Release 3.1.8

• Don't use native types for invalid literals.
• Disallow @reverse within an embedded node (flattening can't handle it).
• Update flatten algorithm to handle embedded nodes and annotation objects.
  • Perform bnode renaming before creating the node map to handle embedded nodes. Adds test option to do result comparison by remapping bnode labels based on dataset bijection.
• Make renaming bnodes default to false for Reader and to_rdf_spec.
• Fix expansion bug where a property is @reverse and the same subject has an explicit @reverse.
• Don't compact an @id or @value to a string if it contains an annotation.
• Make sure list members don't have annotations.

Release 3.1.7

• JSON-LD* compaction. Also fixes writer error where prefixes are supplied.
• Fix bug where container could be returned as an Array and not a Set.

Release 3.1.6

• Preliminary support for RDF* parsing using the embedded triple mode overloading @id (JSON-LD*).
  • Add --rdfstar option to script/parse.

Release 3.1.5

• Improve documentation.
• When expanding, check that value objects have at most a single string value for @type, unless framing.
• Use unlicense for doap:license.
• Fix option handling in jsonld CLI.
• Update CLI options with base coming from base_uri, and local frame/context files opened.
  • For arguments which are URLs, if it is not an absolute URI, read it into a StringIO.

Release 3.1.4

• Performance Improvement
  • Reduce object duplication in context processing.
  • Keep document base out of context, and pass in through other algorithms. This allows for improved caching of initial contexts.
• Streaming RDF reader
  • Non-event based, but conforms to the JSON-LD 1.1 Streaming note.
  • Invoke using JSON::LD::Reader with the :stream option set to true.

Release 3.1.3

• Don't overload context PRELOADED with local cache.
• Parse contexts without scope, and merge into active context.
• Improve context merging to look for term protection overrides.
• Use nil for default value of @protected rather than false, which makes merge logic work better so that protected:true does not override protected:false.
• Don't cache base from remote context.

Release 3.1.2

• Remove support for empty terms, as they must not be an empty string in 1.1 (1.0, too really).
• Add tests for keyword redefinition of @type has an empty map as a value.
• Allow aliases of @nest to be expanded term definitions, as well as simple.
• Missing script element should be LoadingDocumentFailed not InvalidScriptElement.
• Loading an HTML with no script element should be an error, not an empty document.
• Fix HTML libraries to deal with loading when both have been required by being explicit about which library to use.
• Use validate_scoped option to context and term creation.
• Fix bug in merge_node_map_graphs for @type.
• Update message text for "context overflow" exception.

Release 3.1.1

• Accept application/xhtml+xml.
• Pass option to expand_iri to return non-nil results as string, rather than just doing #to_s.
• Make sure Datatype API is valid, not just absolute.
• Default for requireAll is false, not true.
• Normalize language tags to lower case when creating an i18n datatype or compound literal.
• Resolve expandContext in test suite relative to the manifest location.
• Fix some CLI option descriptions. Specifically, --[no-]requireAll is changed to --[no-]require-all.
• Fix bug loading a default script tag if a profile tag is missing.
• Attempt to load json-ld-preloaded gem with silent fallback.

Release 3.1.0

• Support for [JSON-LD 1.1 API][https://www.w3.org/TR/2019/CR-json-ld11-api-20191212/] and JSON-LD Framing Candidate Releases.
  • Scoped Contexts
  • Nested Properties
  • JSON Literals
  • Lists of Lists
  • String direction
  • Indexed Values
  • Imported Contexts
  • Protected Term Definitions
• Requires RDF.rb 3.1 and supports Ruby 2.7 (versions < 2.4 now unsupported)

Release 3.0.0

Sync with updates from JSON-LD 1.1 Working Group

• Updates maps and term assignment to use @none.
• Improve inverse context creation and compact_iri to use concatenated containers.
• Change prioritization for values not matching a container to be after @none.
• Don't force contents of @graph into an array.
• Added support for omitGraph for framing, defaulting to true if processingMode is json-ld-1.1
• The omitDefault flag now defaults to false, and is indepdendent of processing mode.
• Change to use "@vocab": "" semantics, and vocabulary expansion/compaction just as if the effective document base was specified as @vocab.
• Sort @type when looking for scoped contexts.
• Make sure multiple objects with @container: @graph are maintained as multiple graphs.
• Simplify processing mode conflicts to allow mixing different types of contexts.
• Support for lists of lists.
• Raise error if too many recursive contexts loaded.
• Have toRdf emit only valid triples, excluding anything invalid. Literal validity is based on the value being a string, language having the proper form, and datatype IRI being valid.