* added highly experimental startup option `--query.parallelize-gather-writes`
Turning this option on will enable the `parallelize-gather` AQL
optimizer rule for certain write queries. This option is `false`
by default.
The feature is highly experimental and should not be used on a
production system. The option is therefore also hidden.
* parallelize certain write operations in AQL
* remove unneeded _vocbase
* remove unneeded _vocbase
* added startup option and tests
* test
* rename shit
* rename shit
* Cancel connections to known failed servers
* do not uncomment code
* Retry on 503 responses
* Make test grey again
* when post cannot be delivered for readlock waiting for put meningless
* devel pull
* we can rely on POST alone
* clean up
* wrong errorCode tested
* fixed nullptr
* correct timeout
* fixed nullptr
* merge devel
* log level
* maintenace default log level raised to INFO, shard synchronisation only logs start/success all WARN/ERR demoted to DEBUG
* message unification
* devel merge
* fixed query registry
* tobias remarks
* mchacki part one
* static string for rebootid
* working tests
* use the reboot tracker from clusterinfo
* use the reboot tracker from clusterinfo
* rebootid be rebootid
* fixed serverid and all working
* static strings
* callguard out of reboottracker
* call callbacks, when query is ditched
* clear priv
* change log
* my oh my
* simon attention
* merge seems fine
* typo
* fix headers
* Ask the handler for direct execution instead of just looking at the RequestLane.
* Update arangod/RestHandler/RestDocumentHandler.cpp
* Do not access handler avter it was moved into lambda.
* Lambda should be mutable otherwise members are const and moving has no effect.
* Added more handlers to allowDirectExecution.
* Only allow direct execution for RestClusterHandler if asking for endpoints.
* Unified toVelocyPack of InputRow and AqlItemBlock. We can now velocypack any range within an AqlItemBlock as a new block.
* This is the fix required here. We now end exhausted false in initialize cursor. For some reason the other side checks this hard-coded value and if it is not the expected value it does ignore the send body ¯\_(ツ)_/¯
* Added test for range-based serialization
* late document materialization for view draft
* MaterializeType
* asserts and removed std::function
* std::map -> std::array
* Common search ast nodes code to a separate file
* Unit tests for create and clone a node
* js tests
* tmp unfinished
* Replace variables for late materialization
* Move removing
* Removed wrong clears
* Fixed variable replacement
* removed old tests (for separate materialization)
* check if late materialization arangosearch rule is enabled
* js test. added space
* moved a function to anonym namespace
* Subquery assert
* Review improvements
* Review improvements 2
* View node state to a separate var (microoptimization)
* Fixed index assert
* alignment
* C++ unit tests
* Fix bug in SubqueryEndExecutor
If a subquery was spliced, and did not get any input rows, the SubqueryEnd
executor still expected a relevant ShadowRow to be handed through, and ran into
an assertion (or otherwise crashed the server) if it didn't get one.
* Add regression test for SubqueryEndExecutor fix
Michael Hackstein
Jan
Simon
Jan Christoph Uhde
Kaveh Vahedipour
Lars Maier
Tobias Gödderz
Wilfried Goesgens
Andrey Abramov
jsteemann
Markus Pfeiffer
Iurii Popov
Max Neunhöffer
hkernbach