Load and save

1 Uniform API for parsers/writers

Observations:

• Reading/writing RDF files is currently spread over multiple modules.

Proposals:

• Core predicates:
  • rdf_load(+Input, +Options)
  • rdf_save(+Input, +Options)
• Wrapper predicates:
  • rdf_load(+Input)
  • rdf_save(+Input)
• Properties or rdf_load/[1,2] and rdf_save/[1,2]:
  • Using open_any/5 and close_any/1, Input can a (1) file, (2) stream, or (3) HTTP(S) IRI.
  • Using archive_data_stream/3, all archive streams of Input are processed in sequence.
• Properties of rdf_load/[1,2]:
  • If format(-) then use rdf_guess_format/3 to determine the serialization format.
  • If Input is an HTTP(S) IRI then option base_iri is automatically set.
  • Option format(?) can either be used to set a serialization format or it can be used to return the serialization format that was heuristically found.

Notes:

• Format guessing should probably be part of a new library, keeping the plugin based infrastructure for library rdf_db.

2 Loading named graphs

Observations:

• rdf_load(+Input, +Options) with option graph(+Graph) does not make clear where quadruples are stored?
• rdf_save(+Input, +Options) with option graph(+Graph) does not allow a collection graphs to be saved to a quadruples file.

Proposal:

• For rdf_load/2 option dataset(-Default:atom, -Named:list(atom)) returns the graphs that data has be loaded into (i.e., Named=[] for triples).
• For rdf_save/2 option dataset(+Default:atom, +Named:list(atom)) allows a subset of the currently loaded graphs to be saved to file.
• Option graph(?) is a shorthand for option dataset(?,[]).

3 Serialization format-specific parsers/writers

Proposal:

• Rename predicates used for loading/saving XML/RDF:
  • rdf_load/[1,2] → rdf_xml_load/[1,2]
  • rdf_save/[1,2] → rdf_xml_save/[1,2]
  • rdf_save_footer/1 → rdf_xml_save_footer/1
  • rdf_save_header/2 → rdf_xml_save_header/2
  • rdf_save_subject/3 → rdf_xml_save_subject/3

4 Missing parsers/generators

The following formats are not supported:

• JSON-LD
• N3