RDF::Query (original) (raw)

Latest Version: RDF::Query v2.100

RDF::Query is an implementation of the SPARQL and RDQL query languages for Perl. It allows queries to be run against any of the following RDF models: RDF::Trine, RDF::Redland, RDF::Core.

RDF::Query may be downloaded from the CPAN. For updates on the project, see my RDF::Query weblog posts.

Version History

Version 2.100 (2009-03-18)

• API
  • Added ::Algebra::BasicGraphPattern::connected method.
  • Added 'node_counts' as a recognized key to ::Model::RDFTrine::suports().
  • Added ::Model::node_count method.
  • Added ::Model::RDFTrine::count_statements() and ::Model::RDFTrine::node_count() methods.
  • Added 'store' field to the data returned by the meta() method in the ::Model::* classes.
  • Added a peek method to ::Iterator to support execution deferment like in ::Algebra::Service.
  • Added ability to force SERVICE calls to abort using $ENV{RDFQUERY_THROW_ON_SERVICE} and RequestedInterruptError.
  • Added bf methods to ::Triple and ::BasicGraphPattern to describe the BGP in terms of bound/free.
  • Added bind_variables method to RDF::Query::Algebra.
  • Added caching to ::Algebra::Service::_names_for_node.
  • Added code to count (and warn) the rate of false positives from a bloomjoin.
  • Added cost model hooks in RDF::Query and ::Algebra::BasicGraphPattern.
  • Added FeDeRate BINDINGS to the list of supported extensions.
  • Added get_basic_graph_pattern to ::Model::RDFTrine (not used yet).
  • Added is_solution_modifier() methods to ::Algebra classes.
  • Added labels and common patterns to service description template.
  • Added more logic and debugging to aggregating triples into BGPs for SERVICE calls.
  • Added optional restriction argument to ::Algebra::subpatterns_of_type.
  • Added RDF::Query::Model::RDFTrine::BasicGraphPattner::graph() method.
  • Added RDF::Query::Node::compare for sorting (either Trine or Query) node objects.
  • Added RDF::Query::plan_class so that ::Federate can overload calls to ::Plan methods.
  • Added support for computed statement generators (like ARQ's computed property support [e.g. list:member]).
  • Added trial subsumes() methods to ::Algebra::BasicGraphPattern and ::Algebra::Triple.
  • Algebra classes now call RDF::Query::algebra_fixup for optimization before falling back on normal fixup code.
  • Allow equality test and disjunction filters in patterns that can be compiled to SQL.
  • Fixed ::Algebra::GroupGraphPattern to use join_bnode_streams in cases where the bloom filter wasn't automatically created but instead provided by the user.
  • Fixed ::Model::RDFTrine::as_string use with Quads.
  • Fixed argument list for RDF::Query::Algebra::Service::_names_for_node called from bloom function.
  • Fixed bug so ::Model::RDFTrine::meta may be called as a class method.
  • Fixed RDF::Query::algebra_fixup to ignore service capabilities that would result in empty BGPs.
  • Modified code for aggregate queries (though they are currently broken).
  • Moved construction of bloom-filter-optimized patterns to RDF::Query::Algebra::GroupGraphPattern::_bloom_optimized_pattern().
  • Moved initial federated service code into RDF::Query::fixup and ::algebra_fixup.
  • Moved QEP generation code to RDF::Query::query_plan so it can be overloaded.
  • Moved RDF::Query::add_service() to RDF::Query::Federate::add_service().
  • Parsing service descriptions now more forgiving in the face of missing values.
  • RDF::Query::Algebra::Triple::new now adds RDF::Query::Node::Variable objects for undefined nodes.
  • RDF::Query::fixup() now only returns a construct pattern (the query pattern now being returned by RDF::Query::query_plan()).
  • RDF::Query::Model::get_computed_statements now doesn't die if there's no query object present.
  • RDF::Query::new() now accepts two-argument form with just $query and \%options.
  • RDF::Query::Node::Literal objects now can compare as equal when they're of numeric type but the lexical values aren't numeric.
  • RDF::Query::prune_plans now takes ExecutionContext as an argument, and in turn calls ::Plan::prune_plans.
  • RDF::Query::query_plan() now returns all possible query plans when called in list context.
  • RDF::Query::query_plans now calls RDF::Query::prune_plans to select from the list of possible QEPs.
  • RDF::Trine::Node::Resource now escapes unicode in base URIs (now just relative URI part) before calling URI->new_abs.
  • Removed now unused RDF::Query::construct() and RDF::Query::fixup().
  • Removed old execute() code from ::Algebra classes.
  • Removed unused redland fallback code from RDF::Query::Model::RDFTrine.
  • Split RDF::Query::execute into prepare() and execute_plan() methods.
  • Converted RDF::Query::execute() to use ::Plan classes.
  • Updated ::Compiler::SQL to recognize ::Algebra::Project objects.
  • Updates to RDF::Query::execute() to support explicit pre-binding lists (instead of just scalar values).
  • ::Algebra::GroupGraphPattern now throws an exception if passed unblessed values as patterns.
  • ::Federate now labels nodes in the QEP tree with applicable services.
  • ::Model::debug() now shows data from the named graph model.
  • ::Model::RDFTrine::add_uri now decodes resulting content as utf8 before proceeding.
  • ::Model::RDFTrine::meta() probes the underlying store object to declare the proper 'store' class.
  • ::Service::_names_for_node updated to use ::Plan classes for query execution (fixes use of the k:bloom filter).
• Classes
  • Added algebra classes for solution modifiers and query forms (construct, project).
  • Added code and tests for Query Execution Plan classes RDF::Query::Plan::*.
  • Added RDF::Query::Federate::Plan for federation-specific code.
  • Added RDF::Query::BGPOptimizer implementing a basic optimizer for basic selectivity-based join ordering.
  • Added RDF::Query::CostModel classes for computing/estimating the cost of executing a specific pattern.
  • Added RDF::Query::ExecutionContext to hold all necessary information for query execution (query, model, bound variables).
  • Added RDF::Query::ServiceDescription for parsing DARQ-style service descriptions.
  • Added RDF::Query::VariableBindings to wrap existing HASH-based variable binding structure.
  • Added stub ::Plan::ThresholdUnion class for running optimistic queries.
  • Added workaround to RDFCore bridge so that RDF::Core doesn't die if getStmts is called with a Literal in the subj or pred position.
  • Added a RequestedInterruptError exception class.
  • Added code for RDF-Trine specific BGP query plans.
  • Changed debugging in RDF::Query modules to use Log::Log4perl.
  • Made SERVICE work again by rolling back streaming socket work (now commented out).
  • Moved federation code to new RDF::Query::Federate class.
  • Plan generation now includes any plans the model object can provide.
  • RDF::Query now always uses a cost model (defaulting to ::Naive).
  • RDF::Query::Federate now defaults to SPARQLP parser.
  • Removed logging/warn calls from forked process in ::Service (was screwing up the parent-child IO pipe).
  • Removed use of "DISTINCT" queries in SERVICE calls (for pipelining).
  • ServiceDescription now only runs sofilter if ?subject/?object are bound.
  • Started work on a more holistic approach to supporting service descriptions (instead of using add_computed_statement_generator()).
  • Updated ::Algebra::Service to fork and use LWP's callback mechanism for concurrency.
  • ::Algebra::Service now defers loading of content until the first result is requested.
  • ::Model::RDFTrine now only produces BGP plans if there is no get_computed_statement_generators in the query object.
  • Code now materializes all node identities before creating the Bloom filter (so capacity arg is accurate).
  • Bloom filter use is now only attempted on SERVICE blocks that don't immediately contain a FILTER.
  • Re-ordered conditionals so that the service-bloom-filter try block is called less frequently.
  • SERVICE execution now uses non-identity reasoning Bloom filter function.
• Syntax and Serialization
  • Added from_sse method to ::Statement, ::Node.
  • Added initial code for ARQ-style property paths.
  • Added initial parser code for SPARQL Update (SPARUL) extension.
  • Added new 'UNSAID' syntax for SPARQLP, implementing negation.
  • Added parse_expr method to RDF::Query::Parser::SPARQL.
  • Added RDF::Query::Algebra::Quad::bf() method.
  • Fixed RDQL parser to qualify URIs before returning from parse().
  • Fixed SSE serialization of Aggregates using '*' instead of a variable as the column.
  • Removed (now unused) parser generator script.
  • SPARQL parser now always adds a ::Algebra::Project (even when the query selects '*').
  • SPARQL parser now creates ::Algebra::Construct objects for CONSTRUCT queries.
  • SPARQL parser now puts ::Distinct above ::Sort algebras in the parse tree.
  • SPARQLP parser now creates ::Project object as parent of any ::Aggregate.
  • Turtle parser now doesn't modify the lexical value of numeric typed literals.
  • Turtle parser now makes universal IDs for all blank node (even those with given IDs like _:xyz).
  • Updated ::Algebra SSE serializations to conform to Jena's serialization syntax.
  • Updated ::Algebra::Limit::sse to emit 'slice' serialization if its child is a ::Algebra::Offset.
  • Updated as_sparql() methods to support the new ::Construct classes.
  • Updated RDF::Query::sse to emit base and prefix serializations.
  • Updated SPARQL parser and serializer tests to always assume an ::Algebra::Project on SELECT queries.
  • Updated SSE serialization of ::Join::PushDownNestedLoop to use 'bind-join' terminology.
  • Updates to SPARQLP parser to support FeDeRate BINDINGS keyword.
  • ::Algebra::Distinct now does proper serialization in as_sparql().
  • ::GroupGraphPattern::sse() updated to output the '(join ...)' only if the GGP has more than one pattern.
• Optimizer
  • Added benchmark/plans.pl to show the runtimes of the available QEPs for a query.
  • Added benchmark/costmodel.pl for testing the RDF::Query::CostModel.
  • Added logging for execution time (time to construct iterator) of Triples, BGPs, GGPs and sorting.
  • Added logging of cardinalities in ::Algebra::Triple, ::Algebra::BasicGraphPattern and ::Algebra::Service.
  • Added logging to plan classes ::NestedLoop, ::Service, ::Triple.
  • Naive plan is used in ::Federate::Plan::generate_plans only if no optimistic plans are available.
  • Updated cost model code to work with ::Plan classes instead of ::Algebra classes.
  • Logging code now uses sse serialization as log keys (because of expressions that can't be serialized as SPARQL).
  • Logging object can now be passed to RDF::Query constructor.
  • Updated logging of algebra execution to use SPARQL serialization as logging key.
• Miscellaneous
  • SPARQL parser now constructs ::Algebra::Project objects (had been in RDF::Query::execute).
  • Updated RDF::Query to require version 0.108 of RDF::Trine.
  • Added new bloom:filter function variant that doesn't use identity reasoning.
  • Added debugging information when RDFQUERY_THROW_ON_SERVICE is in effect.
  • Fixed test plan for t/optimizer.t in cases where no appropriate model is available.
  • Updated plan generation code to use ::BGPOptimizer when the model supports node_counts.
  • Fixed SSE serialization bug in ::Algebra::Sort.
  • Fixed bugs in RDF::Query and RDF::Query::Expression classes that insisted variables be RDF::Query objects (and not simply RDF::Trine objects).
  • Fixed propogation of iterator binding_names when pre-bound values are used in ::Algebra execution.
  • Fixed RDF::Query::Algebra::Triple to correctly set binding_names when pre-binding is used.
  • Fixed use of pre-binding in execution of RDF::Trine optimized BasicGraphPatterns.
  • Fixed bug in SQL compilation when restricting left-joins to specific node types (based on functions like isIRI).
  • Fixed node identity computation based on owl:sameAs.
  • Fixed bitrotted livejournal example script to work with new 2.000 API.
• Tests
  • Added expected result count test in t/34-servicedescription.t.
  • Added initial tests for algebra subsumes method.
  • Added logging and costmodel tests.
  • Added more example capabilities and patterns to the test service descriptions.
  • Added RDF::Trine::Store::Hexastore to test model construction list in t/models.pl.
  • Added sparql:pattern data to test service descriptions.
  • Added sse re-serialization test to t/29-serialize.t.
  • Added support for sparql:pattern in service description parsing.
  • Added t/plan-rdftrine.t to test QEP generation optimized for RDF::Trine BGPs.
  • Added test data for repeat patterns (e.g. { ?a ?a ?b}).
  • Added tests to t/plan.t for QEP sse serialization.
  • Cleaned up federation tests.
  • Commented out in-progress service description tests.
  • Fixed bug in t/23-model_bridge.t to allow two models from the same model class to be used in testing.
  • Fixed error handling in t/plan.t.
  • Fixed t/costmodel-naive.t to provide a serialized query to ::Plan::Service constructor.
  • Fixed test count in algebra-bgp.t.
  • Fixed test in t/31-service.t that relied on identity-reasoning support in bloom filters (which is now off by default).
  • Fixed use of RDFQUERY_NETWORK_TESTS in 31-service.t.
  • Improved use of temporary RDF::Trine stores in RDF::Query tests.
  • Marked tests TODO for federated query serialization.
  • Removed what looks like an accidentally pasted test in t/plan.t.
  • SERVICE tests involving bloom filter handling marked as TODO.
  • ServiceDescription parsing now gets entire sparql:patterns and not just arcs to depth one.
  • Silenced unfinished test debugging in t/logging.t.
  • Updated args to roqet call in failing_earl_tests.sh.
  • Updated costmodel and logging test expected results that changed due to changes to ::Plan::Join code.
  • Updated dawg-eval.t regex to recognize RDFTrine blank nodes.
  • Updated DBPedia test query in t/34-servicedescription.t to reflect new source data.
  • Updated expected cardinalities in t/logging.t for current plan choice.
  • Updated logging tests to use the new sparql serialization keys.
  • Updated SERVICE test in t/plan.t (still broken, but only runs when dev env var RDFQUERY_NETWORK_TESTS in effect).
  • Updated t/plan.t to be less dependent on the specific state of the kasei.us service endpoint.
  • Updated test service description RDF for new tests.
  • Updates to improve logging and test coverage.
• Examples and Documentation
  • Added examples/query.pl to show a simple example of loading data and executing a query.
  • Added examples/create_query_api.pl for generating queries programatically (based on request from KjetilK).
  • Added bin/graph-bgp.pl to produce a png of a query's BGP variable connectivity graph.
  • Added bin/graph-query.pl to graph the (one chosen) QEP tree.
  • Added bin/graph-qeps.pl to vizualize all QEPs of a query with GraphViz.
  • Updated bin/parse.pl to emit the SSE serialization of the query algebra tree.
  • Updated bin/rdf_parse_turtle.pl to warn on any parser error.

Version 2.002 (2008-04-25)

• Updated Bloom::Filter required version in RDF-Query's Makefile.PL.
• Fixed bug in get_function() when called as a class method.
• Updated sparql:logical-* functions to accept more than two operands.
• Added code to make loading of Bloom::Filter optional.
• Added Bloom::Filter to list of recommended modules.

Version 2.001 (2008-04-19)

• Fixed use of "DESCRIBE " (instead of "DESCRIBE ?var").
• Fixed SPARQL serialization of queries with DISTINCT.
• Added ::Algebra::subpatterns_of_type method for retrieving all subpatterns of a particular type.
• Moved sort_rows() into new ::Algebra classes Sort, Limit, Offset and Distinct.
• Updated SQL compiler to handle new ::Algebra classes.
• Bumped required RDF::Trine version to 0.106.
• Added methods to RDF::Query for retrieving lists of supported SPARQL extensions and extension functions.
• RDF::Trine pattern optimization now holds onto the original BGP object for forwarding of calls to as_sparql().
• Removed use of Storable module. Query execution no longer clones parse tree before running.
• Simplified project operation on query stream in RDF::Query::execute().
• fixup() method in ::Algebra and ::Expression now passed the query object as an argument.
• Replaced ::RDFTrine::unify_bgp with more general fixup() implementation.
• ::Algebra classes now defer to the bridge during fixup() to allow model-specific optimizations.
• RDF::Trine::Iterator::smap now allows overriding default construct_args (e.g. binding names).
• sparql:str now throws a TypeError if argument isn't bound.
• Fixed referenced_variables in RDF::Query::Expression.
• Fixed COUNT function to only count bound variables.
• Fixed aggregation to work with expressions.
• Added support for GROUP BY clauses to aggregation queries.
• Removed now unused ::Algebra::OldFilter class.
• Added serialization tests for aggregate and union patterns.
• Moved as_sparql methods from RDF::Trine:: to RDF::Query:: classes.
• Removed context- (quad-) specific code from RDF::Query::Algebra::Triple.
• Fixed serialization of BOUND filter functions.
• Fixed serialization of unary expressions.
• Fixed call to join_streams to use ::Iterator::Bindings instead of ::Iterator.
• var_or_expr_value now takes bridge object as an argument.
• var_or_expr_value will now accept any RDF::Query::Expression object as an argument.
• Added test for using AS syntax for variable renaming: "(?var AS ?newvar)".
• Added support for MIN, MAX, COUNT and COUNT-DISTINCT aggregate operators to the SPARQLP parser.
• Added COUNT DISTINCT variant aggregate operator.
• Aggregates (MIN, MAX, COUNT, COUNT-DISTINCT) now return xsd:decimal values (this shouldn't really happen for non-numeric operands to MIN and MAX...)
• Added as_sparql submethod to RDF::Query::Node::Literal to serialize numeric types unquoted.
• Added var_or_expr_value method in RDF::Query.
• Removed unused _true and _false methods in RDF::Query.
• Fixed existing aggregates code and tests.
• Removed unused (and bitrotted) t/05-stress.t test.
• Made all tests that access the network opt-in with the RDFQUERY_NETWORK_TESTS environment variable.
• Updated POD for ::Expression::Alias.
• Added tests for select expressions.
• Added initial support for selecting expression values.

Version 2.000 (2008-03-19)

• RDF::Query now uses RDF::Trine package for common classes (nodes, statements, iterators).
• Moved RDF::Query::Stream into RDF::Trine::Iterator namespace.
• Reshuffled SPARQL parsers tests.
• Added support for RDF::Trine::Model.
• RDF::Trine::Iterator can serialize to XML and JSON natively; Updated tests accordingly.
• rdf namespace is only added by default for RDQL.
• Fixed bug where the wrong bridge object would be returned for queries with FROM clauses.
• Moved query_more code to Algebra classes.
• Added RDF::Query::pattern method to return the GGP algebra pattern for the query.
• Added RDF::Query::as_sparql method to serialize query as SPARQL string.
• Removed unused RDF::Query::set_named_graph_query method.
• Fixed bug where triples from a FROM clause were loaded into the underlying persistent store.
• Added POD to RDF::Query::Node classes.
• Added equal method to RDF::Query::Node classes.
• RDF::Query::Node::Blank constructor now optionally produces identifiers.
• Merged tSPARQL parsing code into new SPARQLP parser.
• Added definite_variables method to RDF::Query::Algebra classes.
• Triple class now serializes rdf:type shortcut 'a' appropriately.
• Removed 'VAR' type from ::Node::Variable object structure.
• Updated code to always expect a HASH reference from ::Bindings iterator.
• Added more (sparql and see) serialization tests.
• Removed old fixup code, replaced by ::Algebra fixup methods.
• Moved FROM clause data loading to its own method.
• Started removing old code (RDF::Base, direct DBI use, AUTOLOAD, profiling).
• Moved general count_statements() method into bridge superclass.
• Fixed SQL compiler to work with ::Algebra and ::Node classes.
• Added as_native method to bridge superclass for converting ::Node objects to bridge-native objects.
• Updated document namespace declaration for SPARQL XML Results.
• Added support for SERVICE queries (ala ARQ) with Bloom filter semijoins.
• Moved Expression classes out of the Algebra hierarchy and into their own space (RDF::Query::Expression).

Version 1.501 (2007-11-15)

• Fixed CONSTRUCT+OPTIONAL bug.
• Added as_sparql methods to Algebra and Node classes to serialize patterns in SPARQL syntax.
• Added deparse script.
• Fixed jena:sha1sum tests when Digest::SHA1 isn't available.

Version 1.500 (2007-11-13)

• Query Engine
  • URIs are now properly qualified with a BASE when present.
  • Base URI passed to constructor is now used if no BASE clause exists in the query.
  • Fixed BASE qualification when an IRI contains Unicode characters (Emulating IRI support with the URI module).
  • NAMED graph data is now seperated from the default model into new (temporary) models.
  • NAMED graphs now work with RDF::Core.
  • Added new RDF::Query::Algebra:: classes that are used to represent query ASTs.
  • Added new RDF::Query::Node:: classes that are used to represent RDF Nodes and Variables.
  • Major refactoring to RDF::Query::query_more() to enhance extensibility.
  • Added RDF::Query::query_more_triple() and RDF::Query::query_more_bgp() for triple and bgp matching.
  • Improved support of GGP pattern matching.
  • Added sgrep, smap, swatch and concat methods to RDF::Query::Stream class.
  • Refactored query_more() variants and sort_rows() to use new stream methods sgrep, smap, and concat.
  • Continued to fix bugs to more closely align with DAWG tests.
  • Updated DAWG tests to run with the RDF::Core backend.
  • Any DAWG tests with mf:requires are now automatically marked TODO.
  • DAWG tests graph equality is now punted to user verification.
  • Fixed bNode merging code in DAWG tests.
  • query_more() variants and sort_rows() now all return RDF::Query::Stream objects.
  • query_more() (and everything it calls) now expects bridge object as a method argument (to ensure NAMED graph querying works).
  • Added join_streams() to perform netsted-loop natural joins on two Stream objects.
• Filters
  • Added call_function() method to abstract the generation of ['FUNCTION',...] blocks.
  • FILTER operator != is now negative for unknown datatypes.
  • Fixed exception handling in check_constraints().
  • Fixed type-promotion in arithmetic operations and added recognized xsd numeric types.
  • API Chnage: extension functions now take a bridge object as their second argument (after the query object).
  • Fixed equals() method in RDF::Core bridge to properly use the RDF::Core API.
  • Javascript function makeTerm now accepts language and datatype parameters.
  • toString() javascript funtion now returns just the literal value (not including language or datatype).
  • sop:str now will stringify blank nodes properly.
  • sparql:langmatches now properly differentiates between no language tag and an empty language tag.
• Parsers
  • Parsers now generate ASTs using the Algebra and Node classes.
  • Fixed bugs in SPARQL tokenizer for some Unicode strings (with combining accents).
  • RDF::Query::Parser::new_literal() now canonicalizes language tags (to lowercase).
  • Fixed GGP verification in RDF::Query::Parser::SPARQL::fixup().
  • Merged GGPAtom changes from tSPARQL to SPARQL grammar.
• Backends
  • Fixed bug in RDF::Query::Model::RDFCore::equals() when comparing two blank nodes.
  • Changed RDF::Query::Model::RDFCore::as_string to return strings where node type is identifiable (like Redland, URIs as [...], literal \"...\", bnodes (...)).
  • Model methods literal_value, literal_datatype and literal_value_langauge now return undef if argument isn't a literal.
  • Model methods uri_value and blank_identifier now return undef unless argument is of correct type.
  • Model add_string methods now normalize Unicode input.
  • Blank node prefix is now scoped to bridge, not lexically in RDF::Query::Model::RDFCore.
  • RDF::Query::Model::Redland::new_literal now forces argument to a string (XS code breaks on non-PVOK scalars).
  • RDF::Query::Model::Redland::add_uri now uses LWP instead of relying on Redland to fetch content.
  • Redland model class now recognizes DateTime objects as literals.
  • Updated add_uri() method in RDF::Core bridge to support named graphs (only one name per model for now).
• Miscellaneous
  • Added rudimentary profiling code (Devel::DProf seems to crash on RDF::Query)
  • Added initial code for supporting property functions (using time:inside as an example).
  • Removed multi_get code.
  • Removed code for MULTI patterns. Will replace with algebra-based optimization in the future.
  • RDF::Query::Stream constructor now accepts an ARRAY reference.
  • Stopped using the peephole optimizers. Will replace with an algebra-based optimizer in the future.

Version 1.042 (2007-06-13)

• Added support for Javascript URL-based extension functions.
• Added GPG signing support for Javascript functions.
• Added meta methods to model classes.
• Added count_statements methods to model classes.
• Added default format for stringifying query result streams.
• Added SPARQL syntax coverage tests.
• Added stream tests.
• ISIRI() now returns a sop:isIRI function IRI.
• Redland bridge add_uri strips off URI fragment from parser base URI.
• Underlying models aren't loaded if turned off by environment variables.
• Stopped using the (currently broken) multi-get code.
• Model classes aren't used if turned off by environment variable.
• Standardized node testing to is_node (replacing older isa_node).
• GPG verification failure now throws exceptions.
• GPG fingerprint handling is now whitespace insensitive.
• Removed (currently) unused multiget tests.
• Redland bridge uri_value returns undef if not called with a node.

Version 1.041 (2006-11-26)

• Removed unwanted '+' signs on stringified bigints when running under perl 5.6.x.
• Fixed unicode errors when running under perl 5.6.x.
• Added tests for model bridge classes.
• Fixed bugs in SQL bridge class.
• RDFCore and Redland bridge classes now throw error when passed a bad model object.
• Bridge class support() method now responds to feature requests for 'temp_model'.
• Fixed bug in RDF::Query::Model::RDFCore::isa_resource returning true for blank nodes.
• Fixed code in RDF::Query::Model::::SQL::equals().
• Added partial support for new RDF::Storage::DBI storage class.
• Added support for RDF::Query:::Model::SQL models in RDF::Query bridge code.
• Removed RDF::Query::Model::SQL::* code that's now in RDF::Storage::DBI.
• Added tests for backend bridge classes (RDF::Query::Model::*).
• Added checks for which backends support XML serialization.
• Fixed naive peephole tests in cases where model supports cost-based analysis.
• Fixed bugs in tests that were relying on C to act like C.
• Added RDF::Base support.
• Fixed bug in C that prevented running queries against multiple models.
• Added SimpleQueryPatternError exception class for future work on multi-get optimizations.
• Removed forced dependence on Digest::SHA1.
• Makefile.PL now recommends Digest::SHA1 (for jena function) and Geo::Distance (for 07-filters.t).

Version 1.040 (2006-07-21)

• Added support for BOUND() fiters in SQL compilation.
• SQL compiler now produces valid SQL when query variable names are SQL reserved keywords.
• Moved SPARQL parser test data into YAML.
• Added YAML as a build-time prerequisite.
• Fixed SPARQL parsing bug for blank nodes as statement objects.
• Added peephole optimizer code with naive and cost-analysis strategies.
• Added add_string method to RDF::Query::Model::Redland.
• Added node_count method to RDF::Query::Model::Redland (only used for testing at this point).
• RDF::Query::execute() now uses the peephole optimizer.
• Removed punting code in RDF::Query::execute that tried to do JIT optimization.
• Removed calls to getLabel() on model objects in test files.
• Fixed dawg tests (was dying because of multiple plans being printed).
• Fixed cost-based peephole optimizer tests (by forcing Redland to do node counting).

Version 1.039 (2006-07-14)

• Removed dawg tests from distribution. Only used as developer tests now.
• Updated package to use Module::Install instead of ExtUtils::MakeMaker.
• Fixed a spurious uninitialized warning in RDF::Query::get_bridge.

Version 1.038 (2006-07-09)

• Fixed DBI case-sensitivity for PostgreSQL support.
• Cleaned up SQL syntax for easier debugging.
• Removed extra parens in SQL that were causing postgresql to break.
• Reference test foaf file using File::Spec->catfile() instead of directly.
• Fixed SPARQL parsing bug for anonymous nodes in FILTER expressions.
• Fixed major SQL compilation bugs for testing equality in FILTER expressions.
• Fixed bug in hashing code for blank nodes in SQL compiler.

Version 1.037 (2006-07-06)

• execute() method now takes optional 'bind' parameter for pre-binding query variables.
• Updated code to support basic FILTERs in SQL compilation.
• Fixed bug in SQL compilation where no WHERE clause was needed.
• Fixed bug in SQL compilation when using model-specific Statements tables.
• Added Storable to the list of required modules (was missing in the list).
• Fixed typos in metadata about past versions in DOAP Changes.ttl.

Version 1.036 (2006-06-26)

• Fixed memory leak in RDF::Query::Stream that resulted in too many database handles.
• Initial support for OPTIONALs in SQL compiler.
• Removed LWP requirement for systems without libwww.
• Added support for class variable to hold parsing errors. (Beware under threads.)
• RDF::Query now sets error variable upon parsing error. (Access with error().)
• Fixed POD errors, and added tests for POD coverage.
• Added model API methods to SQL model class.
• Added close() method to RDF::Query::Stream.

Version 1.035 (2006-06-04)

• Added DAWG tests and harness.
• Rewrote core logic in OPTIONAL handling code.
• Comparisons on literals now respect numeric datatypes.
• Fixed outdated calling conventions in casting functions.
• Added custom functions:
  • jena:sha1sum
  • jena:now
  • jena:langeq
  • jena:listMember
  • ldodds:Distance
• Added new model methods: equals, subject, predicate, object.
• Relocated external http-based test data to .Mac URLs.

Version 1.034 (2006-05-01)

• Added JSON serialization for bindings and boolean queries.
• Initial support for compiling RDF queries to SQL queries using the Redland schema.
• Added to_string method to query results Stream class.
• Model objects now store the query parse tree for access to data needed in serialization.
• Unquoted number and boolean literals in SPARQL are now datatyped appropriately.
• Fixed crashing bug when RDF::Query::Model::Redland::as_string was called with an undefined value.
• Fixed bug parsing queries with predicate starting with 'a' (confused with { ?subj a ?type}).
• Fixed bug parsing queries whose triple pattern ended with the optional dot.

Version 1.033 (2006-03-08)

• Updated test suite to work if one of the model classes is missing.
• Data-typed literals are now cast appropriately when used in a FILTER.
• Added support for xsd:dateTime datatypes using the DateTime module.
• Added support for LANG(), LANGMATCHES() and DATATYPE() built-in functions.
• Updated TODO list.
• Added more exception types to RDF::Query::Error.
• Added POD coverage.
• Fixed SPARQL parsing bug for logical operators <= and >=.

Version 1.032 (2006-03-03)

• Replaced the Parse::RecDescent SPARQL parser with a much faster hand-written one.
• Updated SPARQL parsing rules to be better about URI and QName character sets.
• FILTER equality operator now '=', not '==' (to match SPARQL spec).
• Initial support for FILTER constraints as part of the triple pattern structure (Will allow for nested FILTERs).
• Implemented support for ordering query results by an expression.
• Fixed bug in expresion handling of unary minus.
• Fixed bug in Redland NAMED GRAPH parsing.
• Fixed bug in RDF::Core parsing code where blank nodes would be accidentally smushed.

Version 1.031 (2006-02-08)

• Added support for NAMED graphs.

Version 1.030 (2006-01-13)

• Added support for SELECT * in SPARQL queries.
• Added support for default namespaces in SPARQL queries.
• Added support for querying RDF collections in SPARQL: (1 ?x 3)
• Fixed a bug where one of two identical triple variables would be ignored ({ ?a ?a ?b }).

Version 1.028 (2005-11-18)

• Added SPARQL functions: BOUND, isURI, isBLANK, isLITERAL.
• Updated SPARQL REGEX syntax.
• Updated SPARQL FILTER syntax.
• Added SPARQL RDF Collections syntactic forms.
• Fixed FILTER support in OPTIONAL queries.
• Added binding_value_by_name method to Query results stream class.
• Added isa_blank methods to RDF::Redland and RDF::Core model classes.
• Fixed RDF literal datatyping when using Redland versions >= 1.00_02.
• Updated SPARQL grammar to make 'WHERE' token optional.
• Added directives to SPARQL grammar.
• Updated SPARQL 'ORDER BY' syntax to use parenthesis.
• Fixed SPARQL FILTER logical-and support for more than two operands.
• Fixed SPARQL FILTER equality operator syntax to use '=' instead of '=='.
• Now requires Test::More 0.52 because of changes to is_deeply().

Version 1.027 (2005-07-28)

• Updated to follow SPARQL Draft 2005.07.21:
• ORDER BY arguments now use parenthesis.
• SPARQL parser now supports ORDER BY operands: variable, expression, or function call.
• Added binding_value_by_name() method to query result streams.

Version 1.026 (2005-06-05)

• Added new DBI model bridge (accesses Redland's mysql storage directly).
• Added built-in SPARQL functions and operators (not connected to grammar yet).
• Added bridge methods for accessing typed literal information.

Version 1.024 (2005-06-02)

• Added missing SPARQL OFFSET grammar rules.
• Added XML Results support for graph and boolean queries (DESCRIBE, CONSTRUCT, ASK).
• Fixed major bugs in RDF::Core bridge:
  • Bridge wasn't using passed model.
  • Literal construction was broken.
  • Blank node construction was broken when no identifier was specified.
• Stream::bindings_count now returns the right number even if there is no data.
• XML Result format now works with RDF::Core models.
• The Model bridge object is now passed to the Stream constructor.
• Internal code now uses the variables method.
• Removed redundant code from ORDER BY/LIMIT/OFFSET handling.
• Removed unused count method.
• Removed unused delegating AUTOLOAD.
• Removed unused parse_files method.
• Removed usused add_file method.
• Removed duplicate net test file.
• Added test file for local file-based SPARQL 'FROM' queries.
• Added test file for SPARQL Result Forms (LIMIT, ORDER BY, OFFSET, DISTINCT).
• Added test file for SPARQL Protocol for RDF (XML Results).
• Added new tests based on Devel::Cover results.
• Some test files now run against both Redland and RDF::Core:
  • 00-simple.t
  • 03-coverage.t
  • 10-sparql_protocol.t
• All debugging is now centrally located in the _debug method.
• Moved Stream class to lib/RDF/Query/Stream.pm.
• Fixed tests that broke with previous fix to CONSTRUCT queries.
• Fixed tests that broke with previous change to ASK query results.

Version 1.021 (2005-06-01)

• Added SPARQL UNION support.
• Broke OPTIONAL handling code off into a seperate method.
• Added new debugging code to trace errors in the twisty web of closures.

Version 1.020 (2005-05-18)

• Added support for SPARQL OPTIONAL graph patterns.
• Calling bindings_count on a stream now returns 0 with no data.
• Added Stream methods:
  • is_bindings
  • binding_name
  • binding_values
  • binding_names
• Added as_xml method to Stream class for XML Binding Results format.

Version 1.016 (2005-05-08)

• Added initial support for SPARQL ASK, DESCRIBE and CONSTRUCT queries.
  • Added new test files for new query types.
• Added methods to bridge classes for creating statements and blank nodes.
• Added as_string method to bridge classes for getting string versions of nodes.
• Broke out triple fixup code into fixup_triple_bridge_variables().
• Updated FILTER test to use new Geo::Distance API.

Version 1.015 (2005-05-03)

• Fixes to the arguments passed to FILTERs.
• Filter tests now show example of two custom filters:
  • Transitive subClassOf testing.
  • Logical comparison operators (range testing lat/lon values).
• Added literal_value, uri_value, and blank_identifier methods to bridges.
• Redland bridge now calls sources/arcs/targets when only one field is missing.
• Fixes to stream code. Iterators are now destroyed in a timely manner.
  • Complex queries no longer max out mysql connections under Redland.
• Cleaned up node sorting code.
  • Removed dependency on Sort::Naturally.
  • Added new node sorting function ncmp().
• check_constraints now calls ncmp() for logical comparisons.
• Added get_value method to make bridge calls and return a scalar value.
• Fixed node creation in Redland bridge.
• Moved DISTINCT handling code to occur before LIMITing.
• Added variables method to retrieve bound variable names.
• Added binding_count and get_all methods to streams.
• get_statments bridge methods now return RDF::Query::Stream objects.

Version 1.014 (2005-04-26)

• Made FILTERs work in SPARQL.
• Added initial SPARQL support for custom function constraints.
• SPARQL variables may now begin with the '$' sigil.
• Added direction support for ORDER BY (ascending/descending).
• Added 'next', 'current', and 'end' to Stream API.

Version 1.012 (2005-04-24)

• Stream objects now handle being constructed with an undef coderef.
• Streams are now objects usinig the Redland QueryResult API.
• RDF namespace is now always available in queries.
• row() now uses a stream when calling execute().
• Added ORDER BY support to RDQL parser.
• SPARQL constraints now properly use the 'FILTER' keyword.
• SPARQL constraints can now use '&&' as an operator.
• SPARQL namespace declaration is now optional.
• Updated tests.

Version 1.010 (2005-04-21)

• execute now returns an iterator
• Added core support for DISTINCT, LIMIT, OFFSET
• Added initial core support for ORDER BY (only works on one column right now)
• Broke out the query parser into it's own RDQL class
• Added initial support for a SPARQL parser
  • Added support for blank nodes
  • Added lots of syntactic sugar (with blank nodes, multiple predicates and objects)
  • Added SPARQL support for DISTINCT and ORDER BY
• Moved model-specific code into RDF::Query::Model::*
• Moving over to redland's query API (pass in the model when query is executed)