SQLAlchemy | Limit join result to one row for one-to-many relationship

问题

I have two entities: Projects and Student Lists. One Project can have many Student lists.

I am attempting to join student list on project and only return the first row for each project based on custom ordering of student lists.

Attempted Subquery:

_whens = {ProjectStatus.APPROVED: 1, ProjectStatus.REJECTED: 2, 
          ProjectStatus.SUBMITTED: 3, None: 4}
sort_order = case(value=StudentList.student_list_status_id, whens=_whens)

return self._session.query(StudentList).
        filter(StudentList.student_list_id==Project.project_id)
       .order_by(sort_order).limit(1).subquery()

Above I define the custom ordering based on student list status id. The function returns the subquery which I then attempt to join to my Project outerquery below (student_list_subquery refers to what is returned above):

projects = self._session.query(models.Project)
            .filter(models.Project.project_year == year)
            .join(student_list_subquery,
            student_list_subquery.c.project_id==Project.project_id)
            .all()

Below is the relevant SQL output

FROM project 
LEFT OUTER JOIN (SELECT student_list.project_id AS project_id, 
 student_list.student_list_id AS student_list_id
 FROM student_list, project
 WHERE project.project_id = student_list.project_id 
 ORDER BY CASE student_list.student_list_status_id WHEN 102 THEN 1 
 WHEN 105 THEN 2 WHEN 101 THEN 3 WHEN NULL THEN 4 END
 LIMIT 1) AS anon_1 ON anon_1.project_id = project.project_id

I am using mySQL so (Distinct On) solutions won't work and neither will row_number/partition solutions either...

I seem to be having the same issue raised here SQLAlchemy: FROM entry still present in correlated subquery

回答1:

Finally Solved the issue. Hope this helps someone else trying to solve the first-n-per-group problem when custom ordering of group required using SQLAlchemy and mySQL.

First I have this function that returns the one student_list_status_id with highest priority for the project(hence the filter).

@staticmethod
def create_student_list_subquery(session):
    '''create a correlated subquery that will limit result to one student
    list per project with custom sorting to retrieve highest priority list
    per project based on status'''

    sl2=aliased(StudentList)
    list_id = sl2.student_list_status_id.label("list_id")
    _whens = {ProjectStatus.APPROVED: 1, ProjectStatus.REJECTED: 2, 
             ProjectStatus.SUBMITTED: 3, None: 4}
    sort_order = case(value=list_id, whens=_whens)

    return session.query(list_id).filter(sl2.project_id==Project.project_id)
                  .order_by(sort_order)
                  .limit(1)

The I join project status which correlates to the student_list_status_id in the query above (aliased as ps) onto the project. Then, I can sort on the project status name which was my goal.

self._session.query(models.Project)
        .filter(models.Project.project_year == year)
        .join(ps, ps.project_status_id==student_list_subq)
        .all()

Note that student_list_subq refers to result of create_student_list_subquery function above.

来源：https://stackoverflow.com/questions/50109243/sqlalchemy-limit-join-result-to-one-row-for-one-to-many-relationship

标签

python

mysql

sqlalchemy

flask-sqlalchemy