0.2.0

Bugfixes and minor API improvements.

Optimizations and refactors:

• Use gRPC tools to generate C# code from proto files #233
• Scan builder API adjustments #234
  • Make ExclusiveUpperBoundRaw private
  • Add SetFullProjection method
  • Fix issue where setting projected column indexes wouldn't apply if projected names were already set
• Small memory optimization #226

Bug fixes:

• Fix a scan bug which will read repetitive rows #229
• Update lastPropagatedTimestamp in subsequent scans #230
• Fix a rare race condition where a faulted RPC could remove a newly discovered leader master. #235

Misc:

• Update dependencies

Testing:

• Test against Kudu 1.16 #222

0.1.0

First non-preview version!

New features:

• Enable source link and deterministic builds #202
• Add simple object mapper #208
• Add scanner CountAsync() extension method #210
• Add ability to require authn/encryption #212

Optimizations and refactors:

• Remove special NaN encoding #204
• Minor cleanup and optimization in SSE key encoder #205
• Improve getting raw/binary values from ResultSet #209
  • Rename GetRawFixed() to GetSpan(), which now works for all data types
  • GetBinary() now returns a byte[]
  • Add GetNullableBinary() for consistency

Bug fixes:

• Fix a rare race condition that could lead to a bad connection being cached indefinitely #206

Testing:

• Various minicluster improvements #217

Preview 5

Contains incremental improvements towards the first stable release.

New features:

• Support Kudu transactions #174, see https://github.com/apache/kudu/blob/master/docs/design-docs/transactions.adoc
• Add support for UPDATE_IGNORE and DELETE_IGNORE operations. This instructs the Kudu server to ignore 'row not found' errors. #163
• Expose scan projection schema on KuduScanner #161
• Expose Location and ClusterID #162
• Add support for table comments #164
• Make SASL proto name configurable #165

Optimizations and refactors:

• Scanner improvements #167, #169
  • Consolidate rowwise and columnar results into a single ResultSet
  • RowResult is now a struct instead of a ref struct, simplifying using it
  • Bump default scanner batch size to 8mb, improving performance
• Support serializing/deserializing scan token metadata. This reduces the load on Kudu master servers by avoiding unnecessary metadata requests #160
• Avoid overloading Kudu masters with ConnectToMaster RPCs for high-throughput applications #166
• Make KuduSessionOptions immutable and improve default session options #172, #190

Bug fixes:

• Fix GetTablesAsync() without a name filter #177

Misc:

• Add .NET 6 target and optimizations #180
• Removed unused DefaultAdminOperationTimeout configuration #159
• Improve xml comments #171
• Use compile-time logging source generation #186
• Enable nullable references #188
• Refactor Negotiator #184, #185

Testing:

• Fix some flaky tests #168, #183

Preview 4

Contains incremental improvements towards the first stable release.

New features:

• Alter table ownership support #139
• Support resolving one master address to multiple addresses #145

Optimizations and refactors:

• Optimize key encoding #141
• Simplify and optimize KuduPartitioner #150
• Improve exception handling #146

Bug fixes:

• Try a scan at a different server when quiescing #151
• Fix negotiation over TLSv1.3 #154

Misc:

• Switch serialization library to Google.Protobuf #153
• Add .NET 5 target #140
• Use .NET threadpool scheduler #155
• Update dependencies

Testing:

• Test against Kudu 1.15

Preview 3

Contains incremental improvements towards the first stable release.

New features:

• Support columnar scan format
• Expose partitioning information through KuduPartitioner
• Add support for scanner keep-alive

Optimizations and refactors:

• Refactor sidecar handling and scan parsing
• Add type and null checks to RowResult.GetRawFixed()

Bug fixes:

• Make empty projection scans consistent between rowwise and columnar formats
• Fix improper use of pipelines during connection shutdown

Misc:

• Update dependencies
• Add more XML documentation comments

Testing:

• Port more tests from the Java client

Preview 2

Contains incremental improvements towards the first stable release.

New features:

• Expose stats about scan filters
• Add support for INSERT_IGNORE
• Expose callback for session exceptions

Optimizations and refactors:

• KuduPredicate type validation and performance improvements
• Improve connection shutdown logic
• Add ISecurityContext interface, which allows testing KuduClient without a real server
• Add missing GetBinary(name) overload to RowResult
• Add ToString() to RowResult
• WriteAsync() improvements:
  • Throw an exception on per-row errors (eg. row already exists for inserts, doesn't exist for updates)
  • Return WriteResponse, which contains the raw write timestamp

Bug fixes:

• Prefer tservers in the same location when multiple are local

Misc:

• Add VS Code support
• Update dependencies

Testing:

• Use Knet.Kudu.Binary for integration tests, the .NET port of the kudu-binary jar, https://kudu.apache.org/2019/03/19/testing-apache-kudu-applications-on-the-jvm.html
• Port more tests from the Java client (client stress test, timeout, key encoding)

Preview 1

Very first release of Knet.Kudu.Client. Get it on nuget: https://www.nuget.org/packages/Knet.Kudu.Client