- (improvement) Avoid redundant BigDecimal cast (#306)
- (bugfix) Fix NPE on clearParameters (#315)
- (bugfix) Support LocalDateTime (#316)
- (bugfix) Fix signed error when type is nullable (#305) Thanks @liuneng1994
- (bugfix) Fix ResultSetMetaData.isSigned (#281 #283)
- (bugfix) Fix LoggerFactory#getLogger (#287)
- (enhance) Set ClickHouseSQLException errorCode correctly (#290) Thanks @baibaichen
- (feature) Rework stuffs related to time api (#288) Thanks @sbouchex
- (bugfix) Fix Negative Decimal serde (#279)
- The project static website is available at Home Page
- The China mirror repo is available at Gitee
- Implemented
BalancedClickhouseDataSource
to support connect to clickhouse cluster - Now the JDBC driver works with DataGrip and DBeaver
- Introduced a log facade which prefer to bind
slf4j
if it exists in classpath, otherwise fallback to JDK Logger
, and added logs for tracing, debugging and warning
This release contains lots of internal refactors, and currently, this driver guarantee none stable API except JDBC API itself, please be careful if you use the internal API which not belongs to JDBC standard interfaces.
- (feature) Implement
BalancedClickhouseDataSource
to support multi instances (#205) - (bugfix) Fix invalid package clickhouse-integration-spark_2.12 (#206)
- (feature) Implement
ClickHouseConnection
catalog and schema method (#208) - (enhance) Enhance
Number
process - (build) Disable deploy
examples
module - (test) Integration tests with connection pool (#212)
- (feature) Support
charset
parameter (#213) - (feature) Implement
ClickHousePreparedStatment
#setBytes
and #getBytes
(#213) - (enhance) Trim space before parsing SQL
- (feature) Support
tcp_keep_alive
parameter - (bugfix) Fix DateTime64 nanos and scale (#217)
- (bugfix) Reset connection
STATE
when recreate connection (#221) - (bugfix) Skip remaining response when ping (#224)
- (refactor) Decouple
RequestOrResponse
and rename PhysicalConnection
to NativeClient
(#225) - (feature) Introduce log classes to avoid hard dependence on
slf4j
(#231) - (build) Add
debug
Maven profile (#235) - (enhance) Add LRU cache for lexer
IDataType
(#236) - (feature) Refactor and implement ClickHouseDatabaseMetadata (#233)
- (feature) Support type alias (#240)
- (bugfix) Explicit upcast ByteBuffer to Buffer (#243)
- (deps) Bump ini from 1.3.5 to 1.3.8 (#250)
- (feature) Support Decimal256 && Decimal128 (#239)
- (refactor) Refactor and fixup (#246)
- (deps) bump alibaba druid 1.2.4
- (deps) bump slf4j 1.7.30
- (deps) bump mockito 3.6.28
- (deps) bump jmh 1.27
- (bugfix) Cast correctly numbers to get a boolean (#256)
- (bugfix) Added check against null data on
ClickHouseResultSet#getXXX
(#258) (#259) - (feature) Handle Nothing datatype (#261)
- (bugfix)
ClickHouseConnection#connect
should return null for unacceptable url (#267) - (enhance)
BalancedClickhouseDataSource
support optional database url - (test) Added user & password for ClickHouse unit tests (#264)
- (feature) Support arbitrary settings (#268)
- (refactor) Rewrite
StringView
implements CharSequence
(#270) - (workflow) Auto report benchmark report.
- (backport) cast correctly numbers to get a boolean (#256)
- (backport) enhance null check on ClickHouseResultSet#getXXX (#259)
- (backport) valid connection by ping instead of SELECT 1
- (backport) explicit upcast ByteBuffer to Buffer (#243)
- (backport) fix Connection#getCatalog (#249)
- (backport) fix ClickHouseStatement#getUpdateCount
- (backport) fix ClickHouseStatement#setMaxRows
- (backport) ClickHouseResultSet#getString support all types
- (backport) implement ClickHouseResultSet#getBoolean
- (backport) implement ClickHouseResultSet#isBeforeFirst, #isAfterLast, #isFirst etc.
- ClickHouseConnection#getMetaData return null instead of throw exception
- Trim space before parsing SQL
- Fix DateTime64 nanos and scale (#217)
- Reset connection STATE when recreate connection (#221)
- Skip remaining responses when consume PingResponse (#224)
- Minor tunes
- Fix invalid package clickhouse-integration-spark_2.12 (#206)
- Since this release v2.4.0, we switch to semantic versioning, and the bug fix version would be available quickly.
- We introduce a new module
clickhouse-integration-spark
base on Spark Jdbc DataSource API for integrating with Spark, check detail at README - In the past few weeks, we were more focus on code quality, fixed all LGTM alerts and got A+ score.
- Implement
clickhouse-integration-spark
base on Spark Jdbc DataSource API (#170 #184) - Batch insert support nullable types in arrays and nested types (#144 #194)
- Support Boolean (#196)
- Fix timezone issue (#195)
- Fix potential concurrence issue (#191)
- Code refactor, fixed all LGTM alerts and got A+ score.
- Refactor
ClickHouseStatement
inheritance tree (#201) - Migrate to mockito3 (#175)
- Migrate to org.lz4:lz4-java (#174)
- Migrate to Junit5 (#171)
- Bump yandex/clickhouse-jdbc 0.2.4
- Bump jmh 1.26
- Enable checkstyle (#172)
- Enable code coverage (#190)
- Improve Docs and Readme
- Add missing license headers
- Provide shaded version
clickhouse-native-jdbc-shaded
since this release, see detail at README.md
- Fixed
SQLFeatureNotSupportedException
in several scenarios (#142 thanks @dcastanier) (thanks @sundy-li) - Fixed return value for executeBatch in PreparedInsertStatement (#145 thanks @tauitdnmd)
- Fixed jdbc url parse (#148 thanks @tauitdnmd)
- Support zoned datetime DataType, i.e.
DateTime(Asia/Shanghai)
(#158 thanks @sundy-li) - Refactor project to multi modules (#156 thanks @pan3793)
- Remove
joda-time
, migrate joda-time
and legacy java.util.Date
to java.time
(#164 thanks @pan3793) - Integration Test against Java 8, Java 11 (thanks @sundy-li)
- Integration Test with Spark 2.4.7&Scala 2.11 (thanks @sundy-li)
- Bump junit from 4.11 to 4.13.1
::: tip
You can find early releases at GitHub Release Page
:::