[core] Introduce Variant data

[Zouxxyy]

Purpose

Subtask of #4471

Variant was born from Apache Spark, and the encoding rules have been incorporated into Parquet https://github.com/apache/parquet-format/blob/master/VariantEncoding.md

This PR add the basic interface of Variant and ported the latest implementation of Spark.

public interface Variant {

    /** Returns the variant metadata. */
    byte[] metadata();

    /** Returns the variant value. */
    byte[] value();

    /** Parses the variant to json. */
    String toJson();

    /**
     * Extracts a sub-variant value according to a path which start with a `$`. 
     */
    Object variantGet(String path);
}

Usage

  String json = "{\"age\":35,\"city\":\"Hangzhou\"}";
  Variant v = GenericVariant.fromJson(json);
  assertThat(v.variantGet("$.age")).isEqualTo(35L);
  assertThat(v.variantGet("$.city")).isEqualTo("Hangzhou");

Tests

API and Format

Documentation

[JingsongLi]

+1