Syntax and semantic changes CDH 6.2.1 to CDP 7.0.3.2
Review the syntax and semantic changes in Hive after migrating to CDP 7.0.3.2 from
CDH 6.2.1. A link to Apache Hive JIRAs, if there is one, provides more information about the
semantic change.
Aliasing tables Hive 3 in CDP follows the SQL 11 standard regarding a table identifier, which cannot begin with a numeral.ANALYZE TABLE ... COMPUTE STATISTICS PARTIALSCAN removed The analyze command collects statistics on tables and partitions, but is slow when scanning many files. A command was added that operated on files in RCfile format to speed up statistics collection. The command was not widely used, not well documented, and evenutally removed.Decimal to string change Hive decimal to string conversion yields results having incorrect padding. LPAD on a decimal(7,1) value of 0 returns "0" (plus padding) but it should be "0.0" (plus padding).Decimal literals The default treatment of Decimal literals in Hive was changed from Double to Decimal as per the SQL standards.hive.stats.collect.rawdatasize removal In an early release of Hive, the hive.stats.collect.rawdatasize property collected statistics when set to true. As Hive matured, hive.stats.autogather was added to gather statistics when data is inserted into tables.HIVE_SUPPORT_SQL11_RESERVED_KEYWORDS To simplify the parser logic and largely reduce the size of generated parser code, the HIVE_SUPPORT_SQL11_RESERVED_KEYWORDS configuration has been removed.Limit scanned partitions Before Hive 1.0 a configuration property was added for limited scanned partitions that was replaced by another property.Overflow handling of decimals An ArrayIndexOutOfBoundsException is returned when a sum operation on a decimal column overflows beyond the maximum precision.Functions that changed A number of changes to Hive functions occurred in CDP.NULL related behaviors Several behavior changes are due to the handling of NULLs.Timestamp or date related behaviors A few behavior changes are related to timestamp and date formats, functions, and data types.