Cheng Lian created HIVE-12069: --------------------------------- Summary: Hive cannot read Parquet decimals backed by INT32 or INT64 Key: HIVE-12069 URL: https://issues.apache.org/jira/browse/HIVE-12069 Project: Hive Issue Type: Sub-task Affects Versions: 1.2.1 Reporter: Cheng Lian
According to the [DECIMAL section|https://github.com/apache/parquet-format/blob/5b806d1e855bf47f5234c768aefc000b704f43ab/LogicalTypes.md#decimal] of parquet-format spec, low precision Parquet decimals can be written as {{INT32}} or {{INT64}} for better storage and query efficiency. However, currently Hive only recognizes decimals backed by binaries (either {{BINARY}} or {{FIXED_LEN_BYTE_ARRAY}}). We can improve this by providing a more general {{PrimitiveConverter}} [here|https://github.com/apache/hive/blob/release-1.2.1/ql/src/java/org/apache/hadoop/hive/ql/io/parquet/convert/ETypeConverter.java#L127-L137]. -- This message was sent by Atlassian JIRA (v6.3.4#6332)