[PIG-3926] ROUND_TO function: rounds double/float to fixed number of decimal places

• Type: New Feature
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.12.0
• Fix Version/s: 0.13.0
• Labels: builtin function math
• Patch Info: Patch Available

Release Note

The ROUND_TO(val, digits[, mode]) function accepts a single float or double value, along with an integer number of digits, and returns the value to that number of decimal places. For example, ROUND_TO(3.14159f, 3) returns 3.142f (a float); ROUND_TO(3.14159d, 0) returns 3.0d (a double); and ROUND_TO(1234.56d, -2) returns 1200d. The optional third argument controls the rounding mode; see the documentation.

• Assignee: Flip Kromer
• Reporter: Flip Kromer
• Created: 07/May/14 01:48
• Updated: 07/Jul/14 18:07
• Resolved: 10/May/14 23:04

Description

The too-often used trick of rounding to a fixed number of decimal places by writing '1000 * Math.round(num / 1000.0)' is not only unsightly, it's numerically imprecise (http://stackoverflow.com/a/12684082/41857). The attached patch adds a function to do this.

ROUND_TO(val, digits) accepts a single float or double value, along with an integer number of digits, and returns the value to that number of decimal places

Examples:

ROUND_TO(3.14159f, 3) -- returns a float 3.142
ROUND_TO(3.14159d, 0) -- returns a double 3.0.

I also added unit tests for ROUND – there were none before that I could find.

I don't know how to write a unit test that the schema correctly routes to FloatRoundTo and DoubleRoundTo – if the ones I provided are insufficient please point me to a simple example to emulate.

Issue Links

• relates to: PIG-3932 Document ROUND_TO builtin UDF