |
|\ | | | | | faster gbdt forest evaluation | ||||
| * | faster gbdt forest evaluation | Håvard Pettersen | 2019-10-02 | 1 | -2/+15 |
| | | | | | | | | | | This is a draft implementation of gbdt forest evaluation doing feature-at-a-time rather than tree-at-a-time. | ||||
* | | Remove some more transport related stuff. | Henning Baldersheim | 2019-09-30 | 5 | -53/+5 |
|/ | |||||
* | Merge pull request #10804 from vespa-engine/balder/remove-leftover-fs4-code | Henning Baldersheim | 2019-09-30 | 32 | -5321/+29 |
|\ | | | | | Balder/remove leftover fs4 code | ||||
| * | Remove unused error code | Henning Baldersheim | 2019-09-30 | 1 | -16/+0 |
| | | |||||
| * | GC some unused code related to fs4 protocol that has now gone missing. | Henning Baldersheim | 2019-09-27 | 27 | -4266/+30 |
| | | |||||
| * | Remove TransportServer | Henning Baldersheim | 2019-09-27 | 7 | -1040/+0 |
| | | |||||
* | | Merge pull request #10811 from ↵ | Henning Baldersheim | 2019-09-30 | 2 | -2/+4 |
|\ \ | | | | | | | | | | | | | vespa-engine/toregge/port-logdatastore-test-to-darwin Port logdatastore unit test to Darwin. | ||||
| * | | Port logdatastore unit test to Darwin. | Tor Egge | 2019-09-29 | 2 | -2/+4 |
| |/ | |||||
* / | Fix same element unit test for clang. | Tor Egge | 2019-09-29 | 1 | -0/+1 |
|/ | |||||
* | Merge pull request #10787 from ↵ | Tor Egge | 2019-09-25 | 2 | -31/+0 |
|\ | | | | | | | | | vespa-engine/toregge/remove-dead-posting-change-method Remove dead method search::PostingChange::apply | ||||
| * | Remove dead method search::PostingChange::apply. | Tor Egge | 2019-09-24 | 2 | -31/+0 |
| | | |||||
* | | Merge pull request #10770 from ↵ | Tor Egge | 2019-09-25 | 7 | -3/+76 |
|\ \ | | | | | | | | | | | | | vespa-engine/toregge/detect-enum-store-entry-ref-count-overflow Detect enum store entry reference count overflow | ||||
| * | | Fix indentation. | Tor Egge | 2019-09-25 | 1 | -1/+1 |
| | | | |||||
| * | | Use a more descriptive test name. | Tor Egge | 2019-09-24 | 1 | -3/+3 |
| | | | |||||
| * | | Adjust reference count for default value enum store entry when shrinking a | Tor Egge | 2019-09-24 | 2 | -0/+52 |
| | | | | | | | | | | | | single value enumerated attribute. | ||||
| * | | Verify that corresponding posting list is empty before removing | Tor Egge | 2019-09-23 | 2 | -3/+21 |
| | | | | | | | | | | | | enum store entry. | ||||
| * | | Detect enum store entry reference count overflow. | Tor Egge | 2019-09-23 | 3 | -0/+3 |
| | | | |||||
* | | | Nonfunctional changes only | Jon Bratseth | 2019-09-24 | 2 | -35/+19 |
| | | | |||||
* | | | Handle NaN in hit aggregation sorting | Jon Bratseth | 2019-09-24 | 1 | -1/+1 |
| | | | |||||
* | | | Handle Nan in group merging | Jon Bratseth | 2019-09-24 | 1 | -7/+1 |
| | | | |||||
* | | | Handle comparison of NaNs | Jon Bratseth | 2019-09-24 | 1 | -1/+1 |
| | | | |||||
* | | | Nonfunctional changes only | Jon Bratseth | 2019-09-24 | 6 | -46/+39 |
| |/ |/| | |||||
* | | added low-level code used to identify matching elements | Håvard Pettersen | 2019-09-23 | 3 | -0/+45 |
| | | |||||
* | | introduce separate docsum matcher class | Håvard Pettersen | 2019-09-23 | 2 | -5/+4 |
|/ | |||||
* | added function identifying which elements matched | Håvard Pettersen | 2019-09-19 | 10 | -0/+238 |
| | | | | | only a skeleton for now; outlines the interface between the summary generator and the matcher. | ||||
* | Merge pull request #10646 from vespa-engine/geirst/enum-store-cleanup-5 | Henning Baldersheim | 2019-09-13 | 31 | -286/+250 |
|\ | | | | | Geirst/enum store cleanup 5. | ||||
| * | Style changes and function renaming in enum store dictionary. | Geir Storli | 2019-09-13 | 7 | -44/+33 |
| | | |||||
| * | Style changes and function renaming in enum store. | Geir Storli | 2019-09-13 | 28 | -169/+166 |
| | | |||||
| * | Remove unused functions. | Geir Storli | 2019-09-13 | 3 | -21/+0 |
| | | |||||
| * | Rename functions used to free unused values in enum store. | Geir Storli | 2019-09-13 | 10 | -47/+47 |
| | | |||||
| * | Rename fixupRefCount(s) -> set_ref_count(s). | Geir Storli | 2019-09-13 | 6 | -14/+13 |
| | | |||||
* | | Don't remove parentheses around comparisons | Lester Solbakken | 2019-09-13 | 1 | -0/+2 |
|/ | |||||
* | Merge pull request #10620 from ↵ | Arne H Juul | 2019-09-12 | 14 | -15/+92 |
|\ | | | | | | | | | vespa-engine/havardpe/unpacking-iterators-optimizer Havardpe/unpacking iterators optimizer | ||||
| * | add expensive flags to phrase/same element | Håvard Pettersen | 2019-09-12 | 9 | -15/+33 |
| | | | | | | | | they will affect the iterator cost tier | ||||
| * | added rank config for split/delay flags | Håvard Pettersen | 2019-09-12 | 5 | -0/+59 |
| | | | | | | | | used to control how unpacking iterators are optimized and sorted | ||||
* | | Add unit tests for enum store loading and fix bug when loading from ↵ | Geir Storli | 2019-09-11 | 12 | -52/+179 |
| | | | | | | | | | | | | | | non-enumerated save files. In this case the posting list reference was not written to the dictionary, the result being empty posting lists for all unique values. Loading from non-enumerated save files is only used when changing a numeric attribute to use fast-search. | ||||
* | | Write unit tests for enum store batch updater. | Geir Storli | 2019-09-11 | 3 | -2/+80 |
| | | |||||
* | | Rewrite enum store tests to gtest. | Geir Storli | 2019-09-11 | 4 | -240/+190 |
|/ | |||||
* | Remove unused functions. | Geir Storli | 2019-09-10 | 2 | -2/+0 |
| | |||||
* | Simplify how we iterate the set of unique values when saving an enumerated ↵ | Geir Storli | 2019-09-10 | 8 | -44/+13 |
| | | | | attribute. |