The Impala Query Language and Built-in Functions - Learning Cloudera Impala - page 62

The CAST() function can be use to convert DOUBLE values to FLOAT ,
TINYINT , SMALLINT , INT , BIGINT , STRING , TIMESTAMP , or BOOLEAN .
Exponential notation in DOUBLE literals can be used when casting from
STRING.
As an example, value 1.0e6 represents one million.
• FLOAT : A FLOAT data type in Impala represents a 4-byte single precision
floating-point data type when used with the CREATE TABLE or ALTER
TABLE statement. The value of DOUBLE in Impala ranges between
1.40129846432481707e-45 and 3.40282346638528860e+38, and this value
can either be positive or negative.
Casting with the FLOAT data type works as follows in Impala:
Impala automatically converts FLOAT to a more precise DOUBLE value but
not from DOUBLE to FLOAT.
For converting FLOAT to other data types, you can use the CAST() function
and the result will be any one of the TINYINT , SMALLINT , INT , BIGINT ,
STRING , TIMESTAMP , or BOOLEAN data types.
Exponential notations can be use in FLOAT when needed. Exponential liter-
als can also be used when casting from STRING ; as an example 1.0e6 rep-
resents one million.
• STRING : A STRING data type holds a maximum of 32767 bytes of data in it
when used with the CREATE TABLE or ALTER TABLE statements. Here are
the key features of using the STRING data type:
• It is suggested that you limit the string values to the ASCII character
set for full support.
• Multibyte characters can also be used; however, their application will
be limited to query operations. String manipulation, comparison oper-
ators, and the ORDER BY clause may not function correctly with mult-
ibyte characters in the STRING data type.