Local label graph workflow

[nomisto]

Currently the flow for a user that does not want to publish a dataset/label graph is:

• Upload files to temp folder on server and insert graph to Apache Fuseki
• Make dataset available via private key
• Delete files from temp folder on an interval basis (e.g. daily) and remove temp graphs from Fuseki

(The workflow is the same for explainations right now, however db files could be easily acessed locally via sql.js.)

This is done as browser javascript implementations of rdf stores seem to be either slow or require a lot of in memory.

Another problem with in-memory stores is that the graph has to be reloaded if the user reloads the page.

Libraries investigated:

• rdfstore-js:
  • Supports only json-ld, turtle, n3
  • memory/Indexed db
  • If memory: Fast but requires a lot of memory
  • If Indexed db: Extremely slow to load
• rdflib.js
  • Supports only RDF/XML, Turtle, N3, JSON-LD
  • Only memory
  • Fast but requires a lot of memory
• node-quadstore
  • requires external parser, f.e. N3
  • requires external backend, f.e. leveldown
  • Parsing is slow (f.e. about 8 min for OpenBioLink)
  • Storage a lot faster than rdfstore-js with Indexed db