[ 
https://issues.apache.org/jira/browse/SOLR-8770?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15187625#comment-15187625
 ] 

Shawn Heisey commented on SOLR-8770:
------------------------------------

I've gone digging into the code a little bit.  I don't entirely understand what 
I'm looking at.

For the binary writer, I cannot see anything that delves into the individual 
fields of a document, so there's nowhere to add a null check.  The code (in 
JavaBinUpdateRequestCodec) appears to use an iterator to add documents to a 
NamedList.

The code in ClientUtils#writeVal seems to indicate that there might be 
legitimate uses for a null object, as there are certain situations where a null 
will be added to the XML.  I would need to understand other areas of the code 
to know what's going on there.

If there are indeed legitimate uses for a null object in SolrInputDocument 
field, then it will not be possible to throw IAE in SolrInputDocument.


> BinaryRequestWriter interprets null object in field as literal "NULL" string
> ----------------------------------------------------------------------------
>
>                 Key: SOLR-8770
>                 URL: https://issues.apache.org/jira/browse/SOLR-8770
>             Project: Solr
>          Issue Type: Bug
>          Components: clients - java
>    Affects Versions: 5.5
>            Reporter: Shawn Heisey
>
> From what I've been able to determine, if a null object is added with 
> SolrInputDocument#addField, the xml writer does not include that field in the 
> request, but the binary writer sends the literal string "NULL".
> This became apparent when upgrading SolrJ to 5.5, which uses the binary 
> writer by default.  Switching back to 5.4.1 fixed it, until I forced the 
> 5.4.1 client to use the binary writer.  My source data is MySQL.  JDBC is 
> where the null objects are coming from.
> Adding a null check to my doc.addField call has fixed my program with the 5.5 
> client, but this is likely to catch other upgrading users off guard.
> At the very least, the 5.5.1 CHANGES.txt file needs a note, but I believe the 
> behavior of the binary writer should match the behavior of the xml writer.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to