JPA Query with count and group by (many to many relationship with join table) - jpa

Question

I have a little problem. In my application I have two entities, which look like on the snippets:
//imports, annotations, named queries
public class WishList implements Serializable {
private static final long serialVersionUID = 1L;
#Id
#SequenceGenerator(name = "wishes_seq",
sequenceName = "wish_list_id_seq",
allocationSize = 1)
#GeneratedValue(strategy = GenerationType.SEQUENCE, generator = "wishes_seq")
#Basic(optional = false)
#NotNull
#Column(name = "id")
private Integer id;
#Basic(optional = false)
#NotNull
#Size(min = 1, max = 80)
#Column(name = "book_title")
private String bookTitle;
#Size(max = 80)
#Column(name = "book_cattegory")
private String bookCattegory;
#ManyToMany(cascade= CascadeType.ALL, fetch= FetchType.EAGER)
#JoinTable(name="wl_authors",
joinColumns={#JoinColumn(name="wl_id")},
inverseJoinColumns={#JoinColumn(name="author_id")})
private List<Author> authorList;
// other methods
and the second one:
//imports, annotations, named queries
public class Author implements Serializable {
private static final long serialVersionUID = 1L;
#Id
#SequenceGenerator(name = "authors_seq",
sequenceName = "authors_id_seq",
allocationSize = 1)
#GeneratedValue(strategy = GenerationType.SEQUENCE,
generator = "authors_seq")
#Basic(optional = false)
#NotNull
#Column(name = "id")
private Integer id;
#Basic(optional = false)
#NotNull
#Size(min = 1, max = 58)
#Column(name = "author_name")
private String authorName;
#Basic(optional = false)
#NotNull
#Size(min = 1, max = 58)
#Column(name = "author_surname")
private String authorSurname;
#Size(max = 58)
#Column(name = "nationality")
private String nationality;
#Size(max = 64)
#Column(name = "birth_place")
private String birthPlace;
#JoinTable(name = "wl_authors",
joinColumns = {#JoinColumn(name = "author_id",
referencedColumnName = "id")},
inverseJoinColumns = {#JoinColumn(name = "wl_id",
referencedColumnName = "id")})
#ManyToMany(fetch = FetchType.EAGER)
private List<WishList> wishListList;
// other methods
As You can see entities, which I prepared represent two tables in database, which are in many to many relationship (I used join table). I want to count all results in WishList entity, which have the same authors, title and cattegory and return all results as follows:
count result
book title
book cattegory
book authors
The results should be ordered by count result.
The JPA query, which I prepared:
SELECT Count(wl.bookTitle) AS POP,
wl.bookTitle,
wl.bookCattegory
FROM WishList wl
GROUP BY wl.bookTitle,
wl.bookCattegory,
wl.authorList
ORDER BY POP DESC
doesn't satisfy me. I can`t return authors of the book from WishList. When I place wl.authorList before 'FROM' EJB exception appears:
...
Caused by: Exception [EclipseLink-4002] (Eclipse Persistence Services - 2.3.0.v20110604-r9504): org.eclipse.persistence.exceptions.DatabaseException
Internal Exception: org.postgresql.util.PSQLException: ERROR: column "t1.id" must appear in the GROUP BY clause or be used in an aggregate function
...
Call: SELECT COUNT(t0.book_title), t0.book_title, t0.book_cattegory, t1.id, t1.author_name, t1.author_surname, t1.birth_place, t1.nationality FROM wl_authors t4, wl_authors t3, authors t2, authors t1, wish_list t0 WHERE (((t3.wl_id = t0.id) AND (t1.id = t3.author_id)) AND ((t4.wl_id = t0.id) AND (t2.id = t4.author_id))) GROUP BY t0.book_title, t0.book_cattegory, t2.id, t2.author_name, t2.author_surname, t2.birth_place, t2.nationality ORDER BY COUNT(t0.book_title) DESC
Query: ReportQuery(referenceClass=WishList sql="SELECT COUNT(t0.book_title), t0.book_title, t0.book_cattegory, t1.id, t1.author_name, t1.author_surname, t1.birth_place, t1.nationality FROM wl_authors t4, wl_authors t3, authors t2, authors t1, wish_list t0 WHERE (((t3.wl_id = t0.id) AND (t1.id = t3.author_id)) AND ((t4.wl_id = t0.id) AND (t2.id = t4.author_id))) GROUP BY t0.book_title, t0.book_cattegory, t2.id, t2.author_name, t2.author_surname, t2.birth_place, t2.nationality ORDER BY COUNT(t0.book_title) DESC")
Caused by: org.postgresql.util.PSQLException: ERROR: column "t1.id" must appear in the GROUP BY clause or be used in an aggregate function
...
Can somebody help me to create apriopriate query? Is it possible to do it in single query?

Answer 1

I don't think you can achieve it using one JPQL query. Perhaps, if native query would be used you might achieve it (not sure, tho).
Consider splitting the problem into two parts:
invoke a query which returns the statistics but without authors,
for each returned row, fetch the authors.
So firstly invoke such query:
SELECT DISTINCT COUNT(wl.bookTitle) AS POP,
wl.bookTitle,
wl.bookCattegory,
MAX(wl.id)
FROM WishList wl
GROUP BY wl.bookTitle, wl.bookCattegory, wl.authorList
ORDER BY POP DESC
According to the query, each returned row will have the same authors. The MAX(wl.id) is used just to get one of these ids you can query for the authors list.
Having this id you can for each record fetch the authors and return such prepared data (popularity, book title, category, authors).
It'll surely not get the highest ranks in the performance category as doing it in one query, I don't see other solution at this point.
By the way - are you sure your model is OK? Each WishList have exactly one book title, category and authors? Shouldn't the book data be separated to a Book entity and each WishList consists of many Books? I think it then should be easier to query for WishLists which consists of the same books.

Answer 2

I don't think you can group by a ManyToMany,
try,
SELECT Count(wl.bookTitle) AS POP, wl.bookTitle, wl.bookCattegory, a.id FROM WishList wl join wl.authorList a GROUP BY wl.bookTitle, wl.bookCattegory, a.id ORDER BY POP DESC

Answer 3

I know it's a bit late, but does the query work on other databases, like MySQL?
Postgres had a limitation and made mandatory in group by section all the columns present in select section - exactly what the error says.
From v9.1, this error should not happen anymore:
Allow non-GROUP BY columns in the query target list when the primary key is specified in the GROUP BY clause [...] The SQL standard allows this behavior, and because of the primary key, the result is unambiguous.