Greenlet-based coroutines for tornado
Greenado is a utility library that provides greenlet-based coroutines for tornado. In tornado, coroutines allow you to perform asynchronous operations without using callbacks, providing a pseudo-synchronous flow in your functions.
When using gen.coroutine in tornado in a large codebase, you will notice that they tend to be ‘infectious’ from the bottom up. In other words, for them to be truly useful, callers of the coroutine should ‘yield’ to them, which requires them to be a coroutine. In turn, their callers need to ‘yield’, and so on.
Instead, greenado coroutines infect from the top down, and only requires the greenado.groutine decorator somewhere in the call hierarchy, but it doesn’t really matter where. Once the decorator is used, you can use greenado.gyield to pseudo-synchronously wait for asynchronous events to occur. This reduces complexity in large codebases, as you only need to use the decorator at the very top of your call trees, and nowhere else.
Installation & Requirements
$ pip install greenado
greenado should work using tornado 3.2, but I only actively use it in tornado 4+
I have only tested greenado on Linux & OSX, but I imagine that it would work correctly on platforms that tornado and greenlet support.
In the below examples, ‘main_function’ is your toplevel function in the call hierarchy that needs to call things that eventually call some asynchronous operation in tornado.
Normal tornado coroutine usage might look something like this:
from tornado import gen @gen.coroutine def do_long_operation(): retval = yield long_operation() raise gen.Return(retval) @gen.coroutine def call_long_operation(): retval = yield do_long_operation() raise gen.Return(retval) @gen.coroutine def main_function(): retval = yield call_long_operation()
With greenado, it looks something like this instead:
import greenado def do_long_operation(): retval = greenado.gyield(long_operation()) return retval def call_long_operation(): retval = do_long_operation() return retval @greenado.groutine def main_function(): retval = call_long_operation()
Functions wrapped by greenado.groutine return a tornado.concurrent.Future object which you must either yield, call result(), or use IOLoop.add_future on, otherwise you may risk swallowing exceptions.
Contributing new changes
- Fork this repository
- Create your feature branch (git checkout -b my-new-feature)
- Test your changes (tests/run_tests.sh)
- Commit your changes (git commit -am ‘Add some feature’)
- Push to the branch (git push origin my-new-feature)
- Create new Pull Request