[
https://issues.apache.org/jira/browse/MAHOUT-756?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13063289#comment-13063289
]
Sean Owen commented on MAHOUT-756:
----------------------------------
I like most of this cleanup. I think we could leave the getQuick()/setQuick()
methods as-is -- no need for additional arg checking. Their intent is to be
used when the caller knows it's OK.
I don't know that it's a problem that this allows different rows to be of
different types. I don't think any caller uses it that way anyway.
It is interesting what "like" means for the method that specifies a new number
of rows -- what are the types of the new rows? Right now it forces them to
DenseVector which doesn't feel right if the other like() method does not do the
same. On the other hand, you're requesting a different number of rows, which
means the type of new rows aren't defined anyway, so how much can the caller
care?
I think the least surprising thing for the caller, who will normally have an
object with all one type of vector, is to have a result of the same type of
vector. So, in that sense I think your change is good.
But there isn't a need for a new likeVector() method; call Vector.like() as the
other method does.
Let me see what happens under these changes.
> VectorList (Matrix implementation) does not maintain cardinality getters
> correctly
> ----------------------------------------------------------------------------------
>
> Key: MAHOUT-756
> URL: https://issues.apache.org/jira/browse/MAHOUT-756
> Project: Mahout
> Issue Type: Bug
> Components: Math
> Reporter: Lance Norskog
> Priority: Minor
> Attachments: VectorList.patch
>
>
> VectorList (implements Matrix) is dynamically expandable, row-wise. There are
> three different ways to query the size of a Matrix, and VectorList does not
> correctly supply these values.
>
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
