birdstorm
birdstorm
Cap lock expire time to 1000 ms to prevent eternal sleep.
In #2089 resolve lock may cost an extra scan when no locks is contained in scan resp. Optimize the process by using a new kvproto API.
The default concurrency of table scan is 512. It seems to be irrational for each executor to use this setting. We should review scan concurrency settings. And enclose a test...
**Describe the bug** TiSession was cached by distinct value of SparkConf. However, it is possible that multiple SparkSession share the same SparkConf, thus the SparkSession that TiSession points to is...
**Issue by [ilovesoup](https://github.com/ilovesoup)** _Mon Nov 20 14:45:45 2017_ _Originally opened as https://github.com/pingcap/tikv-client-lib-java/issues/148_ ---- We need one otherwise will be too hard to test.
**Issue by [ilovesoup](https://github.com/ilovesoup)** _Tue Nov 21 17:52:20 2017_ _Originally opened as https://github.com/pingcap/tikv-client-lib-java/issues/151_ ---- Some users request such examples.
The previous abstract of calling scan/get from RegionStoreClient is confusing for new resolve lock logic. We should implement an independent Sender interface to send those requests to make logic clear.
`select * from t where c > 10 or c < 100` -> should be range (-inf, inf)
We should cache statistics information to reduce recalculation during self-join, since they must have the same timestamp.
**Issue by [birdstorm](https://github.com/birdstorm)** _Tue Dec 5 07:12:38 2017_ _Originally opened as https://github.com/pingcap/tikv-client-lib-java/issues/180_ ---- Should add an option to use index as priority.