HHH-19396 cannot select the same column twice with different aliases while using cte #10158
Conversation
cigaly
changed the title
cigaly
force-pushed
the
HHH-19396-Cannot_select_the_same_column_twice_with_different_aliases_while_using_CTE
branch
from
5126bf6 to
8b3f32e
Compare
|
Previous change was always setting |
cigaly
force-pushed
the
HHH-19396-Cannot_select_the_same_column_twice_with_different_aliases_while_using_CTE
branch
from
8b3f32e to
38b7ede
Compare
| @@ -491,7 +491,7 @@ private static CacheableSqmInterpretation buildCacheableSqmInterpretation( | |||
| executionContext.getQueryParameterBindings(), | |||
| executionContext.getSession().getLoadQueryInfluencers(), | |||
| sessionFactory.getSqlTranslationEngine(), | |||
| true | |||
| sqm.getCteStatements().isEmpty() | |||
There was a problem hiding this comment.
We shouldn't do this for the entire query, only because there are CTEs involved. IIUC, this only matters when rendering the CTE itself. We should rather change the BaseSqmToSqlAstConverter#deduplicateSelectionItems to false before visiting the CTEs themselves (e.g. in org.hibernate.query.sqm.sql.BaseSqmToSqlAstConverter#visitCteContainer) like we do for subqueries.
| String alias = aliases == null ? null : aliases.get( i ); | ||
| if ( alias == null ) { | ||
| alias = component.getAlias(); | ||
| } |
There was a problem hiding this comment.
I don't think you need this double-check: you're already getting the selectable node's alias in extractAliases().
| protected static List<String> extractAliases(SqmSelectQuery<?> subQuery) { | ||
| final SqmSelectClause selectClause = subQuery.getQueryPart() | ||
| .getFirstQuerySpec() | ||
| .getSelectClause(); | ||
| final var aliases = new ArrayList<String>(); | ||
| for (final var selection : selectClause.getSelections()) { | ||
| final var alias = selection.getAlias(); | ||
| selection.getSelectableNode().visitSubSelectableNodes( node -> | ||
| aliases.add( alias == null ? node.getAlias() : alias ) | ||
| ); | ||
| } | ||
| return aliases; | ||
| } |
There was a problem hiding this comment.
I'm guessing we need this to get the SqmSelection#getAlias value, instead of the SqmSelectableNode's one. Edit: have you perhaps looked into why that's needed? Would be nice to add a clarification comment here for future reference.
We're already iterating through selections when calling getSelectionItems() though, maybe it would be worth consolidating the creation of the lists within the extractSqmExpressibles and createStatementTable methods. Also, small nitpick, but I would either get both as arrays or both as lists.
There was a problem hiding this comment.
One SqmSelection can contain more than one SqmSelectableNode. I've tried that what you are saying in first "iteration", but that caused some twenty tests to fail. See, for example, org.hibernate.orm.test.jpa.criteria.subquery.SubqueryTests#testNoFromClauseInSubquery.
I'll do some refactoring into new commit so, please, check what you (don't) like.
…wice in same (sub)query
…mporary tables and subqueries
Selection items representing same column in (sub)query will be represented with single instance, this will hide real alias
Temporary deduplication disabling moved from org.hibernate.query.sqm.internal.ConcreteSqmSelectQueryPlan.buildCacheableSqmInterpretation into org.hibernate.query.sqm.sql.BaseSqmToSqlAstConverter.visitCteContainer
cigaly
force-pushed
the
HHH-19396-Cannot_select_the_same_column_twice_with_different_aliases_while_using_CTE
branch
from
6a83ba0 to
9a986df
Compare
Jira issue HHH-19396
When creating
SqlAstQueryPartProcessingStateImplin methodorg.hibernate.query.sqm.sql.BaseSqmToSqlAstConverter#visitQuerySpecparameterdeduplicateSelectionItemsmust be set tofalse(see comment in moved code block)When
selectStatementcontains same column more than once in methodorg.hibernate.query.sqm.tree.cte.SqmCteTable#createStatementTable,SqmCteTableconstructor will be unable to properly resolve aliases. To prevent this alias in select clause selection should be used if exist, one from selectable node should be used otherwise.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license
and can be relicensed under the terms of the LGPL v2.1 license in the future at the maintainers' discretion.
For more information on licensing, please check here.
https://hibernate.atlassian.net/browse/HHH-19396