Which is the best way to install nipype for Mindboggle/Python 3?

[binarybottle]

Should I run?:

- conda install lxml
- pip install prov traits future simplejson # nose
- git clone https://github.com/nipy/nipype.git $HOME/nipype
- cd $HOME/nipype
- python setup.py install

Or?:

• conda install lxml
• pip install prov traits future simplejson # nose
• pip install nipype
• pip install --upgrade https://github.com/nipy/nipype/archive/master.zip

[nicholsn]

@satra will have the real answer, but this is what I use for a minimal install w/conda:

conda install scipy networkx traits dateutil six nose
pip install nibabel
pip install nipype

[satra]

the simplest thing that should work is:

conda install pip lxml
pip install https://github.com/nipy/nipype/archive/master.zip

this is what i often do to drill into my head all the dependencies, and to leverage conda's scipy and lxml - we hope to simplify a lot of this after the upcoming release. the situation has become a little unfortunate with both lxml (in general) and traits (on python 3.x)

conda install --yes pip scipy nose networkx lxml future simplejson
pip install nibabel prov xvfbwrapper traits
pip install https://github.com/nipy/nipype/archive/master.zip

[binarybottle]

Thank you! Only thing that's missing is I am unable to generate graphs with this setup. Don't I need to install graphviz as well?:

from http://miykael.github.io/nipype-beginner-s-guide/installation.html:

#Install graphviz and pygraphviz separately
sudo apt-get install graphviz libgraphviz-dev
pip install --upgrade pygraphviz graphviz

[satra]

@binarybottle - yes, you will have to do that separately, since graph generation is optional.

[binarybottle]

Thanks -- I'll add it, because I love graphs!