Possible bug/issue when index sorting is enabled.

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Possible bug/issue when index sorting is enabled.

vish.ramachandran
Hi Lucene Developer community,

I may have hit a Lucene issue and I would like clarification if this is
indeed a bug.

Steps:
1. Enable index sorting for the index
2. Add document with id=1 and yVal=Long.MaxValue
3. Delete document with id, followed by add document with same id with
yVal=45.
4. Flush index.
5. Search for documents with yVal=Long.MaxValue. Deleted document is
returned as part of results.

Here is the concise sample code that reproduces bug:

https://gist.github.com/vishramachandran/4a8443fb67ad74b2f597a58ec0913411

The issue does not reproduce if index sorting is disabled in the code.

I checked with Tomas F. Lobbe on this issue and he was able to concur with
me that this may be a bug - but we wanted to check with the group before
filing one.

Thanks
Vish.
Reply | Threaded
Open this post in threaded view
|

Re: Possible bug/issue when index sorting is enabled.

Adrien Grand
Thanks for reporting this Vish, this is a bug indeed. I was able to dig a
bit into why your test fails with delete+add but not with updateDocument
and opened https://issues.apache.org/jira/browse/LUCENE-8466 with some more
context.

Le lun. 27 août 2018 à 07:45, <[hidden email]> a écrit :

> Hi Lucene Developer community,
>
> I may have hit a Lucene issue and I would like clarification if this is
> indeed a bug.
>
> Steps:
> 1. Enable index sorting for the index
> 2. Add document with id=1 and yVal=Long.MaxValue
> 3. Delete document with id, followed by add document with same id with
> yVal=45.
> 4. Flush index.
> 5. Search for documents with yVal=Long.MaxValue. Deleted document is
> returned as part of results.
>
> Here is the concise sample code that reproduces bug:
>
> https://gist.github.com/vishramachandran/4a8443fb67ad74b2f597a58ec0913411
>
> The issue does not reproduce if index sorting is disabled in the code.
>
> I checked with Tomas F. Lobbe on this issue and he was able to concur with
> me that this may be a bug - but we wanted to check with the group before
> filing one.
>
> Thanks
> Vish.
>