Jupyter_contrib_nbextensions: A variable explorer for notebook? (like MATLAB's Workspace)

[cqcn1991]

I want to know if there is anything like MATLAB's Workspace? (or Spyder's Variable Explorer). So we can easily copy the values out. Something look like this:

---

I ask it here (https://github.com/jupyter/notebook/issues/1516), Carreau gives a very good example. In short, you can get the variables by this:

variables = %who_ls
for v in variables:
    print v, type(eval(v)).__name__, str(eval(v))

Another thing worth considering is spyderlib.plugins.variableexplorer from Spyder. This could be seen at http://stackoverflow.com/questions/2960864/how-can-i-save-all-the-variables-in-the-current-python-session

[jankatins]

rodeo has such a thing:
https://github.com/yhat/rodeo/blob/921ad700c432f5510f1b368c16d1acc7b361de2a/src/browser/jsx/containers/variable-viewer.jsx
https://github.com/yhat/rodeo/blob/0b4a9ed5acf052fa9e2eca9427263b5dc44c187d/src/node/kernels/python/langs/python-patch.py

I would also love to see such a thing! On the other hand: isn't that what the jupyterhub jupyterlab stuff is also about?

[cqcn1991]

@janschulz Never thought of Rodeo, I thought it was about desktop version.
Never heard of jupyterlab, is it this? https://github.com/jupyter/jupyterlab

[jankatins]

rodeo isn't a notebook app, but also uses the kernel infrastructure and has a browser UI (Rodeo (and RStudio on which rodeo is based) are basically a browser window).

RE jupyterlab: I've heard some stuff about it (there was a survey which (also) said "we need more IDE feature in the Notebook UI" and I think jupyterlab is building infrastructure for suchthings) but I've actually no clue what the overall vision is for that thing...

[jankatins]

The original answer in the notebook repo linked to https://github.com/ipython/ipywidgets/blob/master/docs/source/examples/Variable%20Inspector.ipynb -> doesn't look so complicated :-)

[cqcn1991]

@janschulz Yes, that widget is pretty amazing and simple. The problem is that how to make it "copy and pastable", i.e. have table display UI for array and matrix.

Addtionally, I personally want a Ruby on Rails-like framework for Jupyter Notebook, in which have pre-load common use nbextensions, libraries and good file organizations. It can help people get to know useful features, extentions at the beginning, saving people tons of time for configuration.

[flying-sheep]

to be cross-language-compatible we need a protocol extension AFAICS:

1. a request to the kernel to return the names and types of variables.
2. a request to the kernel to open an interactive editor for a variable or at least display it (i.e. the kernel can reply with a open-pager message or create a widget able to edit the variable)

[flying-sheep]

@cqcn1991 to get what you want right now, you can do something like

from IPython.core.page import page
from IPython.core.formatters import format_display_data

page(format_display_data(thing, ['text/plain', 'text/html']))

(the second argument to format_display_data is only for performance since the pager can’t handle more mimetypes anyway)

[cqcn1991]

@flying-sheep what is thing in page(format_display_data(thing, ['text/plain', 'text/html']))?

[flying-sheep]

the variable you want to display. to get it there’s a bit more work involved:

1. you have to set up JS event listeners for the cells that send a comm request to the kernel
2. you have to set up event listeners in python that react to that comm request and do thing = ipython.kernel.shell.user_ns[varname]

[PatrickChrist]

Would be a great feature!

[jsekamane]

How about the Variable Inspector extension?

[ashimb9]

Does the variable inspector work in Jupyterlab?

[jcb91]

Does the variable inspector work in Jupyterlab?

No, jupyterlab uses a different (and so far still constantly-changing) javascript model, so I know of no extensions that work there yet

[jfbercher]

A varInspector is planned in the future in JupyterLab, see the roadmap.

Also, it is planned in the notebook development roadmap for Jupyter v 5+ to port/bridge popular nbextension "we will continue to work on porting nbextensions to JupyterLab and building the bridge layers." and "At least some important, popular extensions are ported / bridged to Jupyter Lab". Of course we will be happy to port or help, but for now the interface seems not quite stable and/or I don't find easy keys.

[biocyberman]

How about the Variable Inspector extension?

This works for simple variables, but a better presentation (i.e. like that of in Spyder) of dataframes/matrices/vectors is much appreciated.

Alternatively, I also tried running jupyter notebooks inside spyder, this did not work well.

[max8599]

@ biocyberman, i think i'm exactly looking for the same like you. Did you find a better variable explorer than the variable inspector in Jupyter?