def get_candidates()

in blink/candidate_generation.py [0:0]

• 41 lines of code
• 11 McCabe index (conditional complexity)

    def get_candidates(self, mention_data):
        solr = self.solr

        # Build query
        keys = self.keys
        query = self.query
        if not self.raw_solr_fields:
            query = query.format(
                *[
                    BM45_Candidate_Generator.solr_escape(mention_data[key])
                    if key in mention_data
                    else utils.get_sent_context(mention_data, key)
                    for key in keys
                ]
            )
        else:
            query = query.format(
                *[
                    mention_data[key]
                    if key in mention_data
                    else utils.get_sent_context(mention_data, key)
                    for key in keys
                ]
            )

        try:
            results = solr.search(query, **self.query_arguments)
        except Exception as e:
            exc_type, exc_obj, exc_tb = sys.exc_info()
            print("\nException:", exc_type, "- line", exc_tb.tb_lineno)
            print(repr(e))

            c = self.c
            if c < 10:
                print(
                    "Exception with: \naddress: {} \nquery: {} \nmention_data: {} \n".format(
                        self.solr_address, query, str(mention_data)
                    )
                )
            self.c = c + 1

            return []

        # Filter the data in the retrieved objects, while ignoring the ones without a wikidata_id (only a very small fraction in the dataset; they are noise)
        filtered_results = [
            self._filter_result(cand) for cand in results.docs if "wikidata_id" in cand
        ]
        return filtered_results