Hi Mike,

Yes, NULLS FIRST/LAST is used for pubdate.

I did not connect this issue to the missing normalizer issue.

Thanks for connecting the dots for me.

Liam


On Jan 31, 2014, at 8:26 AM, Mike Rylander <[email protected]> wrote:

> Liam,
> 
> Thanks for looking at this, but I think we should just use the
> approximate_low_date() normalizer and be done.  I believe a recently
> offered branch of yours should take care of making sure it is
> installed, and we can also add it to the normalizer list for pubdate
> without risk.
> 
> As for creating two attributes, that is a waste of space, will slow
> down search, and will lead to confusing user-level behavior.  By
> having differing values for replacement of u (or x, as is sometimes
> used) used for ASC or DESC, we'll just make the display order unstable
> and confusing when switching between the two.
> 
> As for nulls, we can use NULLS FIRST/LAST to push unknown values to
> the end.  We do that already for at least some of the ORDER BY columns
> ... and I thought pubdate was one of the.
> 
> 
> On Fri, Jan 31, 2014 at 10:58 AM, Liam Whalen
> <[email protected]> wrote:
>> We have noticed a problem with sorting by date where records with empty 
>> date1 values or date1 values with 'u' in them are sorted to the top of the 
>> search results in both Newest to Oldest and Oldest to Newest sorts.
>> 
>> I have worked around this with the following fix:
>> 
>> http://git.sitka.bclibraries.ca/gitweb/?p=sitka/evergreen.git;a=commit;h=476cf5aa7b14b9934ae5521046eb79c9d3d4573b
>> 
>> But, this forces calculations to happen when sorting, which fails to 
>> capitalize on the possibilities of an index.
>> 
>> I suggest that we change the pubdate index to pubdate_asc and pubdate_desc.
>> 
>> In the case of records with no date1 value, pubdate_asc would be given a 
>> value of 9999 and pubdate_desc would be given a value of 0.
>> 
>> This will force these records to the bottom of date sorted results.
>> 
>> In the case of records with 'u' characters in date1, we replace all 
>> occurrences of u with 9 in pubdate_asc and all occurrences of u with 0 in 
>> pubade_desc.  This will push these records to the bottom of their relevant 
>> date groups.
>> 
>> I can make these changes.  Is it a good idea to separate pubdate into two 
>> categories like this?
>> 
>> Liam
> 
> 
> 
> -- 
> Mike Rylander
> | Director of Research and Development
> | Equinox Software, Inc. / Your Library's Guide to Open Source
> | phone:  1-877-OPEN-ILS (673-6457)
> | email:  [email protected]
> | web:  http://www.esilibrary.com

Reply via email to