[MarkLogic Dev General] Getting odd results from Search API

[Will Thompson]

Micah,

Everyone is running ML 4.2-2.

Danny,

The Word Query settings are the same on both, as are the Attribute Range Indexes.  Those are the only indexes I have added to or changed so far.  Are there other indexing options I should be looking at?  

There are some differences in the database specification: values for reindexer timestamp, in memory limit/size/tree size/range index size/reverse index size, and journal size.


Thanks,

Will

-----Original Message-----
From: general-bounces at developer.marklogic.com [mailto:general-bounces at developer.marklogic.com] On Behalf Of Danny Sokolsky
Sent: Monday, April 11, 2011 8:15 PM
To: General MarkLogic Developer Discussion
Subject: Re: [MarkLogic Dev General] Getting odd results from Search API

And another thing to check: are the database indexing options the same on the various machines?

-Danny

-----Original Message-----
From: general-bounces at developer.marklogic.com [mailto:general-bounces at developer.marklogic.com] On Behalf Of Micah Dubinko
Sent: Monday, April 11, 2011 6:14 PM
To: General MarkLogic Developer Discussion
Subject: Re: [MarkLogic Dev General] Getting odd results from Search API

Will,

What version of MarkLogic are you running on your local box and on the others?

Since the value that ends up in @total comes comes from a high-speed estimate (cts:remainder actually) it can vary, but I also recall a bug that was filed and fixed regarding very similar behavior, though I can't find it this second.

-m

On Apr 11, 2011, at 4:42 PM, Will Thompson wrote:

> Hi folks,
> 
> I have a search application that uses Seach API, and everything is working fine on my local developer box.  However, when deployed to other developer computers and our QA server, the responses we're getting from Search API seem to be incorrect.
> 
> The first thing I noticed is that the @total attribute on the <search:response> element changes as I increment the result page number - i.e.: 
> 
> search:resolve(q-parsed,options,1,10) -> @total=97
> search:resolve(q-parsed,options,11,10) -> @total=107
> search:resolve(q-parsed,options,21,10) -> @total=117
> ...
> 
> And so on, with the @total value increasing by 10 as I increase the start facet by 10.  On my box (with identical ML content) the @total value is always 875 for the same query, regardless of page number.
> 
> The second thing I noticed is that on the dev and QC boxes the search rank numbers appear to be "stuck" and at pretty high values. - i.e.: score="772" confidence="0.825606" fitness="1", for every result regardless of page number.  On my box the first result has: score="228" confidence="0.386277" fitness="0.61088", and those descend down with the results and on subsequent pages.
> 
> We have a set of scripts that automate populating ML with content, so the data in each database should be identical.  I configured my database through the admin interface, and the other databases were configured using scripts with admin: commands.  There are some differences in my config and the scripted config, but mostly things like memory/index size settings, nothing I would normally think could be responsible for these weird responses from Search API.
> 
> Has anyone seen this before?  If I knew how to make this reproducible I would, but there's a lot going on.  How do you recommend I go about isolating this issue?
> 
> 
> Thanks!
> 
> Will
> _______________________________________________
> General mailing list
> General at developer.marklogic.com
> http://developer.marklogic.com/mailman/listinfo/general

_______________________________________________
General mailing list
General at developer.marklogic.com
http://developer.marklogic.com/mailman/listinfo/general
_______________________________________________
General mailing list
General at developer.marklogic.com
http://developer.marklogic.com/mailman/listinfo/general