I've already implemented something similar to what you've said. Setting
the _root_ into the document and for safety double checking that I don't
have the same ID twice.
I'm interested in the workaround.
Thanks
On 12/10/2015 10:24 AM, Mikhail Khludnev wrote:
I want to fix it in scope of https://issues.apache.org/jira/browse/SOLR-5211
and hope it will be released in 5.5.
It will *copy* $UniqueKey into "_root_" aways (regardless of having a
child).
Thus we will have $UniqueKey really unique. Overriding update will work.
But deleteByID won't work out-of-the-box for parents with child. And
deleteByQuery should be tweaked to handle blocks.
I almost done it in earlier patch.
https://issues.apache.org/jira/secure/attachment/12739197/SOLR-7606.patch
Is it really urgent for you? I'm asking because there is an easy
workaround, despite it seems odd overall.
On Thu, Dec 10, 2015 at 10:41 AM, Bogdan Marinescu <
bogdan.marine...@awinta.com> wrote:
Any suggestions about this ?
On 12/04/2015 08:26 AM, Bogdan Marinescu wrote:
Hi Mikhail,
I would expect the same behaviour as for a database. Meaning if I have a
field declared as an uniqueKey, then there should only be one document with
that key, regardless if it has a child or not.
If you add the childless document first and afterwards the child, then
sol'r should append the child to the already existing document (or rather
delete the existing one as the new one has newer data).
It's weird because when I query sol'r for that ID, I get two documents
when I am only expecting one.
I could do a sort of 'workaround' logic where I would always pick the one
with children but I think better would be to fix this in sol'r.
Regarding your issue SOLR-5211 <
https://issues.apache.org/jira/browse/SOLR-5211>. If a document has
children and you update the parent as childless or in fact delete the
parent altogether, then the child documents should also be deleted.
I've faced this problem where I was just deleting the parent by id and I
had lots of orphan documents just laying around.
Regards,
Bogdan Marinescu
On 12/03/2015 06:26 PM, Mikhail Khludnev wrote:
Hello Bogdan,
You described how it works now. That's how it was implemented. And I can
explain why it was done so.
Could you please describe the expected behavior for you?
Notice, I want to enforce nested (block) behavior always in scope of
https://issues.apache.org/jira/browse/SOLR-5211. So, the fields
assigned to
parent with child and childless single doc will be the same. So, far it's
not clear how to amend <uniqueKey> semantic.
On Thu, Dec 3, 2015 at 6:35 PM, Bogdan Marinescu <
bogdan.marine...@awinta.com> wrote:
Hi,
I have a problem with nested docs. If I create a document with id: 1 and
fieldA:sometext and then add it to sol'r, I get one doc in sol'r.
Afterwards if I add a child/nested doc to this document I additionally
get
a _root_:1 to the document but the problem is I now have two documents
with
the same ID (id: 1) in sol'r, one with _root_ and the child/nested doc
and
one without it.
Any ideas why this happens?
Any ideas how to avoid this?
Thanks,
