Installing graph-tool to use with Anaconda

Hello,

I have successfully installed graph-tool on Ubuntu 1310 using apt-get. So it works with the default python of the system.
However, I would like to use it with the python included in the Anaconda distribution.

Can anybody please help how to do that? Does it need to be compiled from source in that case?

Thanks very much,
Eszter

attachment.html (2.13 KB)

Hello,

You might consider setting up virtual environment where your anaconda
version resides, there is an option to use system packages within this
environment. On a similar setup, I had to uninstall the system's
matplotlib to avoid libraries beeing mixed up while using graph-tool.

Guillaume

attachment.html (3.13 KB)

If the anaconda python version is not the same as the system default,
you will need to compile graph-tool from source. In that case just
pointing the configure script to the desired python interpreter should
work. Note also that you probably may also need to compile Boost python
against the same python version.

Cheers,
Tiago

Thank you! I am trying to compile graph-tool from source, but I don't have much experience in building software from source.
The machine is the mentioned Ubuntu 13.10 running in Virtualbox. The Anaconda python version is 2.7.6 (I specified this in the PYTHON environment variable and the configure script seems to understand it).
It seems all dependencies are satisfied, and the virtual machine is allocated quite a lot of memory (8GB).
I get this error:

make[4]: Entering directory `/.../graph-tool-2.2.27/src/graph/centrality'
  CXX graph_betweenness.lo
  CXX graph_closeness.lo
  CXX graph_eigentrust.lo
  CXX graph_eigenvector.lo
  CXX graph_hits.lo
  CXX graph_katz.lo
  CXX graph_pagerank.lo
virtual memory exhausted: Cannot allocate memory
make[4]: *** [graph_pagerank.lo] Error 1

Any ideas how to solve this?

Thank you,
Eszter

8GB ram should be more than enough to compile it. What GCC version are
you using? Are you compiling it in parallel (with the -j option)?

Cheers,
Tiago

I got some help from a sysadmin colleague -- turned out the problem was that the virtual machine was 32 bit, so it could not use all the allocated memory.
I am sure it will work on a 64 bit one.

Thank you for the quick responses,
Eszter