[
https://issues.apache.org/jira/browse/NUMBERS-156?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17334764#comment-17334764
]
Matt Juntunen commented on NUMBERS-156:
---------------------------------------
[~aherbert], I have an implementation of this ready but the existing unit tests
for {{SafeNorm}} are sparse and I'd like to expand them to verify the
correctness of the new method. My idea was to create a vector generator that
produces random double arrays, computes an exact norm using BigDecimal,
computes a standard norm using double operations, and then compares these to
the output of {{SafeNorm}}. My thought was that the {{SafeNorm}} output should
be at least as close to the exact result as the directly computed norm.
However, I'm running into issues getting an exact result, specifically with
computing the square root of the sum of squares. I'm using Java 8 (which
doesn't have BigDecimal.sqrt) so I've resorted to using Newton's method with
mixed results. I feel like there's a much better way to do this. Do you have
any ideas on how to approach these tests? Is my initial assumption that
{{SafeNorm}} should be at least as close to the exact value as a standard
computation correct?
> SafeNorm 3D overload
> --------------------
>
> Key: NUMBERS-156
> URL: https://issues.apache.org/jira/browse/NUMBERS-156
> Project: Commons Numbers
> Issue Type: Improvement
> Reporter: Matt Juntunen
> Priority: Major
>
> We should create an overload of {{SafeNorm.value}} that accepts 3 arguments
> to potentially improve performance for 3D vectors.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
