These are chat archives for ipython/ipython

21st
Apr 2015
singletS
@singletS
Apr 21 2015 00:18
@jeffhussmann , @jhamrick or Others, Any advice on changing keymaps in v3.1.0? My past hacks to custom.js don't seem to get me the emacs.js keymap anymore. Tried @jhamrick 's snippet from March 4, but no joy lately.
Isaac
@stuntgoat
Apr 21 2015 02:22
@singletS at ~/.ipython/profile_default/static/custom/custom.js https://gist.github.com/stuntgoat/ee43f7e406f300237628
@singletS that works for 4.0.0.dev :) your mileage may vary
Thomas A Caswell
@tacaswell
Apr 21 2015 02:34
@stuntgoat Awesome :smile: Seems to work with 3.1.0 (as shipped by conda)
singletS
@singletS
Apr 21 2015 03:26
@stuntgoat , @tacaswell WooHoo...it works for a pip3 install, too! Maybe soon there will be a simple, canonical way to choose keymaps in the profile? Something like: c.NotebookApp.keymap='emacs'? I suspect there are plans for fancy user settings in a gui form when time allows. At any rate, many thanks...Ctrl-a does what it's 'sposed to again!
Thomas A Caswell
@tacaswell
Apr 21 2015 03:48
The one I find more annoying is C-k which in chrome jumps to the address bar and relpaces what ever was there with '?'. My normal notebook workflow has been to do everything in an org-mode notebook and then copy-paste it over when I have to.
singletS
@singletS
Apr 21 2015 04:19
+1 for org-mode...I create there and manually convert to ipynb for students. A dream: an org-export or org-babel option for ipython notebook format. EIN is ok, but different goals. This is a step closer: https://github.com/rossant/ipymd/ . Guess I better learn more lisp...
@tacaswell In awe you use chrome; I'm always clobbering tabs with C-w. Grr. Thankfully the "leave page" warning now exists. I sooo wish browsers would let us prevent JS from hijacking keyboards...but I'm getting off-topic.
Dale Jung
@dalejung
Apr 21 2015 14:53
does the google drive integration fix the long running cell issue? i.e. executing a long running cell and closing the browser.
Thomas A Caswell
@tacaswell
Apr 21 2015 15:10
@singletS I have also thought about an org->ipynb converter, but I don't know enough lisp (or have enough time to learn) to work on it. It should 'just' be a hijacking the org -> md code and piping it to a json template.
Kyle Kelley
@rgbkrk
Apr 21 2015 15:17
@dalejung the google drive extension (jupyter drive) updates the notebook client side. If you close your tab, bye bye result propagation. Kernel still runs though.
@jasongrout I just watched your Advanced IPython widgets talk from last year; it was quite good
Bas Nijholt
@basnijholt
Apr 21 2015 15:18

I have IPython3 installed for Python 3 in order to work with Jupyterhub.

Now I'm able to use notebooks with a Python2 kernel, because I created /usr/local/share/jupyter/kernels/python2/kernel.json

with

 "argv": ["python2", "-m", "IPython.kernel",
          "-f", "{connection_file}"],
 "display_name": "Python 2",
 "language": "python2"
}

Now I would also like to use IPython.parallel, but when I start a cluster it will automatically start engines in Python 3, how can I change this to Python 2?

Dale Jung
@dalejung
Apr 21 2015 16:02
@rgbkrk thx, wasn't sure if the jupyter-drive was fundamental different than colab
Kyle Polich
@kylepolich
Apr 21 2015 16:42
Is there a preferred channel for submitting feature requests?
Matthias Bussonnier
@Carreau
Apr 21 2015 16:59
@kylepolich bug tracker or ML is fine.
Min RK
@minrk
Apr 21 2015 17:01
Damian Avila
@damianavila
Apr 21 2015 17:45
/@all... not at home now... have fun in the meeting!!
Jason Grout
@jasongrout
Apr 21 2015 17:46
at one point, they suggested using the seaborn color cycle for matplotlib :)
Jason Grout
@jasongrout
Apr 21 2015 17:56
We'd say that the requirements.txt is the transitive closure of the dependencies :)
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 18:08
@minrk has their been history changes in traitlets already?
Jonathan Frederic
@jdfreder
Apr 21 2015 18:09
I just remembered, we will also need to address the CSS problem when the widgets are moved into their own repo - we'll need a build file and some way to push the CSS to the page. For now, vanilla JS will work, but we may want to use require-css.
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 18:09
@jdfreder it is what we do for custom widgets
Jason Grout
@jasongrout
Apr 21 2015 18:10
Sylvain, you and I get pinged every time there is history rewriting courtesy of ipython/ipython@7799fea :)
(at least, whole-repo rewriting...)
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 18:11
actually, a few things are quite annoying with require. For some things, we replaced shims with explicit loading, because requirejs.config is global, and not specific to a given nb extension
Jason Grout
@jasongrout
Apr 21 2015 18:11
@jdfreder - how about require-less
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 18:12
so if two nb extensions define a path for say tagsinput, you might have surprises
Jonathan Frederic
@jdfreder
Apr 21 2015 18:14
Yeah @SylvainCorlay I had that problem yesterday - rendered itself as a timeout... : /
Yeah, that's also a possibility @jasongrout
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 18:15
we found that the in-browser less compilation was often faster
regarding the require shims, I don't know of a clean solution
Jason Grout
@jasongrout
Apr 21 2015 18:15
in-browser less compilation is probably not faster than precompiling css and serving that, right?
(especially if you minify the css)
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 18:15
surely not
Jason Grout
@jasongrout
Apr 21 2015 18:16
okay, I wasn't sure what you were saying it was faster than.
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 18:16
compilation vs compilation
in browser versus offline
Jason Grout
@jasongrout
Apr 21 2015 18:36
do you happen to have those earlier repos still around? Is there the possibility you can patch in history pre-2008?
@fperez ^
Matthias Bussonnier
@Carreau
Apr 21 2015 18:39
I'll ask.
Min RK
@minrk
Apr 21 2015 18:44
@SylvainCorlay yes, traitlets history has been rewritten. I think it's in a sensible state now.
Sylvain Corlay
@SylvainCorlay
Apr 21 2015 19:00
thanks
Just that I need to create a branch with the old commit for the projects having it as a submodule to find the commit number (in old revisions)
(found it)