[IPython-dev] Efficient way to render Javascripts

Brian Granger ellisonbg at gmail.com
Mon Jun 17 00:51:47 EDT 2013 

Jason,

> That is exactly what I wanted to hear. I'm just looking for some clarity on
> how IPython works, how our project can utilize it, and where our resources
> can potentially help you all. I agree firmly with sound design before
> writing code and am glad to here you all are committed to that. Thanks!

Thanks for understanding, glad my comments helped you.

> So the better question for you all may be this:
>
> We're developing some browser based visualizations that we will use a simple
> web server, a socket, or writing js on the fly to communicate between our
> Python packages and the JS libraries. We'd love to design this in a way that
> we can display the same thing in a notebook cell. Right now we can do that
> with the ipython Display mechanisms but the data flow is definitely a hack.
> What would be your recommendation for our approach so that (1) we can
> display our visualization in the notebook now but without investing tons of
> time into code that will be useless once you all have the widget support and
> (2) that our designs be intended to mesh with the future JS widget designs
> for the notebook (i.e. that our design is agnostic to the display mechanism
> in the browser)?

Here is a sketch of the some of the major features of the new design:

* We will create an API that allows Python and JS to communicate data
back and forth using JSON.  The JSON data might be the data you want
to visualize.
* When data is sent back and forth, a handler will be declared for
that data.  There will likely be handlers on both the JS and Python
sides.  These handlers will consume the JSON data and do whatever
needs to be done.
* We want to move away from having to write "code in strings" that is
passed to Kernel.execute or the Javascript object.  The handler names
will be used instead.
* We expect the current Javascript display object to continue to work though.

So, if you can work to cleanly separate your JS code from the JSON
data that goes back and forth, you should have a clean path for the
transition.

> And a secondary question, what's the best way for us to follow progress in
> the ipython js developments. Is the roadmap wiki the best place for the
> looking for sumarized info on these parts of the project?

The roadmap is pretty good, but we will probably also have other
design documents on the wiki just for the interactive widgets.  I
would check back in early August and we will have a lot more to say.
We will also send out requests for comments on ipython-dev during the
design and implementation process.

Looking forward to see what you are all up to...

Cheers,

Brian

>
> Jason
> moorepants.info
> +01 530-601-9791
>
>
> On Sat, Jun 15, 2013 at 9:44 AM, Brian Granger <ellisonbg at gmail.com> wrote:
>>
>> I don't think you are going to like this, but I want to be honest
>> about our situation wrt these things:
>>
>> * It is not supported to do direct calls from the browser to the
>> kernel.  You can hack it, but in general there is no promise the
>> kernel is even running on publicly visible host.  Personally I would
>> not invest the time to do direct GET/POST to the kernel.
>> * If you write the data to a file, you can load the data using URLs of
>> the form "files/mydata.json".
>> * There is no way for JS to directly write data to a file.  You will
>> have to embed the data in the string passed to Kernel.execute.
>> * If you really have to work on all of this during the summer, it is
>> going to be somewhat painful and you will have to redo a good part of
>> it once IPython's support improves later this year.  If it were me, I
>> would focus on other things.  This type of thing should be *trivial*
>> once we have done our work.
>> * We will *start* on this in late summer, but the work will be done
>> gradually between late summer and early next year.  But, the low level
>> pieces that you probably need should be done by Fall.
>> * We are committed to getting the design of these things right.
>> Within the IPython code base we are strictly focused on our 1.0
>> release and roadmap:
>>
>> https://github.com/ipython/ipython/wiki/Roadmap:-IPython
>>
>> * This means we are actually committed to *not* doing anything on the
>> interactive widgets within IPython until late Fall.  Even if other
>> people want to contribute to the code base for these things, it will
>> be ignored.  It is not a matter of writing code - that is the easy
>> part - the hard part is the design.  IPython is a big project and it
>> is extremely important for us to be focused, otherwise the project
>> starts to look like a random walk and progress on big things is
>> difficult.  Until later summer our focus is on nbconvert,
>> multidirectory support and the 1.0 release.
>> * IPython has been in this situation multiple times before - IOW
>> people waiting long periods of time for us to finish important work.
>> In all cases, waiting and being patient has been the right thing to
>> do.
>>
>> I know this is probably bad news, but in the end, if you simply wait
>> you will be able to accomplish what you want with a smile on your face
>> saying "wow, I can't believe how awesome and easy this is...just
>> magic"
>>
>> Cheers,
>>
>> Brian
>>
>> On Sat, Jun 15, 2013 at 9:23 AM, Jason Moore <moorepants at gmail.com> wrote:
>> > Brian,
>> >
>> > So it seems that for us to get some data from a python object into our
>> > javascript, we write the data to file on the fly so that a js cell can
>> > pull
>> > it in from the filesystem or we hard code the data into the js string.
>> > Is
>> > that the way it should be done with the current setup?
>> >
>> > So we'd like the functionality for GET/POST in the notebook for our
>> > project.
>> > And I have one full time student that starts this week. What would be
>> > the
>> > best way for us to contribute some patches to jumpstart this on the
>> > ipython
>> > side?
>> >
>> >
>> >
>> > Jason
>> > moorepants.info
>> > +01 530-601-9791
>> >
>> >
>> > On Sat, Jun 15, 2013 at 9:02 AM, Brian Granger <ellisonbg at gmail.com>
>> > wrote:
>> >>
>> >> Right now it does not have any easy way of doing that.  For now you
>> >> will have to encode your data into the call to kernel.execute, which
>> >> is a completely pain. That is exactly what we will be tackling later
>> >> in the summer...
>> >>
>> >> Cheers,
>> >>
>> >> Brian
>> >>
>> >> On Sat, Jun 15, 2013 at 8:54 AM, Jason Moore <moorepants at gmail.com>
>> >> wrote:
>> >> > Brian,
>> >> >
>> >> > In IPython, is there a standard mechanism to pass data from the
>> >> > client
>> >> > to
>> >> > the server? We'd like to make some python calls that generate a data
>> >> > stucture, jsonify it and then make it available to the javascript
>> >> > that
>> >> > is
>> >> > running in a cell via a GET request. I assume the kernel architecture
>> >> > has
>> >> > this in place?
>> >> >
>> >> >
>> >> > Jason
>> >> > moorepants.info
>> >> > +01 530-601-9791
>> >> >
>> >> >
>> >> > On Sat, Jun 15, 2013 at 8:14 AM, Brian Granger <ellisonbg at gmail.com>
>> >> > wrote:
>> >> >>
>> >> >> Right now (today) the best way is to use the Javascript object along
>> >> >> with the display function.  However, starting in late summer we will
>> >> >> be improving this API so it may change some, but all for the
>> >> >> better...
>> >> >>
>> >> >> Cheers,
>> >> >>
>> >> >> Brian
>> >> >>
>> >> >> On Fri, Jun 14, 2013 at 11:33 PM, TARUN GABA <tarun.gaba7 at gmail.com>
>> >> >> wrote:
>> >> >> > Hi,
>> >> >> >
>> >> >> > I am wondering on what method would be more efficient to serve
>> >> >> > Javascripts on IPython notebooks
>> >> >> >
>> >> >> > I need to visualize/simulate multibody dynamics, and there are two
>> >> >> > methods which are in my mind.
>> >> >> > The rough API of the project is being formed at ..
>> >> >> >
>> >> >> > http://pydy.org/visualization
>> >> >> >
>> >> >> > the two methods are:
>> >> >> > 1) Via http requests:
>> >> >> >
>> >> >> >   on calling something like ..
>> >> >> >
>> >> >> > myobject.visualize(numerical_data)
>> >> >> >
>> >> >> > The browser sends a POST request to the kernel, and python methods
>> >> >> > send a POST request to the required javascripts alongwith POST
>> >> >> > parameters as the numerical data required for visualization, most
>> >> >> > likely a 4*4 transformation matrix, and shape information.
>> >> >> > the Javascripts recieves the post and then it is displayed in the
>> >> >> > IPython output cell.
>> >> >> >
>> >> >> > 2) Another method can be to make Javascript objects and save them
>> >> >> > as
>> >> >> > text in the classes,
>> >> >> > something like ..
>> >> >> >
>> >> >> > def MyClass():
>> >> >> >
>> >> >> >     def __init__(relevant,args):
>> >> >> >         self.js_data = ''' A call to the Javascript functions,
>> >> >> > alongwith arguments..
>> >> >> >                              var myobject =
>> >> >> > Shape_creating_function(%s)'''%(args)
>> >> >> >
>> >> >> > and then on calling ..
>> >> >> > MyClass.visualize()
>> >> >> >
>> >> >> > from IPython notebook, first calling all the required
>> >> >> > javascripts(files) via publish_javascript()/publish_html() display
>> >> >> > methods and then
>> >> >> >
>> >> >> > display(Javascript(MyClass.js_data))
>> >> >> >
>> >> >> > Which of these would be more feasible and efficient..
>> >> >> > _______________________________________________
>> >> >> > IPython-dev mailing list
>> >> >> > IPython-dev at scipy.org
>> >> >> > http://mail.scipy.org/mailman/listinfo/ipython-dev
>> >> >>
>> >> >>
>> >> >>
>> >> >> --
>> >> >> Brian E. Granger
>> >> >> Cal Poly State University, San Luis Obispo
>> >> >> bgranger at calpoly.edu and ellisonbg at gmail.com
>> >> >> _______________________________________________
>> >> >> IPython-dev mailing list
>> >> >> IPython-dev at scipy.org
>> >> >> http://mail.scipy.org/mailman/listinfo/ipython-dev
>> >> >
>> >> >
>> >> >
>> >> > _______________________________________________
>> >> > IPython-dev mailing list
>> >> > IPython-dev at scipy.org
>> >> > http://mail.scipy.org/mailman/listinfo/ipython-dev
>> >> >
>> >>
>> >>
>> >>
>> >> --
>> >> Brian E. Granger
>> >> Cal Poly State University, San Luis Obispo
>> >> bgranger at calpoly.edu and ellisonbg at gmail.com
>> >> _______________________________________________
>> >> IPython-dev mailing list
>> >> IPython-dev at scipy.org
>> >> http://mail.scipy.org/mailman/listinfo/ipython-dev
>> >
>> >
>> >
>> > _______________________________________________
>> > IPython-dev mailing list
>> > IPython-dev at scipy.org
>> > http://mail.scipy.org/mailman/listinfo/ipython-dev
>> >
>>
>>
>>
>> --
>> Brian E. Granger
>> Cal Poly State University, San Luis Obispo
>> bgranger at calpoly.edu and ellisonbg at gmail.com
>> _______________________________________________
>> IPython-dev mailing list
>> IPython-dev at scipy.org
>> http://mail.scipy.org/mailman/listinfo/ipython-dev
>
>
>
> _______________________________________________
> IPython-dev mailing list
> IPython-dev at scipy.org
> http://mail.scipy.org/mailman/listinfo/ipython-dev
>



--
Brian E. Granger
Cal Poly State University, San Luis Obispo
bgranger at calpoly.edu and ellisonbg at gmail.com

More information about the IPython-dev mailing list