This project is mirrored from https://gitee.com/mirrors/neo4jsource.git.
Pull mirroring failed .
Repository mirroring has been paused due to too many failed attempts. It can be resumed by a project maintainer.
Repository mirroring has been paused due to too many failed attempts. It can be resumed by a project maintainer.
- 01 Aug, 2018 12 commits
-
-
Mattias Finné authored
-
Mattias Finné authored
so that such will be thrown as real exceptions, i.e. treated as tree inconsistencies, after consistent read.
-
Mattias Finné authored
Does this by introducing a test utility QueryResultComparingIndexReader so that this can easily be in other places too.
-
Mattias Finné authored
Keeps the one that has support for multiple threads adding sub-progressors because it supports all cases and the overhead of a concurrent queue is negligible in the grander scheme of things because it's only consulted once per sub-provider in a fusion scenario
-
MishaDemianenko authored
-
Satia Herfert authored
-
Satia Herfert authored
Fallback to interpreted
-
Satia Herfert authored
Slotted runtime cannot properly handle PatternExpressions or PatternComprehensions. In 3.4 we solve this by falling back to interpreted. In 3.5 the planner will plan these Expressions away, but the code to fallback should be kept in case we forgot a scenario where the planner does not plan such expressions away. The FallbackAcceptanceTest is useless in 3.5 though, so do not forward merge it.
-
Pontus Melke authored
-
Pontus Melke authored
Mostly for testing and benchmarking but in the case of bugs it's good to be able to disable compiled expressions without completely disabling the slotted runtime.
-
lutovich authored
Which was sent in response to a HELLO message. This field is currently not used by clients.
-
lutovich authored
Make metadata parsing errors result in a FAILURE message. Previously, they forced Bolt server to close the connection. Such errors can happen when bookmark format is invalid, transaction timeout or metadata can't be parsed or if of illegal type.
-
- 31 Jul, 2018 24 commits
-
-
Mattias Finné authored
-
Mattias Finné authored
-
Mattias Finné authored
Fixes #11888
-
Mattias Finné authored
Caching for temporary importer data structures, that is. This allows user to become aware of the fact that an performance of an import can degrade heavily when that happens. Some amount of information is also provided as to why of/off-heap allocators couldn't be used.
-
Mattias Finné authored
-
Mattias Finné authored
By printing a dash instead of a dot in the progress. This can help understand changes in timings between different parts of each import stage "group".
-
Mattias Finné authored
-
Mattias Finné authored
-
Mattias Finné authored
In BatchInserter the label index and created indexes are populated when calling shutdown(). Previously this was done single-threaded and with some home-grown code that was similar to what IndexingService would do and would also fail shutdown(), i.e. throw exception from it if there was an index entry conflict in a uniqueness index. This would leave the database, and indexes especially in an arbitrarily undefined state and would be confusing to user making the insert. Was the insert successful or not, should it be performed from scratch again? This has been changed to instead instantiate an IndexingService and let it populate the indexes in a multi-threaded fashion. Any failed indexes and would neither affect other indexes nor shutdown() itself. Such failed indexes will instead simply be marked as FAILED and user can decide what to do about it. Fixes #10738
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
Andrei Koval authored
-
- 30 Jul, 2018 4 commits
-
-
Przemek Hugh Kaznowski authored
-
Przemek Hugh Kaznowski authored
Performing a store copy against a machine that is too far behind will cause a checkpoint to be performed before the erroneous response is sent. This change sends the response first then asynchronously performs a checkpoint.
-
Craig Taverner authored
-
Craig Taverner authored
-