Question

I am writing a server application using Tornado web server. Application works with a MySQL database through tornado.database module. In the DB there is a table

CREATE TABLE SampleTable (
    id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
    txt VARCHAR(256)
);

As a reaction on certain queries I need to write a new record to this table and return the id of newly created record to client. So in the server code I have these lines:

class Application(tornado.web.Application):
    def __init__(self):
        handlers = [(r'/somequery', queryHandler)]
        tornado.web.Application.__init__(self, handlers)
        self.db = tornado.database.Connection(
            host="localhost", database="sampledb",
            user="sampleuser", password="samplepassword")

and a handler:

class queryHandler(tornado.web.RequestHandler):
    def post(self):
        lastid = self.db.execute_lastrowid("INSERT INTO \
                                            SampleTable (txt) \
                                            VALUES('some text');")
        print lastid

As far as I know, tornado.database is a relatively simple wrapper around MySQLdb python module, which guarantees correct work of last row id for each connection only. In my case there is only one connection for the whole applcation.

So, in case of several concurrent client queries to /somequery is it possible that last row id would mess between clients or tornado takes care of it?

I tried looking into source of tornado.database, but found nothing regarding my question.

If this wouldn’t work fine, how to do?

Answer 1

execute_lastrowid (actually tornado.database in general) is not asynchronous, that is, it will block your application.

So, you won’t get the id from another client’s request, but if one insert takes a long time, other clients will have to wait for it.

Not related to your question, but typically data changes (e.g. INSERT) should not be accessible via GET requests. Use POST instead.