Schemaless mode and solrconfig.xml

[Erick Erickson <er...@gmail.com>]

How do all the components that define a "text" field for various
purposes play nice in schemaless mode?

The <df> entry is commented out of the /select handler which at least
lets us get started.

But there are a variety of places where "text" is specified, the
/browse handler, spell checking, etc. What ideas exist or are already
under way for making it easier to make Solr run OOB in schemaless with
these bells and whistles? Or will it just be required that people edit
solrconfig.xml to reflect fields that actually get created via
schemaless?

Or is the best solution just to take everything out of the
solrconfig.xml file that appears in the example-schemaless
configuration except, say the select handler?

Or pre-define a "text" field in the schemaless schema.xml file to handle these?

I'm not even sure we _could_ make things like the spell checker play
nice with schemaless. For the other handlers we can specify &df= on
the URL once we know there's a field available, but that's tricky for
spell checking...

And I think browse is just totally out of the question since it's
coded up for the default schema. Pull that too?

Thoughts?
Erick

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org

Re: Schemaless mode and solrconfig.xml

[Gregory Chanan <gc...@cloudera.com>]

Long term, it seems like the solution is to have a SolrConfig API that is
powerful enough to define things like "df" at runtime.  Definitely worth
discussing what we should do until then, though.

Or pre-define a "text" field in the schemaless schema.xml file to handle
> these?
>

I'd lean towards not defining any "df"s in the schemaless solrconfigs over
this solution.  I'd imagine if we define "text" in the schemaless
schema.xml, a user would do the following:
- index some data
- query and get no results and be confused

Alternatively, if we don't define any "df"s they pattern would be:
- index some data
- query and get error about field not specified or df not defined
from there the user would at least have a starting point and decide to
specify the df in the query, specify the field, or edit the solrconfig.xml
and restart the service.  That seems like it would result in less
unexpected behavior and fewer user questions.

Note, this means we should probably get rid of the "df" for /query in the
schemaless example.


>
> I'm not even sure we _could_ make things like the spell checker play
> nice with schemaless. For the other handlers we can specify &df= on
> the URL once we know there's a field available, but that's tricky for
> spell checking...
>
> And I think browse is just totally out of the question since it's
> coded up for the default schema. Pull that too?
>

+1 on pulling both of those.

>
> Thoughts?
> Erick
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>
>