TODO

• Update term sorting code to use SPARQL comparison rules
• Implement simple filtering over BGP matching (1-variable SARGs plus simple 2-variable ops)
  • Numeric logical testing (var, const)
  • Date logical testing (var, const)
  • LANGMATCHES (LANG(var), const)
• Implement aggregates over BGP matching, with support for grouping by variable (but not complex expressions)
  • Use an AVL tree with item = (nodeid_t* group_key, table_t* table), then evaluate just like the materialization used for sorting
  • SUM
  • AVG
  • MIN
  • MAX
  • COUNT
• Implement subset of property paths (does the obvious implementation correlate to the ALP algorithm?)
  • p*
  • ^p (disregard; this can be avoided during planning)
  • p/q
  • p|q
  • !p
• Add optional text indexing (map words to list of graph node IDs; specify participating predicates)
  • Optimize matching of BGPs when there is a filter which is subsumed by keyword matching (CONTAINS filters where the pattern contains at least one whole word)
• Graph algorithm to produce schema statistics useful for query planning (identify (inverse-) functional properties, predicate cardinality, etc.)

Done

• Canonicalize language tags on input
• Translate document-scoped bnode IDs to unique IDs on import
• Perform simple lexical verification/canonicalization on input for known D-types (numerics, dates)
• Add support for property paths to AtteanX::Store::MemoryTripleStore::Query
• Ensure arguments are argument compatible for string filter functions.
• Add native numeric slot in rdf_term_t to allow native numeric operations
• Make AtteanX::Store::MemoryTripleStore::Query consume Attean::API::Plan
• Implement count of triple matching, expose in XS (run triple matching, but avoid materializing perl objects for results)
• Generalize BGP matching to a query structure that can nest operations (BGPs, filters, aggregates, property paths)
• Implement simple filtering over BGP matching (1-variable SARGs plus simple 2-variable ops)
  • ISIRI/ISLITERAL/ISBLANK (var)
  • SAMETERM (var, const)
  • STRSTARTS (var, const)
  • STRENDS (var, const)
  • REGEX (var, const) (link with pcre; base on code in ts.c)
  • CONTAINS (var, const)
• Implement dump to/load from disk
  • Serialize term values AVL tree (simple loading code as it's guaranteed to be unique)
  • Dump edges array directly to disk (with int elements in network order)
• Implement subset of property paths (does the obvious implementation correlate to the ALP algorithm?)
  • p+