Question

I’ve stuck with this SQLAlchemy-Python problem: here’s a class that should do simple update job in multiple threads

class InThreadUpdater():
    def __init__(self):
        self.portion_size = 5
        self.select_portion_func = db.session.query(FBPostStats).with_lockmode("update").limit(self.portion_size)

    def run(self):
        for post in self.select_portion_func.all():
            post.locked_by_thread = True
        db.session.commit()

if __name__ == "__main__":
    updater = InThreadUpdater()
    thread = threading.Thread(target=updater.run)

But after finish there’s no real changes committed to db. It works if I move initialization of
member select_portion_func to function run() like this

def run(self):
    self.select_portion_func = db.session.query(FBPostStats).with_lockmode("update").limit(self.portion_size)
    for post in self.select_portion_func.all():
        post.locked_by_thread = True
    db.session.commit()

What’s the difference between these two variants? Why the first gives some kind of copy for post?

Thanks in advance!

Answer 1

some kind of copy for post?

Because it is actually a copy – when new thread is created, updater variable is copied into another memory region in the heap on the new thread. This is probably a bad idea to share db-connections in such a way.