summaryrefslogtreecommitdiffstats
path: root/document
Commit message (Collapse)AuthorAgeFilesLines
...
* | minor cleanupArne H Juul2022-03-291-24/+1
|/ | | | * simplify and cleanup some outdated code and comments
* always use GeoPosType for position structsArne H Juul2022-03-281-3/+4
| | | | * should not cause any changes in observed behavior
* Handle unaligned IdString reads in a well-defined mannerTor Brede Vekterli2022-03-281-2/+16
| | | | | | Should compile down to effectively the same instructions on x86-64, and might save us some headaches on architectures that are less lenient with data alignment for reads.
* Avoid code duplication.Henning Baldersheim2022-03-281-2/+1
|
* Use both lvalue and rvalue specifier to avoid explicit std::move()Henning Baldersheim2022-03-285-86/+94
|
* Avoid the need for clone by using unique_ptr.Henning Baldersheim2022-03-2821-169/+141
|
* GC unused codeHenning Baldersheim2022-03-281-1/+0
|
* Avoid need to copy/clone FieldUpdateHenning Baldersheim2022-03-277-114/+112
|
* Avoid multiple inheritance.Henning Baldersheim2022-03-2710-28/+45
|
* fix two getDataType callsArne H Juul2022-03-241-2/+2
|
* mark getDataType(String) as deprecatedArne H Juul2022-03-244-13/+30
| | | | | * internal code can use getDataTypeInternal when necessary * external code should always get their data types in other ways
* avoid DocumentTypeManager.getDataType(name)Arne H Juul2022-03-246-28/+36
|
* Merge pull request #21804 from vespa-engine/arnej/split-get-declare-structArne H Juul2022-03-243-2/+41
|\ | | | | split into two methods (review feedback)
| * split into two methods (review feedback)Arne H Juul2022-03-243-2/+41
| |
* | expand tabs and reindentArne H Juul2022-03-2412-385/+385
|/
* just use data type known by the annotation typeArne H Juul2022-03-231-2/+3
|
* Abort when something unexpected happens.Henning Baldersheim2022-03-221-1/+1
|
* GC unused hasChanged concept. Only done on StructFieldValue as we never ↵Henning Baldersheim2022-03-1829-245/+34
| | | | reference internal data.
* Implement className on document::FieldValueHenning Baldersheim2022-03-1714-26/+65
|
* Add explicit and GC unused getAddZeroTermHenning Baldersheim2022-03-175-19/+13
|
* Add noexceptHenning Baldersheim2022-03-171-10/+10
|
* Avoid requiring identifiable for FieldValueHenning Baldersheim2022-03-1674-354/+271
|
* Add basic support for BoolFieldValues in C++ document selection implTor Brede Vekterli2022-03-153-11/+36
|
* Add comparison support for BoolFieldValues in Java document selection implTor Brede Vekterli2022-03-152-0/+20
| | | | | | Field _value_ itself is already extracted via field path iteration, but since bool fields do not inherit from an existing numeric field value type we need to explicitly handle the bool type when evaluating AST comparison nodes.
* GC unused convenience methods.Henning Baldersheim2022-03-153-14/+2
|
* add getDeclaredStructType in DocumentTypeArne H Juul2022-03-135-3/+170
| | | | | * getType() in DocumentTypeManager will be deprecated soon, this will be the recommended replacement.
* even TemporarySDField should be connected to a "repo"Arne H Juul2022-03-111-0/+1
|
* Move BucketIdListResultHenning Baldersheim2022-03-093-8/+7
|
* surely the TAG type is a tag typeArne H Juul2022-03-091-1/+3
|
* Merge pull request #21598 from vespa-engine/toregge/avoid-ambiguous-overloadHenning Baldersheim2022-03-081-1/+1
|\ | | | | Avoid ambiguous overload.
| * Avoid ambiguous overload.Tor Egge2022-03-081-1/+1
| |
* | Merge pull request #21586 from vespa-engine/balder/gc-unused-threadpoolHenning Baldersheim2022-03-081-1/+0
|\ \ | | | | | | Balder/gc unused threadpool
| * | Get VESPA_DLL_LOCAL from vespa/fastos/types.hHenning Baldersheim2022-03-071-1/+0
| |/
* / deprecate useless constructorArne H Juul2022-03-082-10/+19
|/ | | | | | * always ignored non-null typeName * the intention was probably something else * was not used anywhere
* Merge pull request #21580 from vespa-engine/arnej/verify-namecollision-failsJon Bratseth2022-03-071-3/+8
|\ | | | | struct/document name collision fails
| * struct/document name collision failsArne H Juul2022-03-071-3/+8
| | | | | | | | | | * the unit tests claims that this works, and seems to verify that. But the generated config isn't valid.
* | Use corrcet type for short.Henning Baldersheim2022-03-071-1/+1
| |
* | Use XXFieldValue::make over std::make_uniqueHenning Baldersheim2022-03-0715-55/+54
| |
* | No reuse, so just use return value optimisation.Henning Baldersheim2022-03-072-12/+9
| |
* | GC unused non-const iterate interface.Henning Baldersheim2022-03-072-15/+7
| |
* | GC redundant convenience wasy of assigning primitive field values.Henning Baldersheim2022-03-0723-303/+78
|/
* As a first step to reduce code complexity around FieldValue move test-only ↵Henning Baldersheim2022-03-059-124/+115
| | | | | | convenience methods out to test-only helpers.
* Remove clone from DataType.Henning Baldersheim2022-03-0427-120/+61
|
* Remove identifiable from document::DataTypeHenning Baldersheim2022-03-0447-421/+294
| | | | Also remove the virtual == operator and the need for default constructor.
* Reduce use of Identifiable for document::DatatTypeHenning Baldersheim2022-03-0343-129/+162
|
* GC cloneableHenning Baldersheim2022-03-036-17/+10
|
* Explicitly precompute GlobalId for reference values to avoid lazy init raceTor Brede Vekterli2022-03-032-1/+4
| | | | | | | | | | | | | This is a workaround for `DocumentId::getGlobalId()`'s lazy GID computation currently not being thread safe in a `const` context. The following race was previously possible: * Attribute writer thread calls `DocumentId::getGlobalId` when applying the reference value update to a reference attribute. Writes GID and flag. * Shared executor thread calls `DocumentId::operator=` with the same ID instance as the rhs argument when applying the reference value update to the doc store. Causes a read race with the above write.
* Expect updated error messageJon Bratseth2022-03-011-2/+1
|
* No functional changesJon Bratseth2022-03-013-23/+16
|
* Add noexcept specifiers.Tor Egge2022-02-282-3/+3
|