[DRILL-1123] how scale (significant digits after decimal point) is determined for decimal operations

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.4.0
• Component/s: Functions - Drill
• Assignee: DrillCommitter
• Reporter: Chun Chang
• Created: 09/Jul/14 22:02
• Updated: 31/Jul/14 06:31
• Resolved: 21/Jul/14 02:48

Description

#Wed Jul 09 10:28:32 PDT 2014
git.commit.id.abbrev=810a204

The way we determine how many significant digits after decimal point during decimal operation seems confusion, and in some cases, just doesn't make sense. Here is the experiments I did:

0: jdbc:drill:schema=dfs> select cast('12.3456' as decimal(18,6)) / cast('2.0' as decimal(38,1)) from data limit 1;
------------

EXPR$0

------------

6.1728000000000000000000000

------------
1 row selected (0.163 seconds)
0: jdbc:drill:schema=dfs> select cast('12.3456' as decimal(18,6)) / cast('2.0' as decimal(38,10)) from data limit 1;
------------

EXPR$0

------------

6.1728000000000000

------------
1 row selected (0.173 seconds)
0: jdbc:drill:schema=dfs> select cast('12.3456' as decimal(18,6)) / cast('2.0' as decimal(38,20)) from data limit 1;
------------

EXPR$0

------------

6.172800

------------
1 row selected (0.136 seconds)
0: jdbc:drill:schema=dfs> select cast('12.3456' as decimal(18,6)) / cast('2.0' as decimal(38,30)) from data limit 1;
------------

EXPR$0

------------

6

------------
1 row selected (0.185 seconds)

Look at the last example, no decimal digits are displayed, even though I increased decimal scale.