You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Emir Arnautović <em...@sematext.com> on 2018/04/02 07:49:59 UTC

Re: MatchMode in Dismax parser

Hi,
In case you want to round up, you can use negative numbers and percentage of failed matches so 75% of matches rounded up can be written as -25%.

HTH,
Emir
--
Monitoring - Log Management - Alerting - Anomaly Detection
Solr & Elasticsearch Consulting Support Training - http://sematext.com/



> On 29 Mar 2018, at 17:00, Shawn Heisey <ap...@elyograg.org> wrote:
> 
> On 3/29/2018 1:42 AM, iamluckysharma.0910@gmail.com wrote:
>> Just a suggestion , Shouldn't we need to use Math.round instead of direct int when watch mode is in %,
>> example i have 3 boolean clauses if i go for mm=50%, currently it reduce it to ~1, instead it can be ~2,
>> 
>> another example could be when we have 5 boolean clauses and mm=75%, we get calc as 3.75 currently it took 3, so instead of 3 it should have taken 4. as Math.round()
> 
> Maybe that is what it SHOULD do, but in most languages, converting a float value to an integer truncates the decimal portion, it doesn't round.  To do that requires a deliberate choice in the code, and that probably doesn't exist in dismax/edismax.  If your assertion is that this should have been done from day one, I'd say you're right.  But that decision is now ancient history.  The person who wrote the code might have had a very good reason to NOT do it that way.
> 
> At this point, if the functionality were changed, it would result in an upgraded Solr version behaving VERY differently than the previous version.  While new functionality is often added in any new minor release, changing existing behavior that users rely on without a configuration option is usually only done in a major version.  So for 7.x, dismax/edismax would need an option to enable rounding on minimum-should-match calculations.  Sounds like a great feature request to put into Jira, and patches are always welcome.
> 
> Thanks,
> Shawn
>