These are chat archives for ipython/ipython

22nd
Apr 2015
Fernando Perez
@fperez
Apr 22 2015 00:37
@jasongrout, yup, we should try to fetch that... Sadly, I've realized after some digging that I only have a working tree of the old SVN data, not an actual repo dump. With SVN, that's basically useless. I've contacted folks at Enthought to see if by chance they have an old backup of that SVN raw repo somewhere. If not, we're screwed. I'd be double-pissed (only at myself) if that were the case: I knew we'd lost that history in git, but I thought I had a local copy. I was sure I had a copy of the actual repo, sadly that's not the case... I'm an idiot.
Brian E. Granger
@ellisonbg
Apr 22 2015 03:18
Did we lose the data going from SVN->bzr or bzr->git?
Scott Sanderson
@ssanderson
Apr 22 2015 03:19
it's a rare software project that lives through multiple version control systems
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 13:46
Hello,
Do we have a strong usecase for traitlet.link and traitlet.dlink to have potentially multiple targets
This message was deleted
I wanted to rollback on the registration of on_trait_change in case of a TraitError at creation of the link
In the case of multiple targets, it involves caching old values, and old old values
(This type of rejection may occur with cross validation even if the types are ok)
Jason Grout
@jasongrout
Apr 22 2015 14:58
hmmm...pre-2008? Do you know what the last date was?
@fperez - I wonder if possibly Sage might have an old copy.
Jared Forsyth
@jaredly
Apr 22 2015 15:01
Has anyone tried dockerizing an ipython kernel? (not the main runtime, just a kernel)
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 15:03
@jdfreder regarding requirejs, another big issue is that it requires absolute paths, hence, not URL prefexing possible in nbextensions
Scott Sanderson
@ssanderson
Apr 22 2015 15:17
@jaredly I think @rgbkrk has toyed with what you're talking about here: https://github.com/rgbkrk/kernels-service
Jason Grout
@jasongrout
Apr 22 2015 15:27
requirejs only needs absolute paths in the config statements
for loading, use the "nbextensions/myextensions/myfile" to get the right (url-prefixed) nbextension path
Jared Forsyth
@jaredly
Apr 22 2015 15:30
I just made a nodejs kernel for jupyter! I'd love for people to try it out
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 15:35
@jasongrout yes, but any nbextensions requiring a shim will have this issue
this is what I meant
Jason Grout
@jasongrout
Apr 22 2015 15:39
right
It would be nice if there were some way from js to get the url prefix
you could hack it out by checking the window.url, but it'd be nicer if IPython set some variable somewhere with the prefix
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 15:41
for jquery and all, the shims are in the html templates, which allow setting the prefix
Jonathan Frederic
@jdfreder
Apr 22 2015 15:57
@jaredly awesome!
@jasongrout I thought we set base_url... Maybe it's only accessible in the template, and doesn't trickle down to the Javascript.
@SylvainCorlay I'm pretty sure requirejs does support relative paths
I saw it in action yesterday
when I didn't want to
Scroll down to
Ends in ".js".
Starts with a "/".
Contains an URL protocol, like "http:" or "https:".
Jason Grout
@jasongrout
Apr 22 2015 16:04
by the way, the base url is at IPython.notebook.base_url
Jonathan Frederic
@jdfreder
Apr 22 2015 16:04
Awesome
Min RK
@minrk
Apr 22 2015 16:05
relative urls work in require
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 16:07
in shims?
Min RK
@minrk
Apr 22 2015 16:08
making a new shim in require config?
Jason Grout
@jasongrout
Apr 22 2015 16:08
according to the docs @jdfreder listed above, the requirejs config paths argument does do things relative by default, using the right base_url
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 16:09
requirejs.config({
     paths: {
Jason Grout
@jasongrout
Apr 22 2015 16:09
unless you start the url with /, in which case it's treated like a browser urls
Min RK
@minrk
Apr 22 2015 16:09
right.
Jason Grout
@jasongrout
Apr 22 2015 16:10
so it appears that using a path url that starts with, for example, nbextensions/... should work fine.
so you were right when we were talking before, Sylvain
(and I was wrong)
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 16:11
in any case, requirejsconfig is global, hence every time one reloads a widget model / view, the requirejs.config command has to be re-run to avoid conflicts
so requirejs.config is bad for custom widgets
Min RK
@minrk
Apr 22 2015 16:12
I'm curious. Why is setting global config needed?
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 16:13
initially, we were defining the path for things like "d3" with requirejs.config
Jonathan Frederic
@jdfreder
Apr 22 2015 16:15
@SylvainCorlay that's correct, that's what caused a "timeout" error for me, despite no 404 and a correct path.
oic, you're talking about something slightly different
but still, that solution would fix the problem I saw too
Sylvain Corlay
@SylvainCorlay
Apr 22 2015 16:19
yes, I think we are on the same page. Nice to know about the / prefix for shims
Jason Grout
@jasongrout
Apr 22 2015 20:28
Getting all of these commit notifications from the repo rewriting is like a trip down memory lane :)
Jared Forsyth
@jaredly
Apr 22 2015 20:41
Is there some sort of test suite that can be run against kernels to ensure they comply to the messaging protocol? I'm devving a few kernels, and would like a test suite rather than trial + error :) so I've started into one, but wondered if anyone else had done the same thing.
Jason Grout
@jasongrout
Apr 22 2015 20:50
@jaredly: +1
Thomas Kluyver
@takluyver
Apr 22 2015 21:02
@jaredly yes!
there's a PR open against IPython. We probably want to make it a standalone tool, but we haven't got round to it yet.
ipython/ipython#7646
Jared Forsyth
@jaredly
Apr 22 2015 21:06
super cool