在SQLAlchemy多对多关联表中存储对其他行的引用

Question

I'm working on a project in SQLAlchemy. Let's say (to simplify the problem) there are three tables at play-- Blog_posts, Users, and Users_daily_posts. The first two look like this:

class Blog_posts(db.Model):
    __tablename__ = 'Blog_posts'
    post_id = db.Column(db.Integer(), primary_key = True)
    date = db.Column(db.Integer())
    creator_id = db.Column(db.Integer())

class Users(db.Model):
    __tablename__ = 'Users'
    user_id = db.Column(db.Integer(), primary_key = True)
    likes = db.Column(db.Integer())
    users_daily_posts = db.relationship('Users_daily_posts', lazy='dynamic', backref=db.backref('user', lazy='dynamic'))

For the third table, the expectation is that it has as primary keys a user_id and a date (say date is an integer), a column for likes, and a column that should keep track of which blog posts are on that particular date. When a post is liked, the like propagates to this table and to the Users table, but Blog_posts does not store it. I originally thought date should be a foreign key, but it turns out there is a uniqueness constraint on this in the original table, and they won't be unique in the Blog_posts table, so I would think the design would be something like:

class Users_daily_posts(db.Model):
    __tablename__ = 'Users_daily_posts'
    user_id = db.Column(db.Integer(), db.ForeignKey('Users.user_id', ondelete="CASCADE"), primary_key = True)
    date = db.Column(db.Integer(), primary_key = True)
    likes = db.Column(db.Integer())
    blog_posts = ?

For blog_posts, I tried:

db.relationship('Blog_posts', lazy='dynamic', backref=db.backref('posts', lazy='dynamic'))

thinking that this might allow me to add a list of blog posts, but this gives the error:

ArgumentError: Could not determine join condition between parent/child tables on relationship Users_daily_posts.blog_posts.  Specify a 'primaryjoin' expression.  If 'secondary' is present, 'secondaryjoin' is needed as well.

I've seen a couple of posts about this error, but I really don't understand what the primaryjoin expression would be, as I'm about the furthest you can get from a databases whiz. Could somebody explain what the primary join would be and why? Or at least what I'm doing wrong? If there is a different SQLAlchemy idiom that is better suited to this purpose, I'd love to hear about it. Thanks so much for your time!

Answer 1

There are many cases that you need an association table between two classes. But looking at your case, I believe that you need just a column_property or do a simple query is enough.

In your case, you will have:

• Each post have just a one creator or author, so is Many to one relationship;

• The "likes" can't no be on users, but you can have a
  column property that count how many likes;

• Finally, "likes" is more related to the visitors and posts. So you
  will need another model in order to keep information about the
  visitors. And then make a relationship with users.

Try this:

from flask import Flask
from flask.ext.sqlalchemy import SQLAlchemy

import datetime

app = Flask(__name__)
app.config['SQLALCHEMY_DATABASE_URI'] = 'sqlite:////tmp/test.db'
db = SQLAlchemy(app)

class Blog_post(db.Model):
    __tablename__ = 'Blog_posts'
    post_id = db.Column(db.Integer, primary_key = True)
    creator_id = db.Column(db.Integer, db.ForeignKey('Users.user_id'), nullable=False)
    headline = db.Column(db.String(255), nullable=False)
    date = db.Column(db.Date, nullable=False)
    likes = db.Column(db.Integer)
    def __init__(self, headline, body,creator_id):
        self.headline = headline
        self.creator_id = creator_id
        self.date = datetime.datetime.today() 

class User(db.Model):
    __tablename__ = 'Users'
    user_id = db.Column(db.Integer(), primary_key = True)
    user_name = db.Column(db.String(80), unique=True)
    users_posts = db.column_property(
        db.select([db.func.count(Blog_post.post_id)]).\
            where(Blog_post.creator_id==user_id).\
            correlate_except(Blog_post)
    )
    def __init__(self, user_name):
        self.user_name = user_name

Testing...

>>> db.create_all()
>>> u = User('dedeco')
>>> db.session.add(u)
>>> db.session.commit()
>>> 
>>> p = Blog_post('I am dedeco 1','About dedeco 1',u.user_id)
>>> db.session.add(p)
>>> p = Blog_post('I am dedeco 2','About dedeco 2',u.user_id)
>>> db.session.add(p)
>>> p = Blog_post('I am dedeco 3','About dedeco 3',u.user_id)
>>> db.session.add(p)
>>> db.session.commit()
>>> u.users_posts
3

Query:

>>> d = db.session.query(db.func.count(Blog_post.post_id),db.func.strftime('%d/%m/%Y', Blog_post.date)).filter(Blog_post.creator_id==u.user_id).group_by(db.func.strftime('%d/%m/%Y', Blog_post.date))
>>> d.all()
[(3, u'04/11/2015')]

You can see some information about Many to Many on Sqlalchemy documentation