Run a SPARQL query

Sends a query by POST (a VALUES block of a hundred URIs is too long for a GET URL) and returns the results as a table. sparql_values() formats URIs as a VALUES block.

Usage

sparql_query(
  query,
  endpoint = "https://publications.europa.eu/webapi/rdf/sparql",
  pause = 1,
  max_tries = 12L
)

sparql_values(variable, uris)

Arguments

query

Character. A SPARQL query.

endpoint

Character. URL of the endpoint. By default the endpoint of the EU Publications Office (CELLAR).

pause

Numeric. Minimum seconds between two queries to the endpoint.

max_tries

Integer. Attempts per query.

variable

Character. Variable name, without ?.

uris

Character vector of URIs.

Value

sparql_query(): a data.table with one character column per selected variable, NA for an unbound variable. A query that still fails after the retries is an error of class politecollect_request_failed. sparql_values(): a string.

Details

Results are requested as SPARQL JSON rather than CSV: abstracts contain quotes and line breaks that make CSV parsing fragile.

The defaults come from build_eu_database. CELLAR answers 5xx when overloaded, with outages of up to 14 minutes observed (2026-10-20), and rate-limits bursts of queries with HTTP 429 and a Retry-After header (1,800 s observed on 2026-10-16), which is honoured: a run can pause for half an hour and then resume by itself. Up to 12 tries, over about 33 minutes. Do not run several harvests of an endpoint at once.

robots.txt is not read: an endpoint is an API, governed by its terms of use.

See also

Other APIs: api_paginate(), api_request(), crossref_request(), ia_search(), openalex_api_key(), scopus_api_key()

Examples

sparql_values("work", c("http://example.org/1", "http://example.org/2"))
#> [1] "VALUES ?work { <http://example.org/1> <http://example.org/2> }"
if (FALSE) { # \dontrun{
options(politecollect.user_agent = "my_project (me@example.org)")
sparql_query("SELECT ?s WHERE { ?s ?p ?o } LIMIT 3")
} # }