Choose the most recent release version from incubator-datasketches-xxx.
Or, clone or fork the current SNAPSHOT directly from the relevant repository.
Apache DataSketches uses semantic versioning. Version numbers use the form major.minor.incremental and are incremented as follows:
The zip files downloaded from incubator-datasketches-xxx include a version number in the name, as in apache-datasketches-java-1.1.0-incubating-src.zip. This same number is also in the top section of the pom.xml file.
If you are developing using Maven and want to use, for example, incubator-datasketches-java, add the following dependencies to your pom.xml file:
<dependency> <groupId>org.apache.datasketches</groupId> <artifactId>datasketches-java</artifactId> <version>1.1.0-incubating</version> </dependency>
These are fully tested, production quality releases, and hopefully as bug-free as humanly possible. However, the code is continuously evolving and improvements in performance, documentation, additions to the API, and bug fixes do occur. When enough of these build up, especially for bug fixes or performance improvements, a new release will be issued and the minor digit will be incremented. The incremental digit will only be used for bug fixes as stated above.
As stated above, the major digit is being reserved for major refactorings of the library where backward API compatibility may not be guaranteed.
For the repositories that depend on java core, such as incubator-datasketches-hive, the version number may be incremented just to be in sync with the java core repository, and may not reflect any actual code changes other than a change in the pom.xml to reflect the new dependency.
If you just want to run Hive and don't require direct access to the incubator-datasketches-java it is recommended that you download the “with-shaded-core.jar”, which includes the Hive jar as well as shaded versions of the core jar and memory jar. The shading avoids conflicts with other possible versions of core and memory that you might have in your system.
If you want the latest and greatest version of the code, it is certainly OK for you to create your own snapshot jars from a clone or fork. The code is automatically tested using the current test suite, but you might catch the code in transition to a new future release. Caveat Emptor.
Please use GitHub revisions history on the respective repositories