tag:blogger.com,1999:blog-6848574.post736856052823901895..comments2024-02-14T04:44:39.043-06:00Comments on Working notes: Paulahttp://www.blogger.com/profile/03653112583629043593noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-6848574.post-86246046344742057952010-05-13T11:35:07.964-05:002010-05-13T11:35:07.964-05:00I wasn't specifically thinking about Mulgara, ...I wasn't specifically thinking about Mulgara, so you're partly right. The initial search will certainly take the same number of seeks as searching for a subject in the SPO index. However, you can keep the results of this search, and use that for all subsequent searches on subjects. That means that you've cut the size of the search space significantly, which was my point here.<br /><br />I was actually coming at this from my <a href="http://code.google.com/p/jsparqlc/source/browse/trunk/src/main/java/org/mulgara/rdf/Graph.java" rel="nofollow">Graph API</a>, rather than Mulgara. This interface has a method that returns a List in the way that I've described. At that moment, the only place that appears (outside of my own projects at home) is in jSPARQLc. It's just that I'm finding the API so useful that I'm wondering if I should pack it into it's own Jar somewhere. It's particularly useful as a lightweight API for reading small RDF graphs.Paulahttps://www.blogger.com/profile/03653112583629043593noreply@blogger.comtag:blogger.com,1999:blog-6848574.post-49080209591094778002010-05-13T10:11:50.829-05:002010-05-13T10:11:50.829-05:00Interesting points about the differences in index ...Interesting points about the differences in index structures. However, isn't it the case that, at least with the Mulgara XA1.1 index structures, the PSO index is still indexing statements in a tree and not predicates? In other words, the tree is just as deep as the one ordered by subject, it's just that you'll only have to do the lookup based on predicate once and you'll get a larger slice of the tree once you do. Then again, if you're keeping predicates in a hashtable then it's a moot point.<br /><br />Also, are you talking about this in the context of adding low-level support in the triple store itself for list retrieval? In Mulgara terms, it sounds like what you're talking about is opening two Tuples at once, and iteratively re-applying prefixes to build up the list contents. That certainly isn't possible with the cursor-based Answer API.Alexhttps://www.blogger.com/profile/08316209066205143873noreply@blogger.com