> -----Ursprüngliche Nachricht----- > Von: Georg Nozicka [mailto:georg.nozi...@nozicka.at] > Gesendet: Samstag, 31. März 2012 17:25 > An: users@openjpa.apache.org > Betreff: Unnecessary Join when using the Type construct in JPA Query > > Setup > > Imagine the following entities: > > A Company with a 1:Many to AddressLocation and another 1:Many > to AddressPostal. > > AddressLocation and AddressPostal are two different kind of > addresses. They are both subclasses of Address. > > > > The goal > > > The goal is to have to following two queries which both have the same > result: > > Match just by the name of the company. > Match by the name of the company and street of the location address. > > > The result is always the name of the company and the street > of the location address. > > Version1: Hand written native queries > > > Query by company name only > > > > Query by company name and street of location address > > > > The question is if it is possible to formulate these queries > with JPA Query Language in such a way, that also just one > join is used. > > So I tried the versions below. > > Version 2: JPA Query > > Query by company name only > > > > > Results is the following SQL: > > > Query by company name and street of location address > > > > > Results is the following SQL: > > > Well, both queries do almost what I original wanted, i.e. > just two joins, except, that they do not limit the queries to > the location address. > Therefore the queries are not semantically correct. > > > Version 3: JPA Query with Type construct > > To limit the query to a specific type, in my case the type > AddressLocation, the type construct must be used. > > Query by company name only > > > > Results is the following SQL: > > > Query by company name and street of location address > > > > Results is the following SQL: > > > From a conceptional point of view these queries do exactly > what I want, except, that two joins are used. Actually I have > no idea why in both cases the second join to Address is done, > the alias t2 is only used in the select > (t2.id) where I do not need it. > > Conclusion and Question > > The question is, if there is another way to express the JPA > Query so that only one join is used and that only addresses > of type AddressLocation are taken into account. Or is the > only solution, to step back to native queries which would > really be a pity because we would like to use higher level > APIs like the Criteria API or Querydsl which rely on JPA > Queries. I.e. as long as at the JPA Query level these kind of > problems are not solved, we can not use Criteria API or Querydsl. > > -- > View this message in context: > http://openjpa.208410.n2.nabble.com/Unnecessary-Join-when-usin g-the-Type-construct-in-JPA-Query-tp7424993p7424993.html > Sent from the OpenJPA Users mailing list archive at Nabble.com.
Hello Georg, It would seem that your SQL Statements are missing in the original mail, which makes it difficult to visualize where your problem may lie. Can you provide them?  John ---- Who is General Failure, and why is he reading my hard disk?
