Print

Print


I do not think Solr facets are over-engineered. Trying to hack away to 
make Solr facets act kind of sort of like traditional "browse" search, 
I'm using all the features that are there, and wish they had a few more 
features -- in some ways they are 'under-engineered'.

I just think Solr facet interface is actually a fairly low-level 
interface at present, that may not be suitable for a standard like SRU.

I've needed those features, but I haven't needed em through SRU.

What features did your developers actually need through SRU?  They 
claimed to need complete access to every Solr facet feature, directly?

I mean, some SRU end points may be backed by an SQL rdbms. There are 
some things you can do with SQL that you can't do with SRU.  Does that 
mean that SRU ought to expose a parameter that lets you send arbitrary 
SQL queries?   Nope, for a variety of reasons. Including the obvious one 
that some SRU end points are NOT backed by an SQL rdbms.  And some SRU 
endpoints -- including ones that might be able to expose some facetting 
behavior -- are not backed by Solr, and simply mimicing the Solr API 
(which is even less of a 'standard' than SQL) and requiring SRU 
end-points that want to support facets to mimic it, is not appropriate.

There will still be things you could only do wtih direct SQL access (or 
direct Solr access), and not with SRU.  That's just the nature of the 
game. It can't be solved by trying to shove SQL or Solr's own API into 
SRU/CQL.

On 3/3/2011 2:42 PM, LeVan,Ralph wrote:
>> -----Original Message-----
>> From: Code for Libraries [mailto:[log in to unmask]] On Behalf
> Of
>> Jonathan Rochkind
>>
>> [I agree that simply copying the Solr API for a standard like SRU is
> not
>> the way to go -- Solr is an application that supports various
> low-level
>> things that are not appropriate in that level of detail for a standard
>> like SRU or what have you, at least not until they've been shown to be
>> needed.]
> Sadly, the use case for me was a group of SOLR developers who wanted an
> SRU interface and claimed to want SOLR facets exposed.  Now, if as I
> suspect, SOLR facets are overengineered and have features that have not
> been shown to be needed, then I've overengineered too.  But I don't feel
> like I can impose an extreme programming attitude to their requirement
> and make them justify each SOLR feature.
>
> Ralph
>