lsmyrnaios
/
UrlsController
1
0
Fork 1
Code Issues Pull Requests Releases 10 Wiki Activity

Improve performance and reduce memory usage of the "findAssignmentsQuery": 

- Reorder JOINs and predicates to reduce the computational cost.
- Remove the memory-costly "pu.url" predicates from the "where" clause, as the DB has no empty urls anymore.
This commit is contained in:
Lampros Smyrnaios 2023-10-31 15:59:48 +02:00
parent db929d8931
commit 9b1f2c4931
1 changed files with 8 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
src/main/java/eu/openaire/urls_controller/services/UrlsServiceImpl.java
View File

@ -115,23 +115,21 @@ public class UrlsServiceImpl implements UrlsService {
"select pubid, url, datasourceid, datasourcename\n" + // Select the final sorted data with "assignmentsLimit".
"from (select distinct p.id as pubid, pu.url as url, pb.level as level, attempts.counts as attempt_count, p.year as pub_year, d.id as datasourceid, d.name as datasourcename\n" + // Select the distinct id-url data. Beware that this will return duplicate id-url pairs, wince one pair may be associated with multiple datasources.
" from " + DatabaseConnector.databaseName + ".publication p\n" +
" join " + DatabaseConnector.databaseName + ".datasource d on d.id=p.datasourceid and d.allow_harvest=true"+
((excludedDatasourceIDsStringList != null) ? // If we have an exclusion-list, use it below.
(" and d.id not in " + excludedDatasourceIDsStringList + GenericUtils.endOfLine) : "") +
" join " + DatabaseConnector.databaseName + ".publication_urls pu on pu.id=p.id\n" +
" join " + DatabaseConnector.databaseName + ".datasource d on d.id=p.datasourceid\n" + // This is needed for the "d.allow_harvest=true" check later on.
" left outer join " + DatabaseConnector.databaseName + ".publication_boost pb\n" +
" on p.id=pb.id\n" +
" left outer join (select count(at.original_url) as counts, at.original_url from " + DatabaseConnector.databaseName + ".attempt at group by at.original_url) as attempts\n" +
" on attempts.original_url=pu.url\n" +
" left anti join (select a.original_url from " + DatabaseConnector.databaseName + ".assignment a\n" +
" union all\n" +
" select pl.original_url from " + DatabaseConnector.databaseName + ".payload pl\n" + // Here we access the payload-VIEW which includes the three payload-tables.
" ) as existing\n" +
" on existing.original_url=pu.url\n" +
" where d.allow_harvest=true\n" +
((excludedDatasourceIDsStringList != null) ? // If we have an exclusion-list, use it below.
(" and d.id not in " + excludedDatasourceIDsStringList + GenericUtils.endOfLine) : "") +
" and coalesce(attempts.counts, 0) <= " + maxAttemptsPerRecordAtomic.get() + GenericUtils.endOfLine +
" left outer join " + DatabaseConnector.databaseName + ".publication_boost pb\n" +
" on p.id=pb.id\n" +
" left outer join (select count(at.original_url) as counts, at.original_url from " + DatabaseConnector.databaseName + ".attempt at group by at.original_url) as attempts\n" +
" on attempts.original_url=pu.url\n" +
" where coalesce(attempts.counts, 0) <= " + maxAttemptsPerRecordAtomic.get() + GenericUtils.endOfLine +
" and not exists (select 1 from " + DatabaseConnector.databaseName + ".attempt a where a.original_url=pu.url and a.error_class = 'noRetry' limit 1)\n" +
" and pu.url != '' and pu.url is not null\n" + // Some IDs have empty-string urls, there are no "null" urls, but keep the relevant check for future-proofing.
" and (p.year <= " + currentYear + " or p.year > " + (currentYear + 5) + ")\n" + // Exclude the pubs which will be published in the next 5 years. They don't provide full-texts now. (We don't exclude all future pubs, since, some have invalid year, like "9999").
") as distinct_results\n" +
"order by coalesce(attempt_count, 0), coalesce(level, 0) desc, coalesce(pub_year, 0) desc, reverse(pubid), url\n" + // We also order by reverse "pubid" and "url", in order to get the exactly same records for consecutive runs, all things being equal.