Release 3.3.0

• Update documentation links to use gh-pages, and add action to publish gh-pages from Yard docs.
• Rspec-mocks recieve.with become more sensitive to hash vs keyword arguments for Ruby > 3.
• Minor change to queryable.
• Add specs for Enumerable#canonicalize.
• Use inferred format when dumping, if not defined in info.
• Updates for version 3.3.0 with minimum Ruby version 3.0.

Release 3.2.0

• Minimum Ruby version now 2.6.
• Replace dependency on awesome_print with amazing_print.
• Improve Literal#inspect to show language and built-in datatype.
• Fix Struct initialization for upcoming change to Ruby after 3.1.

Release 3.1.2

• Update log formatting for constant-width severity, which makes tracing indented logs simpler.
• Update PDD info in the README.
• Allow other options to be used when serializing results in be_equivalent_graph along with base and prefixes, which are passed to dump.
• Add development runtime on rexml, which seems to be needed for ruby-head and is used for webmock/crack.
• Run CI on GitHub.

Release 3.1.1

• Update DOAP and related files in /etc.
• Normalize URLs to HTTPS, where possible.
• Change doap:license to https://unlicense.org/1.0/.

Release 3.1.0

• Remove support for Ruby < 2.4.
• Make sure buffer encoding set using format or passed option.
• Fix transaction check and enable local running of spec.
• Rename instance variables used in specs to be localized so as not to confuse with the same variable that could be named in the surrounding context.
• Allow readers to define more default prefixes.
• Allow dump format to be specified in be_equivalent_graph.
• Add check for default and user-specified User-Agent.
• Fix write matcher to handle exceptions in block.
• Don't override Array.inspect, and use awesome_print in produce matcher.
• Be explicit when argument is a hash and not keyword arguments.

Release 3.0.0

Changes since 2.2

• Add failure_message_when_negated for be_equivalent_graph.
• Don't run reader or writer open tests over all formats, just that being tested.
• Verify that property Accept header is sent when a specific reader is determined in Reader.open.
• Change validation on Boolean, to reflect changes in RDF.rb.

Release 2.2.0

• Updates for RDF.rb 2.2.
• Update minumum ruby version to 2.2.2. See ruby-rdf/rdf#307.
• Tests Transaction#mutate?
  • We deliberately avoid checking for false positives, since the interface makes no guarantee that they won't appear.
• Generalize protected method check for Queryable
• Remove false as default assumption in Queryable specs
  • This assumption has only held for RDF::Repository in the past. The test should be replaced in either Repository or Dataset.
• Remove duplicate Transactable tests from Repository
• Isolate Transactable#transaction test from blank nodes
  *This test used to implictly test stablity of blank node identifers across multiple #statements calls. By using URIs instead of blank nodes, we avoid failures when they change.
• Filter for #supports?(:literal_equality)
• Use #to_h instead of #to_hash. Test for DEPRECATION on uses of #to_hash. This is associated with ruby-rdf/rdf#333.

Pre-release 2.2.0-rc1

• Updates for RDF.rb 2.0.
• Update minumum ruby version to 2.2.2. See ruby-rdf/rdf#307.
• Tests Transaction#mutate?
  • We deliberately avoid checking for false positives, since the interface makes no guarantee that they won't appear.
• Generalize protected method check for Queryable
• Remove false as default assumption in Queryable specs
  • This assumption has only held for RDF::Repository in the past. The test should be replaced in either Repository or Dataset.
• Remove duplicate Transactable tests from Repository
• Isolate Transactable#transaction test from blank nodes
  *This test used to implictly test stablity of blank node identifers across multiple #statements calls. By using URIs instead of blank nodes, we avoid failures when they change.
• Filter for #supports?(:literal_equality)
• Use #to_h instead of #to_hash. Test for DEPRECATION on uses of #to_hash. This is associated with ruby-rdf/rdf#333.

Release 2.0.0

Updates for RDF.rb 2.0

• Specs for Queryable#query tested internal implementation rather than behavior, this caused issues for classes mixing in Queryable, but delegating #query to some collaborator. See, e.g. ActiveTriples/ActiveTriples#169. This tests the contract directly, instead.
• Add tests for Transaction, Changeset, and Dataset
• Refactor Queryable specs
  • #33 created some duplicate tests; this refactors by moving them into a shared example set.
  • Additionally, it skips specs for #query_pattern and #query_execute if they are not defined by the Queryable. This allows implementations to fulfill the #query interface directly, if desired.
• Disable an enumerable test unless the subject supports named graphs. (Moved some other checks inside the spec, where it's defined).
• In Mutable, when doing deletes with patterns, make sure to not use BNodes, which might not be supported by repository implementation.
• Add RDF::Spec.logger, which creates a Logger instance using a StringIO. Also implements #clear and #to_s instance methods.
• Added be_equivalent_graph and produce matchers.
• Simplify Enumerable#has_statement? spec as both Repositories and Graphs should have a statement, either matching on graph_name, or ignoring it. Previously, if a Graph had a bound graph_name it would need to match that in the statement, but this is inconsistent with the notion that a graph has only triples.
• Change Queryable#first_* specs to not depend on query result ordering.
• Add a round-trip check for Writers also having a Reader to ensure that different BNodes having the same identifier get written out with unique identifiers (see ruby-rdf/rdf/issues/262).
• Soften RDF::Transaction buffering requirements
• Not every Transaction needs to keep an up-to-date Changeset in buffer. This change loosens the requirements. some of the tests are removed to the rdf test suite in spec/transaction_spec.rb.
• Introduce RDF::Dataset as a distinct concept
  • Dataset is introduced as a base class of RDF::Repository. It represents a (default read only) set of graphs. Some of the specs are removed from the repository specs, but are re-included indirectly through dataset's tests.
• Test atomic/transactional #delete_insert application
  • When a Mutable #supports?(:transactions), #delete_insert is guaranteed to update atomically.
  • Repository#delete_insert always applies within a transaction scope and conforms to the individual Repository`s transaction semantics.
• Check durability by re-creating the repository using uri and verifying that when subject is updated, the new instance is also updated. Fixes #52.
• Skip mutable and writable tests unless resource is mutable or writable. This is better than testing separately in each test.
• Avoid changing Repository while tx is in progress
  • Adding or deleting statements in the Repository causes serializable transactions to break (correctly). Instead, prepare the repository state before beginning the test transaction.
• Test isolation semantics for transactions.
• Fix array inspector when values are hashes.
• Allow formats to not have readers (vocabulary is a write-only format).

2.0 Beta Release

Updates for RDF.rb 2.0 Beta

• Specs for Queryable#query tested internal implementation rather than behavior, this caused issues for classes mixing in Queryable, but delegating #query to some collaborator. See, e.g. ActiveTriples/ActiveTriples#169. This tests the contract directly, instead.
• Add tests for Transaction, Changeset, and Dataset
• Refactor Queryable specs
  • #33 created some duplicate tests; this refactors by moving them into a shared example set.
  • Additionally, it skips specs for #query_pattern and #query_execute if they are not defined by the Queryable. This allows implementations to fulfill the #query interface directly, if desired.
• Disable an enumerable test unless the subject supports named graphs. (Moved some other checks inside the spec, where it's defined).
• In Mutable, when doing deletes with patterns, make sure to not use BNodes, which might not be supported by repository implementation.
• Add RDF::Spec.logger, which creates a Logger instance using a StringIO. Also implements #clear and #to_s instance methods.
• Added be_equivalent_graph and produce matchers.
• Simplify Enumerable#has_statement? spec as both Repositories and Graphs should have a statement, either matching on graph_name, or ignoring it. Previously, if a Graph had a bound graph_name it would need to match that in the statement, but this is inconsistent with the notion that a graph has only triples.
• Change Queryable#first_* specs to not depend on query result ordering.
• Add a round-trip check for Writers also having a Reader to ensure that different BNodes having the same identifier get written out with unique identifiers (see ruby-rdf/rdf/issues/262).
• Soften RDF::Transaction buffering requirements
• Not every Transaction needs to keep an up-to-date Changeset in buffer. This change loosens the requirements. some of the tests are removed to the rdf test suite in spec/transaction_spec.rb.
• Introduce RDF::Dataset as a distinct concept
  • Dataset is introduced as a base class of RDF::Repository. It represents a (default read only) set of graphs. Some of the specs are removed from the repository specs, but are re-included indirectly through dataset's tests.
• Test atomic/transactional #delete_insert application
  • When a Mutable #supports?(:transactions), #delete_insert is guaranteed to update atomically.
  • Repository#delete_insert always applies within a transaction scope and conforms to the individual Repository`s transaction semantics.
• Check durability by re-creating the repository using uri and verifying that when subject is updated, the new instance is also updated. Fixes #52.
• Skip mutable and writable tests unless resource is mutable or writable. This is better than testing separately in each test.
• Avoid changing Repository while tx is in progress
  • Adding or deleting statements in the Repository causes serializable transactions to break (correctly). Instead, prepare the repository state before beginning the test transaction.
• Test isolation semantics for transactions.