Tornado is intriguing, but the main thing I don't like about it is how it reinvents the wheel with a lot of its components. For example:

- They created a lot of Web-stuff parsing code themselves, when Werkzeug provides a tested and thorough implementation of a lot of that. (Especially routing, regex-based routing just looks horribly kludgy once you have used werkzeug.routing.)

- They created their own template language when they could have used Jinja. (In fact, tornado.template is basically a half-as-powerful copy of Jinja.)

- They created their own database access layer when they could have used SQLAlchemy instead.

Sometimes I think people take the concept of "minimal dependencies" way too far.

tornado.database has no real ties to the rest of the system; you can (and I usually do) use SQLAlchemy or another database layer instead. You can also call another template engine if you prefer, but you're sort of stuck with the url routing.

I'd be curious as to how many people actually use it. Since it's MySQL specific, it holds little value for anyone not using MySQL ;-)

20% of survey respondents reported using it in february: http://dl.dropbox.com/u/19694335/tornado/Tornado%20usage%20s...

If you bring in tons of third party code you end up with a mess like Pylons. For people new to a framework, it's a pain to learn that when a problem arises you need to go look up the docs for a separate framework.

Or you end up with something really solid and enjoyable to work with, like Flask. Which is just a combination of Werkzeug and Jinja with some glue and excellent documentation.

90% made up my mind to use Tornado for several upcoming projects, would love to hear some comments (good or bad) from anyone thats used it or using it.

I've used both Django and Tornado and I vastly prefer Tornado.

Some personal experiences: With Django, I specifically don't like ORM and it's template language. Also, I thought the admin interface would reduce the need to write input UIs, but at the end of the day, it's not that suitable for end-user when you have several end-user roles with different access right patterns.

If you (read: I) are going to replace Django's template language and leave out ORM and admin interface out of Django, Django doesn't really shine anymore and simpler frameworks like Tornado provide cleaner base for your development.

Hm, you could take a look at pyramid, if you dont like djangos orm or tempalting it might just be right for you.

When speed isn't important and you know you want to use a SQL database; then use Django. For everything else, use Tornado. :)

when he says Django he really means Flask :-p

I've noticed that Tornado comes with its own MySQL wrapper, however I'm concerned it might have blocking issues.

Does anyone have experience with using this ?

Also, what do people use for form validation with Tornado ?

The theory is, you shouldn't need async database access. If your database is a bottleneck, you are screwed, regardless of how your app layer is scaling.

The main advantage of async is not blocking if you call an external web API.

That sounds wrong. You need async database access so that while you're waiting for a query to finish, you can still process other requests.

Imagine an app that receives an HTTP request every second. One in five of these requests requires making a query that takes 3 seconds to complete. The rest is serviced immediately. You don't want the long request freezing your entire server and preventing the fast requests from being serviced.

From this thread https://groups.google.com/forum/#!topic/python-tornado/mgj18... (posting here because Google Groups sometimes require login):

  We experimented with different async DB approaches, but settled on 
  synchronous at FriendFeed because generally if our DB queries were 
  backlogging our requests, our backends couldn't scale to the load 
  anyway. Things that were slow enough were abstracted to separate 
  backend services which we fetched asynchronously via the async HTTP 
  module. 

  I may open source the async MySQL client I wrote, but I am still 
  skeptical of the long term value given the code complexity it 
  introduces. 

  Bret

Erm, so if things are fast it doesn't matter and if things are slow you should write a whole new service, put a HTTP interface into it and use that? Not sure I buy that :)

Disclaimer: I wrote an async DB module for Twisted/PostgreSQL and it did not turn out to be all that complex.

There's async libraries here: https://github.com/facebook/tornado/wiki/Links - You can get just about any db except MySQL. There's ones for Couch, Mongo, HBase, a bunch of other No-SQLs and Psycopg (PostGres).

Also, Tornado is now introducing some kind of interface to Twisted's event loop, which does have everything ported.

Bret just doesn't seem to want async My-SQL in the main code base.

Not sure friendfeed ever got enough users to know scaling issues...

Personally I would go with async DB, or have DB access in its own thread(s) and async communicate with that.

From Wikipedia: [FriendFeed] had on average one million monthly visitors ... [and] was bought for $15 million in cash, and $32.5 million in Facebook stock.

I figure after the first million users, your embarrassingly parellel app server layer is not your bottleneck, it's getting the database to scale.

The impact of an occasional slow query is mitigated by the fact that you're running multiple processes. You need multiple processes anyway because of the GIL, so just run a few more to make sure you're able to keep the CPU busy. An async DB interface would be nice, but as long as you keep your database performance under control it's not crucial, so it hasn't been written yet.

I am curious about what kind of libraries are compatible with Tornado as well. Is there a list that someone maintains? I am specifically interested in things that can bring a bit of the goodies that django provides.

I've recently been blogging about my tornado-utils library which, for example, brings a email sending library similar to Djangos http://www.peterbe.com/oc-Tornado

I wrote a web service in Tornado around a year ago, pre-1.0. It's a very nice, albeit minimal framework. In terms of comparisons to others:

* it feels easier to scale up than Flask, but that could be my minimal Flask experience showing. Certainly, it is more performant than basic Flask installations and configurations.

* it makes it harder to separate your concerns than, say, Django. I've been working in Django for a long time, and while I have definite issues with the "app"-style packaging (ie. if I add an external forum framework, for instance, I need to munge my "profiles" app to support it), it does feel easier to separate your concerns in Django than in most other frameworks I've used.

* it's fast, and reasonably agnostic - the MySQL DB layer, for instance, can be completely eschewed for something else.

I imagine they've added a lot to it since then, and the service itself isn't something I've looked at for a while (failed startup idea). But from what I used back then, I'd definitely recommend it.

I'm a convert from Twisted.

It is simple enough, the code is clear and small enough to dive into if you have any problems, and it works just great.

I too am in the Tornado's camp. I love its simplicity. It has decent documentation. And if there's something you don't understand you just peek at the source. I tend to write my code independent of tornado, so if the need arises I can switch to a different framework.

Easy to get started, easy to deploy, pythonic/magic-free code, responsive maintainer.

Coming from php(mvc background), the only thing i don't like about tornado is that basically you have a Class per request, currently i'm in a process to refactor this so that:

www.site.com/controller/action/params/?vars -> would route to ->

class controller(requestHandler):

    def get_action(self, *params, **vars):
        #do something
        self.write(response)
    def  post_action2(self, *params, **vars):
        #.... and so on ....

I don't get it. Your example is what Tornado looks like::

  class LoginHandler(RequestHandler):
      def get(self):
          self.render('login.html')
      def post(self):
          self.set_secure_cookie(self.get_argument('username'))
        self.redirect('/')

Having one class per URL is very common and perhaps it's more than a decade of Python web programming but I like that. Makes for neat code.

The one class per action pattern helps with async calls and callbacks. The handler class has methods for each HTTP method plus the callbacks it needs.

That would get really messy in a controller-with-multiple-actions pattern as the controller classe now has a lot more methods in a single class: all the HTTP methods it handles plus the callbacks for all of them.

i dont agree with you, i would rather have a controller with multiple actions, rather then "controller"/requestHandler per action, because in most cases you want the controller to handle all the actions of a page, rather then one action.

Having controller per one action would put overhead on the code lines you write.

Im a PHP developer who uses Tornado and I recently wrote a PHP micro framework that works like tornado's controllers where rest verbs were the only publicly (uri) accessible methods

For a multiplayer, realtime, text-based game, like a quiz, should I dive in nodejs or Tornado?

I know Python but never played with server-side javascript before.

It all boils down to what you prefer. They're both relatively mature software packages that are being used by tons in production.

Whichever you choose, you may find that http://socket.io/ saves you some time by inventing various wheels you may have needed to reinvent (for use with tornado, see the tornadio project).

Thank you! Didn't know about tornadio.

I'm especially looking forward to trying the new interface to Twisted and the gen module. I was about to write some async interfaces to things that are currently not - these new modules look like they'll be helping me out quite a bit.