On Monday, May 12, 2014 1:23:12 AM UTC-7, Michael Hunger wrote:
>
> if you are a bit familiar with Java you can also use the batch-inserter 
> API yourself to implement the things you need. 
>
> This also applies to other JVM languages too, like JRuby, Jython, ...


I realize this, but I'd much rather use your solution than hack up my own 
(:-).

 

> >  I'd like to express these as node properties, but the Batch Importer 
> >  uses TSV syntax, which has a fixed set of properties per node. 
> Yep, good insight, you don't want to store those value triples as 
> relationships. 
>

The YAGO distribution is divided into 25 "themes" (eg, Facts, 
Labels, MetaFacts). I could easily imagine wanting to filter results based 
on the theme.  Adding a "theme" property to a relation is easy, but I'm not 
sure how to encode the theme for properties.
So far, the best solution I have is to create a property ("yago:hasISBN") 
and a meta-property ("yago:hasISBN:theme"). Am I missing a better approach?

 

> Yes, it skips empty cells 
>

Cool.  There are ~50 possible properties, but I assume that's OK...
 
 

> >      Given that the nodes file no longer has ID numbers, how do I 
> >      tell the Batch Importer which entities to modify? 
>
> If it would work You could state the properties to look-up from an index 
> and then use those to find and update the nodes. But the index read 
> performance is much slower than the batch-inserter write performance. 
>
> Usually what I'd do is to programmatically read all nodes of the graph and 
> store the relevant lookup property (eg. url) and the node-id in a Map or 
> sorted array. Then you can find the node quickly by id and update it. 
>

Yes, but that would get me back into writing code. Fortunately, it looks 
like I can do the entire import in a single run, based one (HUGE) 
specification file.  

-- 
You received this message because you are subscribed to the Google Groups 
"Neo4j" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to