These are chat archives for ipython/ipython

12th
Aug 2015
Min RK
@minrk
Aug 12 2015 00:56
@/all IPython 4.0 is about to be released, so we will finally be done with the first complete release of IPython/Jupyter.
Matthias Bussonnier
@Carreau
Aug 12 2015 00:57
:cake:
let's celebrate with notebook 4.1 !
Younggun Kim
@scari
Aug 12 2015 00:58
Nice!
Zachary Jones
@bigzachattack
Aug 12 2015 00:58
:+1: :+1:
Jonathan Frederic
@jdfreder
Aug 12 2015 01:14
:D
K.-Michael Aye
@michaelaye
Aug 12 2015 01:43
:sparkles:
Steven Silvester
@blink1073
Aug 12 2015 02:04
Well done!
Adriano
@rsvp
Aug 12 2015 02:46
TADA.jpynb !!
Damian Avila
@damianavila
Aug 12 2015 02:52
\o/
Sylvain Corlay
@SylvainCorlay
Aug 12 2015 03:06
:+1:
Ai MAKABI
@amacbee
Aug 12 2015 05:09
:clap:
Andreas Klostermann
@akloster
Aug 12 2015 10:38
is there a recommended way to install the Jupyter development version?
Andreas Klostermann
@akloster
Aug 12 2015 10:49
I think I found it... I was expecting more trouble though
Andreas Klostermann
@akloster
Aug 12 2015 13:43
I'm trying to add an asyncio integration to the gui integrations in the "ipykernel" repository, but I can't figure out how to run "ipython kernel" using my ipykernel clone...
Andreas Klostermann
@akloster
Aug 12 2015 14:52
No, I guess that was the entirely wrong route to go down.....
Andreas Klostermann
@akloster
Aug 12 2015 15:58
I'm also trying @takluyver 's aiokernel, but it's very rough.
Min RK
@minrk
Aug 12 2015 17:02
A dev install of ipykernel should be just like any other Python package - clone ipython/ipykernel and pip install -e .
Andreas Klostermann
@akloster
Aug 12 2015 17:22
ok
I was prodding at the asyncio stuff again... ZMQ runs on a Tornado loop, and Tornado can theoretically run on an asyncio Loop, but without changing ipython and pyzmq I probably can't let the normal IPyKernel run on asyncio. (For integration into blender and kivy, and running asyncio stuff in the notebook)
Min RK
@minrk
Aug 12 2015 17:26
it shouldn't need any change to pyzmq, but IPython's startup can probably be changed to run asyncio on aiozmq, and tornado on asyncio.
There may need to be other changes to IPython beyond that, though.
Andreas Klostermann
@akloster
Aug 12 2015 17:27
ok
the more fundamental problem I've discovered in blender is that I can't easily install pyzmq because of cython
so I think I am pretty much stuck there
Min RK
@minrk
Aug 12 2015 17:37
you don't need cython for pyzmq, only to develop it
Or do you mean it can't handle compiling anything at all?
Sylvain Corlay
@SylvainCorlay
Aug 12 2015 18:01
Hello, I wanted to bring up the series of PR that we postponed in mid-june to after 4.0 regarding the comm_info_[request/reply] shell message
jupyter/jupyter_client#34, ipython/ipykernel#25, jupyter/notebook#166
Using the shim for that message that is currently implemented with comms in ipywidgets, I realized we may want to add a target_name attribute to the request message, which, when provided, triggers a reply that only contains comm info for the specified target.
Min RK
@minrk
Aug 12 2015 18:07
that makes sense
Sylvain Corlay
@SylvainCorlay
Aug 12 2015 18:13
Another question is whether a discussion on adding a request/reply type of message between comms would be in scope (only from the front-end to the back-end). The one usecase we have is requesting widget state from the front-end.
For each comm, we could register a handler to a request type (specified by key in the shell message)
Min RK
@minrk
Aug 12 2015 18:34
I guess it's a question of whether to do that for all comms or do that in a specific on_msg handler, which you could do right now.
Sylvain Corlay
@SylvainCorlay
Aug 12 2015 18:35
the thing is that you cannot register a callback to a specific message id at the moment with comm messages, unless you maintain a dictionary in the general handler
Min RK
@minrk
Aug 12 2015 18:39
which is exactly what an internal implementation would do anyway.
do you want to sketch out a real use case that works now, and a version that would be better with your proposal?
Hamzeh Alsalhi
@hamsal
Aug 12 2015 18:49
Is there a way to make ipython notebook outputs wider to better fill widescreen monitors?
Min RK
@minrk
Aug 12 2015 18:49
It can be done with CSS, I'm not sure exactly what, though.
Sylvain Corlay
@SylvainCorlay
Aug 12 2015 19:08
@minrk sorry I was out to get some lunch
yes, I saw the internal implementation already does that. I was just thinking that it is better to reuse the existing machinery from shell [reply/request]
@jdfreder also implemented the same logic as part of the widget model for the request state.
What I was doing is move the state request outside of the widget model because the _view_name and _model_name are part of the model state and need to be retrieved from the backend. At the moment we rely on the local storage.
Min RK
@minrk
Aug 12 2015 19:13
Probably so. I just like to try to prototype some of these things in less stable areas like widgets, then promote to general comms if they make more sense there.
Sylvain Corlay
@SylvainCorlay
Aug 12 2015 19:14
i.e. I need to request the state for a widget before I instantiate the model.
it makes sense.
ok
David Warde-Farley
@dwf
Aug 12 2015 20:19
So, IPython autocomplete and namespace packages...
mila-udem/blocks-extras#21
David Warde-Farley
@dwf
Aug 12 2015 20:28
@minrk any idea if this is even supposed to work, according to IPython?
I seem to recall that this is why scikits.* switched away from them.
Matthew Rocklin
@mrocklin
Aug 12 2015 21:29
In the notebook when a user elects to interrupt the kernel what happens? Is there a way that I can intercept this, gracefully cleanup active threads, and then reraise?
Jason Grout
@jasongrout
Aug 12 2015 23:33
Is Jupyter 4.0 actually released?
or is it still in RC stage?
It hit hackernews as being released, because of the blog post: https://blog.jupyter.org/2015/08/12/first-release-of-jupyter/
but conda doesn't seem to have the package yet, despite what the docs indicate: http://jupyter.readthedocs.org/en/latest/install.html#installing-the-jupyter-notebook
okay, pypi reports jupyter 1.0.0 and ipython 4.0.0 out. So I guess it's conda that's lagging.
it'll probably be up before long