[MATH-57] [math] NaN handling in StatsUtil.min, StatsUtil.max methods

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.0

Environment

Operating System: Windows 2000
Platform: PC

• Bugzilla Id: 37019
• Assignee: Unassigned
• Reporter: Mark Osborn
• Created: 11/Oct/05 20:58
• Updated: 23/Apr/09 02:26
• Resolved: 23/Apr/09 02:26

Description

Data with NaN appears to be incorrectly handled in some cases when calculating
min and max using StatsUtil package. See examples below:

data=

{NaN, 4.0, 5.0, NaN}

max (Apache)=NaN <-- correct.
min (Apache)=NaN <-- correct.

data=

{4.0, 5.0, NaN}

max (Apache)=NaN <-- correct.
min (Apache)=NaN <-- correct.

data=

{4.0, 5.0, 7.0}

max (Apache)=7.0 <-- correct.
min (Apache)=4.0 <-- correct.

data=

{NaN, 4.0, 5.0}

max (Apache)=5.0 <-- incorrect? NaN.
min (Apache)=4.0 <-- incorrect? NaN.

data=

{2.2343, NaN, NaN, 3435.32}

max (Apache)=3435.32 <-- incorrect? NaN.
min (Apache)=3435.32 <-- incorrect? NaN.

data=

{3435.32, NaN, NaN, 2.2343}

max (Apache)=2.2343 <-- incorrect? NaN.
min (Apache)=2.2343 <-- incorrect? NaN.

data=

{NaN, NaN, NaN}

max (Apache)=NaN <-- correct.
min (Apache)=NaN <-- correct.