[
https://issues.apache.org/jira/browse/DATAFU-60?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14108012#comment-14108012
]
Matthew Hayes edited comment on DATAFU-60 at 8/23/14 2:25 PM:
--------------------------------------------------------------
Tried running TestRange tests and I saw a failure:
{code}
PASSED: testFloatingBuckets
PASSED: testIntegerBuckets
PASSED: testMalformedBuckets1
PASSED: testMalformedBuckets2
PASSED: testNegInfinity
PASSED: testScientificBuckets
FAILED: testInfinity
{code}
The problem is 0 is supposed to be in the range:
{code}
@Test
public void testInfinity()
{
NumericalRange range = NumericalRange.fromRangeString("[0,*)");
assertEquals(range.isInRange(0), false);
assertEquals(range.isInRange(1e10), true);
}
{code}
was (Author: matterhayes):
Tried running TestRange tests and I saw a failure:
{code}
PASSED: testFloatingBuckets
PASSED: testIntegerBuckets
PASSED: testMalformedBuckets1
PASSED: testMalformedBuckets2
PASSED: testNegInfinity
PASSED: testScientificBuckets
FAILED: testInfinity
{code}
> Support NDCG calculation within a UDF
> -------------------------------------
>
> Key: DATAFU-60
> URL: https://issues.apache.org/jira/browse/DATAFU-60
> Project: DataFu
> Issue Type: New Feature
> Reporter: Joshua Hartman
> Labels: features
> Attachments: DATAFU-60-v2.patch, DATAFU-60.patch
>
> Original Estimate: 48h
> Remaining Estimate: 48h
>
> Ndcg is a common evaluation metric for the quality of a list of items
> presented to a user. It is an especially common metric used in the search
> literature.
> This feature request is to implement a udf to calculate ndcg.
> NDCG can be calculated using any function to represent the value of a
> position. Several useful functions should be available as part of the datafu
> library. First is the standard 1/logarithmic discounting factor. Another
> option should be the ability to supply a custom positional value for any
> range of positions in the case that a positional "value" is already well
> understood. However, the actual discounting function used should be easily
> pluggable in the event something custom is needed
--
This message was sent by Atlassian JIRA
(v6.2#6252)
