[ 
https://issues.apache.org/jira/browse/ARROW-5336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17251442#comment-17251442
 ] 

Weston Pace commented on ARROW-5336:
------------------------------------

[~wesm] [~bkietz] I'm planning on taking this for 3.0.0 (after discussion with 
Ben).  A few questions:

The I/O APIs and builder API make use of delta dictionaries.  However, these 
seem to not be required in the `Array` itself.  By the time we arrive at 
`Array` any "deltaing" has been taken care of.  So I believe I do not have to 
worry about delta dictionaries for this correct?

There is an existing `DictionaryUnifier` which seems to be unused.  There is 
some comment referring to `Transpose` but that method does not appear to make 
use of this.  However, it appears it is a good place to start.  Is there any 
reason this would not be so?

The existing `DictionaryUnifier`has a check on `null_count` with a message 
`Cannot yet unify dictionaries with nulls`.  From what I've read `null_count` 
is only referring to the nulls in the array's buffer and not nulls in the array 
data's dictionary itself.  I'm not sure how these would affect unification.  
These fields don't have an index in the dictionary so they can be safely 
ignored correct?

> [C++] Implement arrow::Concatenate for dictionary-encoded arrays with unequal 
> dictionaries
> ------------------------------------------------------------------------------------------
>
>                 Key: ARROW-5336
>                 URL: https://issues.apache.org/jira/browse/ARROW-5336
>             Project: Apache Arrow
>          Issue Type: Improvement
>          Components: C++
>            Reporter: Wes McKinney
>            Assignee: Weston Pace
>            Priority: Major
>             Fix For: 3.0.0
>
>
> Currently (as of ARROW-3144) if any dictionary is different, an error is 
> returned



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to