Join multiple tables in SQLAlchemy/Flask

Question

I am trying to figure out the correct join query setup within SQLAlchemy, but I can't seem to get my head around it.

I have the following table setup (simplified, I left out the non-essential fields):

class Group(db.Model):
    id            = db.Column(db.Integer, primary_key = True)
    number        = db.Column(db.SmallInteger, index = True, unique = True)
    member        = db.relationship('Member', backref = 'groups', lazy = 'dynamic')

class Member(db.Model):
    id            = db.Column(db.Integer, primary_key = True)
    number        = db.Column(db.SmallInteger, index = True)
    groupid       = db.Column(db.Integer, db.ForeignKey('group.id'))
    item          = db.relationship('Item', backref = 'members', lazy = 'dynamic')

class Version(db.Model):
    id           = db.Column(db.Integer, primary_key = True)
    name         = db.Column(db.String(80), index = True)
    items        = db.relationship('Item', backref='versions', lazy='dynamic')  

class Item(db.Model):
    id           = db.Column(db.Integer, primary_key = True)
    member       = db.Column(db.Integer, db.ForeignKey('member.id'))
    version      = db.Column(db.Integer, db.ForeignKey('version.id'))

So the relationships are the following:

1:n Group Member
1:n Member Item
1:n Version Item

I would like to construct a query by selecting all Item-Rows from the database, that have a certain version. Then I would like to order them by Group and then by Member. The output using Flask/WTForm should look something like this:

* GroupA
  * MemberA
     * ItemA (version = selected by user)
     * ItemB ( dito )
  * Member B
     * ItemC ( dito )
  ....

I have come up with something like the following query, but I am pretty sure that it is not correct (and inefficient)

   session.query(Item,Member,Group,Version)
    .join(Member).filter(version.id==1)
    .order_by(Group).order_by(Member).all()

My first intuitive approach would have been to create something like

Item.query.join(Member, Item.member==Member.id)
    .filter(Member.versions.name=='MySelection')
    .order_by(Member.number).order_by(Group.number)

but obviously, this doesn't work at all. The join operation on the Version table does not seem to produce the type of join between the two tables that I expected. Maybe I am totally misunderstanding the concept, but after reading the tutorials this would have made sense to me.

van · Accepted Answer · 2014-02-25 17:13:12Z

55

Following will give you the objects you need in one query:

q = (session.query(Group, Member, Item, Version)
        .join(Member)
        .join(Item)
        .join(Version)
        .filter(Version.name == my_version)
        .order_by(Group.number)
        .order_by(Member.number)
        ).all()
print_tree(q)

However, the result you get will be a list of tuples (Group, Member, Item, Version). Now it is up to you to display it in a tree form. Code below might prove useful though:

def print_tree(rows):
    def get_level_diff(row1, row2):
        """ Returns tuple: (from, to) of different item positions.  """
        if row1 is None: # first row handling
            return (0, len(row2))
        assert len(row1) == len(row2)
        for col in range(len(row1)):
            if row1[col] != row2[col]:
                return (col, len(row2))
        assert False, "should not have duplicates"

    prev_row = None
    for row in rows:
        level = get_level_diff(prev_row, row)
        for l in range(*level):
            print 2 * l * " ", row[l]
            prev_row = row

Update-1: If you are willing to forgo lazy = 'dynamic' for the first two relationships, you can a query to load a whole object network (as opposed to tuples above) with the code:

q = (session.query(Group)
        .join(Member)
        .join(Item)
        .join(Version)
        # @note: here we are tricking sqlalchemy to think that we loaded all these relationships,
        # even though we filter them out by version. Please use this only to get data and display,
        # but not to continue working with it as if it were a regular UnitOfWork
        .options(
            contains_eager(Group.member).
            contains_eager(Member.items).
            contains_eager(Item.version)
            )
        .filter(Version.name == my_version)
        .order_by(Group.number)
        .order_by(Member.number)
        ).all()

# print tree: easy navigation of relationships
for g in q:
    print "", g
    for m in g.member:
        print 2 * " ", m
        for i in m.items:
            print 4 * " ", i

edited Feb 25, 2014 at 17:13

answered Feb 25, 2014 at 5:54

van

77.6k13 gold badges174 silver badges179 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

user3347953 Over a year ago

Thanks alot. I have experimented and came up with a very similar query to your solution. I just had the feeling that it seemed not elegant, as I had all those relationships defined between the tables in the class definitions. The print_tree code is great!

van Over a year ago

I know of a more elegent way to get what you want, but it would not work with lazy = 'dynamic' relationships. Is this something you insist on having?

Collectives™ on Stack Overflow

Join multiple tables in SQLAlchemy/Flask

1 Answer 1

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related