These are chat archives for ipython/ipython

12th
Sep 2014
Brian E. Granger
@ellisonbg
Sep 12 2014 02:42
@minrk let me know when you are around - I am trying to get jupyterhub running - I think the README is missing some steps
Kyle Kelley
@rgbkrk
Sep 12 2014 16:24
@takluyver How does the Python3 install from master work?
If I run pip3 install .[all] it's not happy
If I just pip3 install ., it's missing dependencies for jsonschema and jsonpointer.
I just pip3 installed those directly for now
Screen Shot 2014-09-12 at 11.26.14 AM.png
Thomas Kluyver
@takluyver
Sep 12 2014 16:34
What goes wrong with .[all]? @rgbkrk
Kyle Kelley
@rgbkrk
Sep 12 2014 16:36
It thinks it is downloading a package named .[all]
$ sudo docker build -t ipython/latest .
Sending build context to Docker daemon 96.19 MB
Sending build context to Docker daemon
Step 0 : FROM ipython/scipystack
 ---> 5cf6d689c955
Step 1 : MAINTAINER IPython Project <ipython-dev@scipy.org>
 ---> Using cache
 ---> 386739a33073
Step 2 : ADD . /srv/ipython/
 ---> a942c169d130
Removing intermediate container 00df162f31bc
Step 3 : WORKDIR /srv/ipython/
 ---> Running in 6f9858522adc
 ---> 3461a424ba12
Removing intermediate container 6f9858522adc
Step 4 : RUN pip3 install jsonschema jsonpointer
 ---> Running in e9c77c702665
Downloading/unpacking jsonschema
  Downloading jsonschema-2.4.0-py2.py3-none-any.whl
Downloading/unpacking jsonpointer
  Downloading jsonpointer-1.4-py3-none-any.whl
Installing collected packages: jsonschema, jsonpointer
Successfully installed jsonschema jsonpointer
Cleaning up...
 ---> 9e6ddb42b8d7
Removing intermediate container e9c77c702665
Step 5 : RUN pip2 install .[all]
 ---> Running in a24edbde2cd7
Downloading/unpacking .[all]
  Could not find any downloads that satisfy the requirement .[all]
Cleaning up...
No distributions at all found for .[all]
Storing debug log for failure in /.pip/pip.log
2014/09/12 16:35:52 The command [/bin/sh -c pip2 install .[all]] returned a non-zero code: 1
Thomas Kluyver
@takluyver
Sep 12 2014 16:37
silly pip...
Kyle Kelley
@rgbkrk
Sep 12 2014 16:37
yeah, could have sworn I've done it this way before
oh wait
I usually use -e when doing .[all]
But since this wasn't meant for editing I took that flag off
Thomas Kluyver
@takluyver
Sep 12 2014 16:37
I think it should work without, but...who knows
Kyle Kelley
@rgbkrk
Sep 12 2014 16:38
Yup, that made it work
I'll just leave the -e flag
Thomas Kluyver
@takluyver
Sep 12 2014 16:38
ok
Kyle Kelley
@rgbkrk
Sep 12 2014 16:38
Current issue now is getting anything to run in a kernel...
It just sits there hanging
Interesting.
After running pip3 install -e .[all], ipython2 no longer exists
Step 7 : RUN ipython2 kernelspec install-self
 ---> Running in 3d299541db23
Traceback (most recent call last):
  File "/usr/local/bin/ipython2", line 9, in <module>
    load_entry_point('ipython==3.0.0-dev', 'console_scripts', 'ipython2')()
  File "/usr/lib/python2.7/dist-packages/pkg_resources.py", line 351, in load_entry_point
    return get_distribution(dist).load_entry_point(group, name)
  File "/usr/lib/python2.7/dist-packages/pkg_resources.py", line 2362, in load_entry_point
    raise ImportError("Entry point %r not found" % ((group,name),))
ImportError: Entry point ('console_scripts', 'ipython2') not found
2014/09/12 16:38:29 The command [/bin/sh -c ipython2 kernelspec install-self] returned a non-zero code: 1
Thomas Kluyver
@takluyver
Sep 12 2014 16:40
oh, yeah
-e is equivalent to the develop command in setuptools
which breaks having commands for different versions installed
Kyle Kelley
@rgbkrk
Sep 12 2014 16:41
Maybe I'll open a PR with this so you can comment on the mess I'm making
ah
Thomas Kluyver
@takluyver
Sep 12 2014 16:41
I hate setuptools
Kyle Kelley
@rgbkrk
Sep 12 2014 16:41
You do?
And here I thought humble setuptools was an homage to your favorite package.
Thomas Kluyver
@takluyver
Sep 12 2014 16:41
we wrote a 'setup.py symlink' option to work around that particular bit of stupidity
but it doesn't get you dependency installation
I think you'll have to install dependencies as a separate step
Kyle Kelley
@rgbkrk
Sep 12 2014 16:42
This may be an issue on master I'm facing actually
Screen Shot 2014-09-12 at 11.43.02 AM.png
For the "why won't my code run"
Thomas Kluyver
@takluyver
Sep 12 2014 16:43
one moment, I'll try locally
Kyle Kelley
@rgbkrk
Sep 12 2014 16:44
I thought it was exec nonsense so I wrote a wrapper script. I'll submit this as a [WIP] PR for the moment
Thomas Kluyver
@takluyver
Sep 12 2014 16:45
Interesting. I see "InvalidStateError: An attempt was made to use an object that is not, or is no longer, usable"
works after a refresh. Check for browser caching issue.
s
Kyle Kelley
@rgbkrk
Sep 12 2014 16:46
ah,
ah, k
ipython/ipython#6457
yay for caching
I'm going to send you a link
This is my first time playing with the new LaTeX -> Unicode mode
that's rad
Thomas Kluyver
@takluyver
Sep 12 2014 16:51
:)
Kyle Kelley
@rgbkrk
Sep 12 2014 16:51
I really only wanted Python 3 in the base image so I could try that mode out.
:P
Thomas Kluyver
@takluyver
Sep 12 2014 16:51
heh
we'll take any excuse ;-)
Kyle Kelley
@rgbkrk
Sep 12 2014 16:54
Now, this kernel issue
I need to run python setup.py symlink?
I posted about it in #6457
Thomas Kluyver
@takluyver
Sep 12 2014 16:55
setup.py symlink is equivalent to setup.py develop and pip install -e.
But it doesn't install dependencies, so for this purpose it's no better than doing setup.py install, or pip install .
I think the easiest thing is to install dependencies as a separate step (pip install blah), and then install ipython.
Kyle Kelley
@rgbkrk
Sep 12 2014 16:58
Ok, I can do that. We have most of them already. The only ones that were missing were jsonschema and jsonpointer
Thomas Kluyver
@takluyver
Sep 12 2014 16:58
OK
Kyle Kelley
@rgbkrk
Sep 12 2014 16:58
But then if we need updated dependencies we'll have to update the Dockerfile
Thomas Kluyver
@takluyver
Sep 12 2014 16:58
unfortunately, yes
Kyle Kelley
@rgbkrk
Sep 12 2014 16:58
That's a fine route to me right now
Thomas Kluyver
@takluyver
Sep 12 2014 17:00
cool
@ellisonbg @damianavila - decision meeting now? I'll host a hangout
Damian Avila
@damianavila
Sep 12 2014 17:03
ok
Thomas Kluyver
@takluyver
Sep 12 2014 17:04
invitations should have gone out
Kyle Kelley
@rgbkrk
Sep 12 2014 17:05
Did you call me too just now?
I say yes to all my PRs
Thomas Kluyver
@takluyver
Sep 12 2014 17:05
I just called everyone on the team
Kyle Kelley
@rgbkrk
Sep 12 2014 17:05
I tried to answer, it was weird
Thomas Kluyver
@takluyver
Sep 12 2014 17:05
hmm, strange
do you want a new invite
Kyle Kelley
@rgbkrk
Sep 12 2014 17:06
yeah
I think it conflicted with a call that is in progress with SWC + Mozilla folks
Even though I exited that call
Or just send me the link that goes to the hangout
Damian Avila
@damianavila
Sep 12 2014 17:07
I did not get any invitation...
can you post the link of the hangout?
I invited you both again as well
Kyle Kelley
@rgbkrk
Sep 12 2014 17:09
That will let us watch, not participate
Matthias Bussonnier
@Carreau
Sep 12 2014 17:09
If you post link to pr in this chat, I can try to look from my phone while commuting :-P
Matthias Bussonnier
@Carreau
Sep 12 2014 17:13

5659 I think one of us could rebase and that we should accept small fixes for 3.0

Pr #6106 no, feature creep. Maybe once widget et separate repo and people took care of then.
Jason Grout
@jasongrout
Sep 12 2014 17:24
@minrk: did you originally do the notebook trust framework?
Min RK
@minrk
Sep 12 2014 17:24
I did much of the implementation
Jason Grout
@jasongrout
Sep 12 2014 17:24
It didn't work until I fixed some bugs :) ipython/ipython#6460
Min RK
@minrk
Sep 12 2014 17:24
the design came from a dev meeting
ah, then it was broken by the require-refactor
it definitely did work
Jason Grout
@jasongrout
Sep 12 2014 17:25
Sure; I'm not sure when it broke. You can see there were a few small problems.
Anyways, I think that is ready for review. It's ironic there aren't tests to cover the trust framework :).
Min RK
@minrk
Sep 12 2014 17:26
Trust actually works just fine
Jason Grout
@jasongrout
Sep 12 2014 17:26
can you set a notebook to be trusted using the menu?
for me, it seemed that a notebook always had undefined trust
because the event handler for the trust change always set it to undefined
Min RK
@minrk
Sep 12 2014 17:27
and has several tests, it's just a UI element (by far the rarest way to interact with trust) didn't get updated properly
Jason Grout
@jasongrout
Sep 12 2014 17:28
huh. I couldn't get a notebook to be trusted when I tried.
Min RK
@minrk
Sep 12 2014 17:28
Did you run it?
Jason Grout
@jasongrout
Sep 12 2014 17:28
even without the UI. Yes, I ran it, and then I cleared the cells too. The IPython.notebook.trusted attribute stayed undefined
Because when setting a notebook to be trusted, it invoked the event handler, and the event handler promptly set the trust to undefined
Min RK
@minrk
Sep 12 2014 17:29
and refresh the page?
Jason Grout
@jasongrout
Sep 12 2014 17:30
So: open a notebook, run all the cells, save the notebook, and then refresh the page?
Min RK
@minrk
Sep 12 2014 17:30
That ought to get around the recently introduced bugs your PR fixes
Jason Grout
@jasongrout
Sep 12 2014 17:30
I don't think it will, but I'll try again
nope, doesn't work.
because when the notebook is initially set to be trusted, the event handler is invoked, which promptly sets the trust status to undefined.
You can try it if you don't believe me :)
open up a notebook that has undefined trust
Min RK
@minrk
Sep 12 2014 17:34
I just did
it worked
Jason Grout
@jasongrout
Sep 12 2014 17:34
(the event handler at https://github.com/jasongrout/ipython/blob/trust/IPython/html/static/notebook/js/notebook.js#L190 sets the trust to undefined because it uses data.value, but data is true in the old version
was your notebook initially trusted?
Min RK
@minrk
Sep 12 2014 17:34
nope
Jason Grout
@jasongrout
Sep 12 2014 17:34
hmmm....
Min RK
@minrk
Sep 12 2014 17:34
But in any case, your PR fixes real UI bugs - it doesn't matter if I can avoid running into them
Jason Grout
@jasongrout
Sep 12 2014 17:35
(I guess any odd thing about the trust framework sets up red flags in my mind)
Min RK
@minrk
Sep 12 2014 17:35
it fails with Uncaught TypeError: Cannot read property 'on' of undefined
Jason Grout
@jasongrout
Sep 12 2014 17:35
ah, right. I fixed that bug, and then the next bug was manifest.
Min RK
@minrk
Sep 12 2014 17:36
It's just a failure of the UI to actually trigger 'Trust this notebook'
Jason Grout
@jasongrout
Sep 12 2014 17:36
So you shouldn't end up with a trusted notebook, then.
That's odd that your notebook ended up trusted.
Min RK
@minrk
Sep 12 2014 17:37
It set the value, it just didn't trigger the secondary event (save) to push trust to the server because of the typo
which I did manually by hitting save
Jason Grout
@jasongrout
Sep 12 2014 17:39
Did you click the Trust Notebook menu, or did you just evaluate the cells?
Min RK
@minrk
Sep 12 2014 17:39
both
both worked
had to click save after Trust menu, because of the typos you fixed
Jason Grout
@jasongrout
Sep 12 2014 17:40
Okay; I don't see how just running the cells and then saving would have worked.
But anyways, thanks for reviewing the PR.