iPython: Manipulate-like command

问题:

In Wolfram Mathematica, I can interactively modify the value of a parameter by using the Manipulate[] command.

For example, Manipulate[n, {n, 1, 20}] shows a slider through which is possible to vary the value of n.

Is there any simple way (i.e. something like a magic or a decorator, like in SAGE) to achieve the same result in the IPython notebook?

回答1:

Update
This was added in IPython 2.0 (released Apr 1, 2014), it's called Interactive Widgets and works in web notebooks.

Original answer
This is ongoing work for 2.0 (release December something-ish) Have a look at the IPython-dev meeting on YouTube to see progress. The last meeting from oct 21 at 28min-ish has a widget demo by John then interact by Brian.

回答2:

I am not sure whether this would satisfy all your needs as it is still experimental, but seems to do what you asked for - look at static interactive widgets by Jake VanderPlas.

What I did to get running was the following:

Get the source and install:

git clone https://github.com/jakevdp/ipywidgets
pip install ./ipywidgets

start an ipython notebook and experiment with the notebook example.ipynb in the ipywidgets directory

There's another type of interactive visualisation, where panning, zoom and meta-data are available (possibly more). It is by the same author and reported in his blog D3 Plugins: Truly Interactive Matplotlib In Your Browser.

There are also nice docs: MPLD3: Bringing Matplotlib to the Browser

To experiment with it I did more or less the same:

git clone https://github.com/jakevdp/mpld3
pip install ./mpld3/

One can now run the create_example.py script, or even better, start an ipython notebook and play with the provided mpld3/notebooks/mpld3_demo.ipynb as well as with the attractive examples of mpld3_plugins posted in the blog.

It may be worth noting that I am using the Anaconda distro of python, which includes Jinja2; as far as I know Jinja2 is a requirement for mpld3.

I do agree that an @interact decorator a la sage would be nice.