RDF.ex v0.10.0

Changelog View Source

All notable changes to this project will be documented in this file. This project adheres to Semantic Versioning and Keep a CHANGELOG.

0.10.0 - 2021-12-13

This release adds RDF-star support on the RDF data structures, the N-Triples, N-Quads, Turtle encoders and decoders and the BGP query engine. For an introduction read the new page on the RDF.ex guide. For more details on how to migrate from an earlier version read this wiki page.

Elixir versions < 1.10 are no longer supported

Added

Changed

Compare v0.9.4...v0.10.0

0.9.4 - 2021-05-26

Added

  • RDF.statement/1, RDF.statement/3 and RDF.statement/4 constructor functions
  • the :default_prefixes configuration option now allows to set a {mod, fun} tuple, with a function which should be called to determine the default prefixes
  • the :default_base_iri configuration option now allows to set a {mod, fun} tuple, with a function which should be called to determine the default base IRI
  • support for Elixir 1.12 and OTP 24

Fixed

  • the Turtle encoder was encoding IRIs as prefixed names even when they were resulting in non-conform prefixed names
  • the Turtle encoder didn't properly escape special characters in language-tagged literals
  • the N-Triples and N-Quads encoders didn't properly escape special characters in both language-tagged and plain literals
  • the Inspect protocol implementation for RDF.Diff was causing an error when both graphs had prefixes defined

Note: In the canonical form of -0.0 in XSD doubles and floats the sign is removed in OTP versions < 24 although this is not standard conform. This has the consequence that the sign of -0.0 is also removed when casting these doubles and floats to decimals. These bugs won't be fixed. If you rely on the correct behavior in these cases, you'll have to upgrade to OTP 24 and a respective Elixir version.

Compare v0.9.3...v0.9.4

0.9.3 - 2021-03-09

Added

  • :indent option on RDF.Turtle.Encoder, which allows to specify the number of spaces the output should be indented

Changed

  • the performance of the Enumerable protocol implementations of the RDF data structures was significantly improved (for graphs almost 10x), which in turn increases the performance of all functions built on top of that, eg. the N-Triples and N-Quads encoders
  • improvement of the Inspect forms of the RDF data structures: the content is now enclosed in angle brackets and indented

Fixed

  • strings of the form ".0" and "0." weren't recognized as valid XSD float
    and double literals
  • the Turtle encoder handles base URIs without a trailing slash or hash properly
    (no longer raising a warning and ignoring them)

Compare v0.9.2...v0.9.3

0.9.2 - 2021-01-06

Added

Changed

  • a new option :as_value to enforce interpretation of an input string as a value instead of a lexical, which is needed on datatypes where the lexical space and the value space both consist of strings
  • RDF.XSD.Date and RDF.XSD.Time both can now be initialized with tuples of an Elixir Date resp. Time value and a timezone string (previously XSD date and time values with time zones could only be created from strings)

Compare v0.9.1...v0.9.2

0.9.1 - 2020-11-16

Elixir versions < 1.9 are no longer supported

Added

  • general serialization functions for reading from and writing to streams and implementations for N-Triples and N-Quads (Turtle still to come)
  • a :gzip option flag on all read_file/3 and write_file/3 functions allows to read and write all supported serialization formats from and to gzipped files (works also with the new possibility to read and write files via streams)
  • RDF.Dataset.prefixes/1 for getting an aggregated RDF.PrefixMap over all graphs
  • RDF.PrefixMap.put/3 for adding a prefix mapping and overwrite an existing one
  • RDF.BlankNode.value/1 for getting the internal string representation of a blank node
  • RDF.IRI.in_namespace?/2 for determining whether an IRI lies in a namespace

Changed

  • all read_file/3 and write_file/3 functions on RDF.Serialization and the modules of RDF serialization formats can use streaming via the :stream flag option; for read_file/3 and write_file/3 it defaults to false, while for read_file!/3 and write_file!/3 it defaults to true when the respective format supports streams
  • the Inspect form of the RDF data structures are now Turtle-based and respect the usual :limit behaviour
  • more compact Inspect form for RDF.PrefixMap
  • the RDF.Turtle.Encoder accepts RDF.Vocabulary.Namespace modules as base
  • the performance of the RDF.Turtle.Encoder was improved (by using a for most use cases more efficient method for resolving IRIs to prefixed names)
  • RDF.BlankNode.new/0 creates integer-based blank nodes, which is much more efficient in terms of performance and memory consumption than the previous ref-based blank nodes

Fixed

  • RDF.BlankNodes based on refs weren't serializable to Turtle
  • RDF.Vocabulary.Namespaces couldn't contain terms conflicting with functions from Elixirs Kernel module; most of them are supported now, while for the
    remaining unsupported ones a proper error message is produced during compilation

Compare v0.9.0...v0.9.1

0.9.0 - 2020-10-13

The API of the all three RDF datastructures RDF.Dataset, RDF.Graph and RDF.Description were changed, so that the functions taking input data consist only of one field in order to open the possibility of introducing options on these functions. The supported ways with which RDF statements can be passed to the RDF data structures were extended and unified to be supported across all functions accepting input data. This includes also the way in which patterns for BGP queries are specified. Also, the performance for adding data has been improved.

For an introduction on the new data structure API and the commonly supported input formats read the updated page on the RDF data structures in the guide. For more details on how to migrate from an earlier version read this wiki page.

Added

Changed

  • the format for the specification of BGP queries with RDF.Graph.query/2, RDF.Graph.query_stream/2 and RDF.Query.bgp/1 has been changed to be consistent
    with the supported formats for input data in the rest of the library
  • RDF.Description.new now requires the subject to be passed always as first argument; if you want to add some initial data this must be done with the :init option
  • The put/3 functions on RDF.Graph and RDF.Dataset now overwrite all statements with same subject. Previously only statements with the same subject AND predicate were overwritten, which was probably not the expected behaviour, since it's not inline with the common put semantics in Elixir. A function with the previous behaviour was added on RDF.Graph and RDF.Dataset with the put_properties/3 function.
    • CAUTION: This means the RDF.Graph.put/2 and RDF.Dataset.put/2 function have become more destructive now when not specified otherwise.
    • Note: Although one could argue, that following this route RDF.Dataset.put/3 would consequently have to overwrite whole graphs, this was not implemented for practical reasons. It's probably not what's wanted in most cases.
  • The Access protocol implementation of get_and_update/3 on RDF.Graph and RDF.Dataset previously relied on the put/2 functions with the old behaviour of overwriting only statements with the same subject and predicate, which was almost never the expected behaviour. This is fixed now by relying on the new put/2 behaviour.
  • the values/2 functions of RDF.Statement, RDF.Triple, RDF.Quad, RDF.Description, RDF.Graph and RDF.Dataset now accept on their second argument an optional RDF.PropertyMapwhich will be used to map predicates accordingly; the variant of these values/2 functions to provide a custom mapping function was extracted into a new function map/2 on all of these modules
  • for consistency reasons the internal :id struct field of RDF.BlankNode was renamed to :value
  • allow the base_iri of RDF.Vocabulary.Namespaces to end with a . to support vocabularies which use dots in the IRIs for further structuring (eg. CIM-based formats like CGMES)
  • RDF.Triple.new/1 now also accepts four-element tuples and simple ignores fourth element
  • RDF.Quad.new/1 now also accepts three-element tuples and simple assumes the fourth element to be nil

Fixed

Compare v0.8.2...v0.9.0

0.8.2 - 2020-09-21

Added

  • the Turtle encoder can now produce partial Turtle documents with the :only option and any combination of the following values: :triples, :directives, :base, :prefixes
  • the style of the Turtle directives produced by the Turtle encoder can be switched to SPARQL style with the option :directive_style and the value :sparql
  • the most common conflict resolution strategies on RDF.PrefixMap.merge/3 can now be chosen directly with the atoms :ignore and :overwrite
  • RDF.PrefixMap.prefixed_name/2 to convert an IRI to a prefixed name
  • RDF.PrefixMap.prefixed_name_to_iri/2 to convert a prefixed name to an IRI

Changed

  • when serializing a RDF.Dataset with the Turtle encoder the prefixes of all of its graphs are used now

Fixed

Compare v0.8.1...v0.8.2

0.8.1 - 2020-06-16

Added

  • query functions for basic graph pattern matching (incl. streaming-support)

Compare v0.8.0...v0.8.1

0.8.0 - 2020-06-01

RDF literals and their datatypes were completely redesigned to support derived XSD datatypes and allow for defining custom datatypes. For an introduction on how literals work now read the updated page on literals in the guide. For more details on how to migrate from an earlier version read this wiki page.

Elixir versions < 1.8 are no longer supported

Added

  • a lot of new datatypes like xsd:float, xsd:byte or xsd:anyURI -- all numeric XSD datatypes are now available; see this page of the API documentation for an up-to-date list of all supported and missing XSD datatypes
  • an implementation of XSD facet system now makes it easy to define own custom datatypes via restriction of the existing XSD datatypes
  • RDF.Literal.update/2 updates the value of a RDF.Literal without changing anything else, eg. the language or datatype

Changed

  • the RDF.Literal struct now consists entirely of a datatype-specific structs in the literal field, which besides being more memory-efficient (since literals no longer consist of all possible fields a literal might have), allows pattern matching now on the datatype of literals.
  • RDF XSD datatypes are now defined in the RDF.XSD namespace
  • alias constructor functions for the XSD datatypes are now defined on RDF.XSD
  • matches?, less_than?, greater_than as higher level functions were removed from the RDF.Literal.Datatype modules
  • less_than?, greater_than? now always return a boolean and no longer nil when incomparable; you can still determine if two terms are comparable by checking if compare/2 returns nil
  • the language option is not supported on the RDF.XSD.String.new/2 constructor
  • the language option on RDF.Literal.new/2 is no longer ignored if it's empty (nil or ""), so this either produces an invalid RDF.LangString now or, if another datatype is provided will fail with an ArgumentError
  • canonical now performs implicit coercions when passed plain Elixir values
  • the inspect format for literals was changed and is now much more informative and uniform, since you now always see the value, the lexical form and if the literal is valid
  • RDF.Namespace.resolve_term/1 now returns ok or error tuples, but a new function RDF.Namespace.resolve_term!/1 with the old behaviour was added

Fixed

  • numeric operations on invalid numeric literals no longer fail, but return nil instead
  • Datetimes preserve the original lexical form of the timezone when casting from a date
  • BEAM error warnings when trying to use top-level modules as vocabulary terms

Compare v0.7.1...v0.8.0

0.7.1 - 2020-03-11

Added

  • proper typespecs so that Dialyzer passes without warnings (@rustra)

Fixed

  • RDF.XSD.Time didn't handle 24h overflows with an offset correctly

Compare v0.7.0...v0.7.1

0.7.0 - 2019-11-22

Added

  • RDF.Diff data structure for diffs between RDF graphs and descriptions
  • RDF.Description.update/4 updates the objects of a predicate in a description with a custom update function
  • RDF.Graph.update/4 updates the descriptions of a subject in a graph with a custom update function
  • RDF.Description.take/2 creates a description from another one by limiting its statements to a set of predicates
  • RDF.Graph.take/3 creates a graph from another one by limiting its statements to a set of subjects and optionally also a set of predicates
  • RDF.Graph.clear/1 removes the triples from a graph
  • Mix formatter configuration for using defvocab without parens

Changed

  • RDF.Serialization.Writer.write_file/4 which is the basis used by all the write_file/3 and write_file!/3 functions of all serialization format modules like RDF.NTriples, RDF.Turtle, JSON.LD etc. now opens file in a different mode: it no longer opens them with the :utf8 option. First, this by default slowed down the writing, but more importantly could lead to unexpected encoding issues. This is a breaking change: If your code relied on this file mode, you can get the old behaviour, by specifying the file_mode on these functions accordingly as [:utf8, :write, :exclusive]. For example, to write a Turtle file with the old behaviour, you can do it like this:
RDF.Turtle.write_file!(some_data, some_path, file_mode: ~w[utf8 write exclusive]a)

Compare v0.6.2...v0.7.0

0.6.2 - 2019-09-08

Added

Changed

  • RDF.Turtle.Decoder saves the base IRI in the RDF.Graph now
  • RDF.Turtle.Encoder now takes the base IRI to be used during serialization in
    the following order of precedence:
    • from the base option or its new alias base_iri
    • from the base_iri field of the given graph
    • from the RDF.default_base_iri returning the one from the application configuration
  • RDF.PrefixMap.new and RDF.PrefixMap.add now also accepts terms from RDF.Vocabulary.Namespaces as namespaces

Fixed

  • Vocabulary namespace modules weren't always detected properly

Compare v0.6.1...v0.6.2

0.6.1 - 2019-07-15

Added

  • RDF.IRI.to_string/1 returns the string representation of an RDF.IRI
    (implicitly resolving vocabulary namespace terms)
  • RDF.Literal.matches?/3 for XQuery regex pattern matching
  • RDF.Decimal.digit_count/1 and RDF.Decimal.fraction_digit_count/1 for
    determining the number of digits of decimal literals

Fixed

  • language literals were not properly unescaped during Turtle parsing
  • RDF.Literal.new/1 can take decimals and infers the datatype xsd:decimal correctly
  • true and false with capital letters are no longer valid RDF.Booleans following the XSD specification; the same applies for booleans in Turtle
  • +INF is no longer a valid RDF.Double (positive infinity doesn't expect a sign)
  • slightly improve output of errors during parsing of Turtle, N-Triples and N-Quads

Compare v0.6.0...v0.6.1

0.6.0 - 2019-04-06

see here for upgrading notes to RDF.ex 0.6

Added

Changed

  • the constructor functions for RDF.Graphs and RDF.Datasets now take the graph name resp. dataset name through a name option, instead of the first argument
  • RDF.Graph.new supports an additional prefixes argument to initialize the prefixes field
  • when RDF.Graph.add and RDF.Graph.put are called with another graph, its prefixes are merged
  • RDF.Turtle.Decoder saves the prefixes now
  • RDF.Turtle.Encoder now takes the prefixes to be serialized in the following order of precedence:
    • from the prefixes option (as before)
    • from the prefixes field of the given graph
    • from the RDF.default_prefixes
  • drop support for OTP < 20, since prefixes can consist of UTF characters which are not supported in atoms on these versions

Compare v0.5.4...v0.6.0

0.5.4 - 2019-01-17

Fixed

  • issue with Elixir 1.8
  • RDF.write_file and RDF.write_file! delegators had wrong signatures

Compare v0.5.3...v0.5.4

0.5.3 - 2018-11-11

Added

Compare v0.5.2...v0.5.3

0.5.2 - 2018-11-04

Added

Fixed

  • RDF.DateTime.equal_value?/2 and RDF.Date.equal_value?/2 did not handle timezones correctly
  • -00:00 is a valid timezone offset on RDF.DateTime

Compare v0.5.1...v0.5.2

0.5.1 - 2018-09-17

Fixed

  • generated Erlang output files of Leex and Yecc are excluded from Hex package

Compare v0.5.0...v0.5.1

0.5.0 - 2018-09-17

Elixir versions < 1.6 are no longer supported

Added

  • Possibility to execute simple SPARQL queries against RDF.Graphs with SPARQL 0.2
  • New RDF.Term protocol implemented for all structs representing RDF nodes and
    all native Elixir datatypes which are coercible to those modules. For now, it
    mainly offers, besides the coercion, just the function RDF.Term.equal?/2 and RDF.Term.equal_value?/2 for term- and value comparisons.
  • New RDF.Decimal datatype for xsd:decimal literals and support for decimal literals in Turtle encoder
  • RDF.Numeric module with a list of all numeric datatypes and shared functions for all numeric literals, eg. arithmetic functions
  • Various new RDF.Datatype function
    • RDF.Datatype.cast/1 for casting between RDF.Literals as specified in the XSD spec on all RDF.Datatypes
    • logical operators and the Effective Boolean Value (EBV) coercion algorithm from the XPath and SPARQL specs on RDF.Boolean
    • various functions on the RDF.DateTime and RDF.Time datatypes
    • RDF.LangString.match_language?/2
  • Many new convenience functions on the top-level RDF module
    • constructors for all of the supported RDF.Datatypes
    • constant functions RDF.true and RDF.false for the two boolean RDF.Literal values
  • RDF.Literal.Guards which allow pattern matching of common literal datatypes
  • RDF.BlankNode.Generator
  • Possibility to configure an application-specific default base IRI; for now it is used only on reading of RDF serializations (when no base specified)

Changed

  • RDF.String.new/2 and RDF.String.new!/2 produce a rdf:langString when given a language tag
  • Some of the defined structs now enforce keys on compile-time (via Elixirs @enforce_keys feature) when not setting the corresponding fields would lead to invalid structs, namely the following fields:
    • RDF.IRI.value
    • RDF.BlankNode.id
    • RDF.Description.subject
    • RDF.List.head

Fixed

  • RDF.resource?/1 does not fail anymore when called with unresolvable atoms but returns false instead
  • RDF.IRI.absolute/2 does not fail with a FunctionClauseError when the given base is not absolute, but returns nil instead
  • RDF.DateTime and RDF.Time store microseconds
  • RDF.DateTime: '24:00:00' is a valid time in a xsd:dateTime; the dateTime value so represented is the first instant of the following day
  • RDF.LangString: non-strings or the empty string as language produce invalid literals

Compare v0.4.1...v0.5.0

0.4.1 - 2018-03-19

Added

Changed

  • RDF.Literal.new/2 can create rdf:langString literals without failing, they
    are simply invalid; if you want to fail without a language tag use the new RDF.Literal.new!/2 function

Compare v0.4.0...v0.4.1

0.4.0 - 2018-03-10

Changed

Added

Compare v0.3.1...v0.4.0

0.3.1 - 2018-01-19

Added

  • Collectable implementations for all RDF.Data structures so they can be used as destinations of Enum.into and for comprehensions

Fixed

  • Fix unescape_map in parse_helper for Elixir 1.6 (@ajkeys)

Compare v0.3.0...v0.3.1

0.3.0 - 2017-08-24

Added

  • RDF.IRI as a more suitable URI/IRI representation for RDF, bringing enormous performance and memory consumption benefits (see here for the details about the improvements)

Changed

Fixed

Compare v0.2.0...v0.3.0

0.2.0 - 2017-08-12

Elixir versions < 1.4 are no longer supported

Added

Changed

  • Rename RDF.Statement.convert* functions to RDF.Statement.coerce*

Fixed

  • RDF.uri/1 and URI parsing of N-Triples and N-Quads decoders preserve empty fragments
  • booleans weren't recognized as coercible literals on object positions
  • N-Triples and N-Quads decoder didn't handle escaping properly

Compare v0.1.1...v0.2.0

0.1.1 - 2017-06-25

Fixed

  • Add src directory to package files.

Compare v0.1.0...v0.1.1

0.1.0 - 2017-06-25

Initial release

Note: This version is not usable, since the src directory is not part of the package, which has been immediately fixed on version 0.1.1.