Class AnnotationRelationDaoImpl

java.lang.Object

ubic.gemma.persistence.service.AbstractDao<AnnotationRelation>

ubic.gemma.persistence.service.common.description.AnnotationRelationDaoImpl

All Implemented Interfaces:

BaseDao<AnnotationRelation>, AnnotationRelationDao

@Repository
public class AnnotationRelationDaoImpl
extends AbstractDao<AnnotationRelation>
implements AnnotationRelationDao

Hibernate implementation of AnnotationRelationDao.

Native SQL throughout. The reads are aggregations over a derived table with an ACL clause that has to be composed per user class, and the ACL helper emits SQL, not HQL - the same reason CharacteristicDaoImpl's EE2C reads are native.

The expensive part happens offline. Every query here is an indexed lookup plus a group-by over a small purpose-built table. There is deliberately no candidate cut, no self-join over EE2C, and no second query stitched to the first by a string key in Java: those existed only to fit a corpus-wide derivation inside a request, and once the derivation runs in the maintenance job there is nothing left to approximate around.

Nested Class Summary

Nested classes/interfaces inherited from interface AnnotationRelationDao

AnnotationRelationDao.Direction, AnnotationRelationDao.RelationQuery, AnnotationRelationDao.RelationSummary

Field Summary

Fields inherited from class AbstractDao

log

Constructor Summary

Constructors

Constructor	Description
AnnotationRelationDaoImpl(SessionFactory sessionFactory)

Method Summary

Modifier and Type	Method	Description
List<String[]>	findRelatedTerms(Collection<String> seedValueUris, Collection<String> seedValues, AnnotationRelationDao.Direction direction, Set<AnnotationRelationBasis> bases, Collection<String> predicateUris, Long taxonId, Collection<Long> excludedExperimentIds, int maximumObjectBreadth, int maxResults)	Set membership: every term related to any of the seeds, with no counting, no ranking and no ambiguity to resolve.
List<AnnotationRelationDao.RelationSummary>	findRelations(AnnotationRelationDao.RelationQuery q)	The ranked, evidenced read.
int	removeByBasis(AnnotationRelationBasis basis, Long experimentId)	Remove every derived row for a basis, optionally for one experiment.
int	removeByBasis(AnnotationRelationBasis basis, Long experimentId, String source)	The same rebuild delete, narrowed to one source.

Methods inherited from class AbstractDao

countAll, create, create, find, findByProperty, findByPropertyIn, findIdByProperty, findOneByProperty, findOrCreate, getBatchSize, getElementClass, getEntityName, getIdentifierPropertyName, getSessionFactory, load, load, loadAll, loadReference, loadReference, reload, reload, remove, remove, save, save, streamAll, streamAll, streamQuery, update, update

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface BaseDao

countAll, create, create, find, findOrCreate, getElementClass, load, load, loadAll, loadReference, loadReference, reload, reload, remove, remove, save, save, streamAll, streamAll, update, update

Constructor Details

AnnotationRelationDaoImpl

@Autowired
public AnnotationRelationDaoImpl(SessionFactory sessionFactory)

Method Details

findRelations

public List<AnnotationRelationDao.RelationSummary> findRelations(AnnotationRelationDao.RelationQuery q)

Description copied from interface: AnnotationRelationDao

The ranked, evidenced read. Empty when the query names neither end of the relation.

Support is counted here, never stored, so the counts are exact for the calling user's ACL.

Specified by:

findRelations in interface AnnotationRelationDao

findRelatedTerms

public List<String[]> findRelatedTerms(Collection<String> seedValueUris,
 Collection<String> seedValues,
 AnnotationRelationDao.Direction direction,
 Set<AnnotationRelationBasis> bases,
 Collection<String> predicateUris,
 @Nullable
 Long taxonId,
 Collection<Long> excludedExperimentIds,
 int maximumObjectBreadth,
 int maxResults)

Description copied from interface: AnnotationRelationDao

Set membership: every term related to any of the seeds, with no counting, no ranking and no ambiguity to resolve.

This is the primitive behind query expansion and behind "is this tag already implied?". It is one indexed lookup and has to stay cheap enough to sit on an interactive path.

Specified by:

findRelatedTerms in interface AnnotationRelationDao

Parameters:

seedValueUris - grounded seeds; may be empty if seedValues is not

seedValues - ungrounded seeds, matched on the value string - ordinary here, since plenty of curated statement objects are free text (aortic banding has none)

Returns:

the related terms, each as [value, valueUri], the URI possibly null

removeByBasis

public int removeByBasis(AnnotationRelationBasis basis,
 @Nullable
 Long experimentId)

Description copied from interface: AnnotationRelationDao

Remove every derived row for a basis, optionally for one experiment.

Rebuild, not upsert - see AnnotationRelation. An upsert can only correct rows the new query still produces, so a row whose source annotation was deleted would outlive it.

Specified by:

removeByBasis in interface AnnotationRelationDao

removeByBasis

public int removeByBasis(AnnotationRelationBasis basis,
 @Nullable
 Long experimentId,
 @Nullable
 String source)

Description copied from interface: AnnotationRelationDao

The same rebuild delete, narrowed to one source.

AnnotationRelationBasis.ONTOLOGY has more than one producer — CLO states which disease a cell line came from, CHEBI which roles a chemical bears — and they are loaded and rebuilt independently. Rebuilding one with the basis-wide delete would silently drop the other's rows and leave the table looking like the other producer had never run.

Specified by:

removeByBasis in interface AnnotationRelationDao

Parameters:

source - the source to remove, or null for every source under the basis