Documentation

Author: lriggs

docs/source/overview.rst

@@ -45,6 +45,9 @@ but some modules are JNI bindings to the C++ library.
    * - arrow-vector
      - An off-heap reference implementation for Arrow columnar data format.
      - Native
+   * - arrow-vector-codegen
+     - Template files for Arrow datatypes suitable for code generation.
+     - Native
    * - arrow-tools
      - Java applications for working with Arrow ValueVectors.
      - Native

vector/pom.xml

@@ -205,6 +205,12 @@ under the License.
             </goals>
             <phase>package</phase>
             <configuration>
+              <!-- Arrow's type system has many types, and projects building operations across Arrow data
+              (comparisons, casts, aggregations) benefit from generating type-specialized code rather
+              than hand-writing implementations for each type.
+              The TDD files provide a machine-readable definition of Arrow's types that enables this.
+              Including them in the distribution allows downstream projects to generate code that 
+              stays in sync as Arrow's type system evolves.-->
               <classifier>codegen</classifier>
               <classesDirectory>${basedir}/src/main/codegen</classesDirectory>
               <includes>