如何使用SQLAlchemy获取与三元组条件列表匹配的行

Question

Having a list of 3-tuples :

[(a, b, c), (d, e, f)]

I want to retrieve all the rows from a table where 3 columns matches the tuples. FOr this example, the query WHERE clause could be something like this :

   (column_X = a AND column_Y = b AND column_Z = c)
OR (column_X = d AND column_Y = e AND column_Z = f)

How can I create such a request using SQLAlchemy ? In my case the 3-tuples list will contains hundred of elements, and I'm looking for the best scallable solution.

Thanks for your help,

Answer 1

Easiest way would be using SQLAlchemy-provided tuple_ function:

from sqlalchemy import tuple_

session.query(Foo).filter(tuple_(Foo.a, Foo.b, Foo.c).in_(items))

This works with PostgreSQL, but breaks with SQLite. Not sure about other database engines.

Fortunately there's a workaround that should work on all databases.

Start by mapping out all the items with the and_ expression:

conditions = (and_(c1=x, c2=y, c3=z) for (x, y, z) in items)

And then create an or_ filter that encloses all the conditions:

q.filter(or_(*conditions))

Here's a simple example:

#/usr/bin/env python
from sqlalchemy import create_engine
from sqlalchemy import Column, Integer
from sqlalchemy.sql import and_, or_
from sqlalchemy.orm import sessionmaker
from sqlalchemy.ext.declarative import declarative_base

engine = create_engine('sqlite:///')
session = sessionmaker(bind=engine)()
Base = declarative_base()

class Foo(Base):
    __tablename__ = 'foo'

    id = Column(Integer, primary_key=True)
    a = Column(Integer)
    b = Column(Integer)
    c = Column(Integer)

    def __init__(self, a, b, c):
        self.a = a
        self.b = b
        self.c = c

    def __repr__(self):
        return '(%d %d %d)' % (self.a, self.b, self.c)

Base.metadata.create_all(engine)

session.add_all([Foo(1, 2, 3), Foo(3, 2, 1), Foo(3, 3, 3), Foo(1, 3, 4)])
session.commit()
items = ((1, 2, 3), (3, 3, 3))
conditions = (and_(Foo.a==x, Foo.b==y, Foo.c==z) for (x, y, z) in items)
q = session.query(Foo)
print q.all()
q = q.filter(or_(*conditions))
print q
print q.all()

Which outputs:

$ python test.py 
[(1 2 3), (3 2 1), (3 3 3), (1 3 4)]
SELECT foo.id AS foo_id, foo.a AS foo_a, foo.b AS foo_b, foo.c AS foo_c 
FROM foo 
WHERE foo.a = :a_1 AND foo.b = :b_1 AND foo.c = :c_1 OR foo.a = :a_2 AND foo.b = :b_2 AND foo.c = :c_2
[(1 2 3), (3 3 3)]

Answer 2

A less conventional approach that I suspect would scale well would be to create a temporary table of all your tuples and then join on that:

import sqlalchemy
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy import Column, Integer, Table
from sqlalchemy.orm import sessionmaker
Base = declarative_base()
engine = sqlalchemy.create_engine('sqlite:///:memory:')
Session = sessionmaker(bind=engine)
session = Session()

class Triple(Base):
    __tablename__ = 'triple'
    id = Column(Integer(), primary_key=True)
    x = Column(Integer())
    y = Column(Integer())
    z = Column(Integer())

ws_table = Table('where_sets', Base.metadata,
        Column('x', Integer()),
        Column('y', Integer()),
        Column('z', Integer()),
        prefixes = ['temporary']
    )

Base.metadata.create_all(engine)

...

where_sets = [(1, 2, 3), (3, 2, 1), (1, 1, 1)]
ws_table.create(engine, checkfirst=True)
session.execute(ws_table.insert(), [dict(zip('xyz', s)) for s in where_sets])
matches = session.query(Triple).join(ws_table, (Triple.x==ws_table.c.x) & (Triple.y==ws_table.c.y) & (Triple.z==ws_table.c.z)).all()

which executes SQL like this:

INSERT INTO triple (x, y, z) VALUES (?, ?, ?)
(1, 2, 3)
INSERT INTO triple (x, y, z) VALUES (?, ?, ?)
(3, 1, 2)
INSERT INTO triple (x, y, z) VALUES (?, ?, ?)
(1, 1, 1)
SELECT triple.id AS triple_id, triple.x AS triple_x, triple.y AS triple_y, triple.z AS triple_z 
FROM triple JOIN where_sets ON triple.x = where_sets.x AND triple.y = where_sets.y AND triple.z = where_sets.z

Answer 3

Would anyone consider creating an extra key in the original table ? ie create a new column with "1"-"2"-"3" instead of another table and check for the uniqueness.