ABSTRACT

XML has dominated the enterprise landscape for fifteen years and still remains the most commonly used data format. Despite its popularity the usage of XML for "Big Data" is challenging due to its semi-structured nature as well as rather demanding memory requirements and lack of support for some complex data structures such as maps. While a number of tools and technologies for processing XML are readily available the common approach for map-reduce environments is to create a "custom solution" that is based, for example, on Apache Hive User Defined Functions (UDF). As XML processing is the common use case, this paper describes a generic approach to handling XML based on Apache Hive architecture. The described functionality complements the existing family of Hive serializers/deserializers for other popular data formats, such as JSON, and makes it much easier for users to deal with the large amount of data in XML format.