replace '_' and ' ' when matching wikipedia article names

diff --git a/wikidata/import.sh b/wikidata/import.sh
index c701b03219050e67d8592194b983a590ab0a96f4..97621f174b9825d44602f9d6ef7ef0176dabab07 100755 (executable)
--- a/wikidata/import.sh
+++ b/wikidata/import.sh
@@ -20,5 +20,9 @@ $PSQL -c "alter table entity add column description_en text"
 $PSQL -c "update entity set description_en = description from entity_description where entity.entity_id = entity_description.entity_id and language = 'en'"
 
 cat totals.txt | $PSQL -c "COPY import_link_hit from STDIN WITH CSV DELIMITER ' '"
-$PSQL -c "insert into link_hit select target||'wiki', catch_decode_url_part(value), sum(hits) from import_link_hit group by target||'wiki', catch_decode_url_part(value)"
+$PSQL -c "truncate link_hit"
+$PSQL -c "insert into link_hit select target||'wiki', replace(catch_decode_url_part(value), '_', ' '), sum(hits) from import_link_hit where replace(catch_decode_url_part(value), '_', ' ') is not null group by target||'wiki', replace(dcatch_decode_url_part(value), '_', ' ')"
+$PSQL -c "truncate entity_link_hit"
 $PSQL -c "insert into entity_link_hit select entity_id, target, value, coalesce(hits,0) from entity_link left outer join link_hit using (target, value)"
+$PSQL -c "create table entity_hit as select entity_id,sum(hits) as hits from entity_link_hit group by entity_id"
+$PSQL -c "create unique index idx_entity_hit on entity_hit using btree (entity_id)"