Write a Query Planner library, focusing on BGP processing

[lecoqlibre]

This Query Planner (QP) library will be used to rewrite SPARQL queries with indexes being taken into account.

This way, developers could write their SPARQL queries into their application without having to know about indexing.

The QP will know what indexes could be used, find the most appropriate ones and rewrite the query accordingly before to pass it to the query engine for execution.

There also should be a possibility for the developer to let his query been executed without rewriting. This way, developers would have only one interface to comply with to execute query (there is no need to use a separated query engine).

The query engine should be abstracted so different approaches could be used. We should support Comunica. Waiting for Comunica to support queries using named graphs, we should provide a workaround.

• We need to think about how to pass indexes to the QP. Could we have one description resource about all the indexes?
• We need a way to allow the QP logic extendable, so new index type could be supported (perhaps using DP strategy, visitor, chain of responsibility).
• Write the SPARQL query for TypeIndex.
• Write the SPARQL query for Implement indexes discovery for skills and cities #30.
• Create a new repository for the QP ("solid-search"?).
• Publish the QP library on NPM.

[pchampin]

We discussed this during the call yesterday.
A conclusion was that rewriting arbitrary SPARQL queries is too hard a problem to address directly, but we will investigate the rewriting of BGPs .