saghul / evergreen Goto Github PK

Hi saghul,

I've just installed evergreen to have a look and perform basic tests:

cat /etc/issue

Ubuntu 14.04.1 LTS

pip install evergreen

Downloading evergreen-0.1.0.tar.gz
Downloading fibers-0.4.0.tar.gz
Downloading pyuv-1.0.1.tar.gz

but I've been stopped at first step:

python

import evergreen
loop=evergreen.EventLoop()

Traceback (most recent call last):
File "", line 1, in
File "/home/francois/.local/share/virtualenvs/py2/local/lib/python2.7/site-packages/evergreen/core/loop.py", line 126, in init
self._ready_processor.unref()
AttributeError: 'pyuv.Check' object has no attribute 'unref'

I thought probably an evergreen/pyuv version mismatch :)
But now that I checked, there seems to be a single version of pyuv available on pypi:

pip install yolk
yolk -V pyuv

pyuv 1.0.1

If you can have a look ?

François

when i using evergreen == 0.1.0 , i install it with pip

pip install evergreen
and it install pyuv == 1.3.0, and it try this example:

#!/usr/bin/env python
# encoding: utf-8
import evergreen

# Create the global loop
loop = evergreen.EventLoop()

# Start the loop
loop.run()

it get following error:

Traceback (most recent call last):
  File "test.py", line 7, in <module>
    loop = evergreen.EventLoop()
  File "/home/yetship/.virtualenvs/evergreen/lib/python2.7/site-packages/evergreen/core/loop.py", line 126, in __init__
    self._ready_processor.unref()
AttributeError: 'pyuv._cpyuv.Check' object has no attribute 'unref'

I really dislike the complete future approach in evergreen. Generally the concept is great (compared to callback passing style), but in this case you are mixing two programming models: greenlets vs. futures.

I would like to see that threads and processes have the same interface than a task. So you would do:

t = loop.spawn_[task|thread|process](func, *args, **kwargs)
t.join()
t.result
t.get()  # shortcut to join + result

p = loop.spawn_ext_process(cmd_args, stdout=True)
for line in p.stdout:
    print line

Internally you should use libuv's thread pool and process capabilities which should be more efficient than the concurrent.futures implementation.

Hi there,

I'm using Evergreen in a project of mine, and I'm running into a strange edge case. Long story short, I need to read from a TUN/TAP device on Mac OS X. If you do pretty much anything except calling select() on the device, you get errno = 35. So, I tried calling select() from a regular Thread, reading the data that shows up, and then putting it on a queue (see here). However, I get an RuntimeError('there is no event loop created in the current thread'), which makes sense. However, I don't know of any other way to do this - sending data to a task from another thread without blocking the entire event loop. Any suggestions?

Thanks!

There are multiple test cases dead-locking on Windows:

• test_future_wait_multiple_wait_first (test_futures.FuturesTests)
• test_stop2 (test_loop.LoopTests)
• test_task_join_timeout (test_tasks.TasksTests)
• test_timeout (test_timeout.TimeoutTests)

Any idea?

Hi,
Just find out that in your documentation, the design part, it mentioned that the Task class is implemented using the greenlet library.
But in the tasks.py, the Task class is using python-fibers instead of greenlet.
I think some people will be confused when reading the document and tracing the source code.
Just to notify this small issue, not sure it is the right place to report this.