Sebastian Weber | 2 Jan 17:07 2012
Picon

graph-tool on Mac OS Lion - running!

Hi!

Wonders seem to happen once in a while. Finally I got graph-tool running on my Lion machine. Here are my steps
which lead to a running graph-tool:

1. deinstall Macports python 2.7 such that it is never in the way (reinstall whatever you need in macports
with python with the python26 branch)
2. install boost 1.47.0 into $HOME/.local; linked against python 2.7 from apple
3. install numpy/scipy from the SuperPack, see scipy.org
4. install cgal 3.9 with gmp and other dependencies coming from macports; install dir is $HOME/.local
5. grab graph-tool from git and configure it with

PYTHON_VERSION=2.7 PYTHON=/usr/bin/python2.7 CXXFLAGS="-I/$HOME/.local/include
-I/opt/local/include" CFLAGS="-I$HOME/.local/include -I/opt/local/include"
LDFLAGS="-L$HOME/.local/lib -L/opt/local/lib" ./configure --prefix=$HOME/.local
--with-boost=$HOME/.local/
--with-numpy=/Library/Python/2.7/site-packages/numpy-2.0.0.dev_3b3735d_20111219-py2.7-macosx-10.7-x86_64.egg/numpy/
--with-scipy=/Library/Python/2.7/site-packages/scipy-0.11.0.dev_c2ff01c_20111219-py2.7-macosx-10.7-x86_64.egg/scipy/
--disable-debug --disable-openmp

Compile it a very long time and then install it. Of course, I use the latest Xcode 4.2.1. I found it really hard
to get graph tool running on Lion.

Cheers,

Sebastian

--

-- 
Sebastian Weber
Group of Cell Communication and Control
(Continue reading)

Tiago de Paula Peixoto | 2 Jan 17:39 2012
Picon

Re: graph-tool on Mac OS Lion - running!

On 01/02/2012 02:07 PM, Sebastian Weber wrote:
> Compile it a very long time and then install it. Of course, I use the
> latest Xcode 4.2.1. I found it really hard to get graph tool running
> on Lion.

I don't even have access to a Lion machine, hence there is very little I
can do to help. But it is also very obvious that none of the problems
you encountered had anything to do with graph-tool in itself, instead they
seem to me like a macports issue. You should really contact them with a
bug report if you want this to improve in the future.

Cheers,
Tiago

--

-- 
Tiago de Paula Peixoto <tiago <at> skewed.de>

_______________________________________________
graph-tool mailing list
graph-tool <at> skewed.de
http://lists.skewed.de/mailman/listinfo/graph-tool
Sebastian Weber | 2 Jan 17:54 2012
Picon

Re: graph-tool on Mac OS Lion - running!

I am not blaming you, Tiago. Don't get me wrong. Problem is that apple will stay forever with their gcc 4.2 due
to license issues and Macports compilations are not usable.

BTW, I forgot one important step:

Download python 2.7.2 sources and use the h2py script to convert the /usr/include/dlfcn.h into DLFCN.py
and place that somewhere where python 2.7 reaches it.

Speaking of bug reporting: What should I report here? boost python binding don't work? scipy/numpy
doesn't work?

Cheers,

Sebastian

Am 02.01.2012 um 17:39 schrieb Tiago de Paula Peixoto:

> On 01/02/2012 02:07 PM, Sebastian Weber wrote:
>> Compile it a very long time and then install it. Of course, I use the
>> latest Xcode 4.2.1. I found it really hard to get graph tool running
>> on Lion.
> 
> I don't even have access to a Lion machine, hence there is very little I
> can do to help. But it is also very obvious that none of the problems
> you encountered had anything to do with graph-tool in itself, instead they
> seem to me like a macports issue. You should really contact them with a
> bug report if you want this to improve in the future.
> 
> Cheers,
> Tiago
(Continue reading)

Noon Silk | 9 Jan 00:59 2012
Picon

Building on windows (windows 7 64 bit building for win32)

Hello,

 Has anyone had much luck with this?

 I've been trying under cygwin with very little luck. Specifically I'm
running ./configure, and it runs for a reasonable period of time
(after I installed various requirements), but am now getting:

checking for boostlib >= 1.38.0... yes
checking whether the Boost::Python library is available... no
configure: error: No usable boost::python found

 Even though I feel like claiming that there is indeed a viable
boost::python available (that is to say, I claim I've installed it).

 Is there any hope for me here? Is there a better way to get it
running (as a part of a .exe) under windows? I'll continue onward, but
would love some words of encouragement, if anyone has them ...

--

-- 
Noon Silk

Fancy a quantum lunch? http://groups.google.com/group/quantum-lunch?hl=en

"Every morning when I wake up, I experience an exquisite joy — the joy
of being this signature."
Tiago de Paula Peixoto | 9 Jan 20:00 2012
Picon

Re: Building on windows (windows 7 64 bit building for win32)

Hi Noon,

On 01/09/2012 12:59 AM, Noon Silk wrote:
> Hello,
> 
>  Has anyone had much luck with this?

I have never tried this, and I don't have any windows machine nearby to
do so. However, I would be very interested in making any necessary
modifications to make it work, and to provide instruction to those who
wish to try.

>  I've been trying under cygwin with very little luck. Specifically I'm
> running ./configure, and it runs for a reasonable period of time
> (after I installed various requirements), but am now getting:
>
> checking for boostlib >= 1.38.0... yes
> checking whether the Boost::Python library is available... no
> configure: error: No usable boost::python found
>
>  Even though I feel like claiming that there is indeed a viable
> boost::python available (that is to say, I claim I've installed it).

You should look at the full output in config.log to see exactly what the
error was, so that you can understand what happened. Most of the time,
it is only an issue with setting the correct path to the libraries.

>  Is there any hope for me here? Is there a better way to get it
> running (as a part of a .exe) under windows? I'll continue onward, but
> would love some words of encouragement, if anyone has them ...
(Continue reading)

Dmitrijs Ledkovs | 14 Jan 03:38 2012
Picon

Veronoi Diagram and Data locality

Hello,

This is my first post =)

To begin with - what an awesome package! The website is great, the docs are amazing, the code is beautiful, sane autotools usage, very fast, pythonic!
Best software package ever!

I'm a beginner in graph theory. Any suggestions on books/tutorials which can quickly take me from basics to advanced topics would be appreciated.

Now the question.

I'm doing numerical simulation over a veronoi graph. I'm storing data on the vertexes and edges. The iterations go as following:

1) Using data on the edges ('incident pulses') I'm calculating the value of the vertex ('node voltage')
2) Using value of the vertex, I'm calculating new values for the edges ('scattered pulses')
3) During next timestep scattered pulses become incident - repeat

In the future I may need to use symmetric directed graph and do more fun things with incident and scattered pulses.

Now the slight problem arrises, when my graph + values is bigger than I can store in RAM.

There is Global Arrays toolkit, which recently got GAiN (Global Arrays in Numpy) which allow seamlessly use numpy/scipy over shared memory cluster. But it will give good performance if arrays are exercising data locality properties.

In my case I want property maps of vertixes & edges to be 'near each other' based on connectivity. Such that my algorithm computes as many vertexes/edges locally as possible. E.g. for a 2D latice I'm using http://en.wikipedia.org/wiki/Z-order_curve.

Now I want something similar but for veronoi diagram.

I'm not familiar with graph theory, but I was playing around with graph-tool and it seems like betweenness and/or communities give me something what I want. Ideally I want to retrieve indexes which will sort PropertyMap to exhibit data locality properties (Similar to Z-order curve).

Does graph-tool already does something like this?

Regards,

Dmitrijs.


_______________________________________________
graph-tool mailing list
graph-tool <at> skewed.de
http://lists.skewed.de/mailman/listinfo/graph-tool
Tiago de Paula Peixoto | 15 Jan 18:06 2012
Picon

Re: Veronoi Diagram and Data locality

Hi Dmitrijs,

On 01/14/2012 03:38 AM, Dmitrijs Ledkovs wrote:
> Hello,
> 
> This is my first post =)
> 
> To begin with - what an awesome package! The website is great, the docs are amazing, the code is beautiful,
sane autotools usage, very fast, pythonic!
> Best software package ever!

Thanks!

> Now the slight problem arrises, when my graph + values is bigger than
> I can store in RAM.
>
> There is Global Arrays toolkit, which recently got GAiN (Global Arrays
> in Numpy) which allow seamlessly use numpy/scipy over shared memory
> cluster. But it will give good performance if arrays are exercising
> data locality properties.
>
> In my case I want property maps of vertixes & edges to be 'near each
> other' based on connectivity. Such that my algorithm computes as many
> vertexes/edges locally as possible. E.g. for a 2D latice I'm using
> http://en.wikipedia.org/wiki/Z-order_curve.
>
> Now I want something similar but for veronoi diagram.
>
> I'm not familiar with graph theory, but I was playing around with
> graph-tool and it seems like betweenness and/or communities give me
> something what I want. Ideally I want to retrieve indexes which will
> sort PropertyMap to exhibit data locality properties (Similar to
> Z-order curve).
>
> Does graph-tool already does something like this?

graph-tool does not yet include vertex-ordering routines. In the case of
Delaunay triangulations, the order of the vertices will correspond
exactly to the order of points supplied. Thus, you may provide an
ordered list of points in order to obtain an ordered graph. Locality, in
this case, can be reasonably identified by (euclidean) distance.

Vertex ordering is relatively easy to implement (for a given predefined
order), so you can expect this to be in the next versions.

Note that if you want to _find_ the best partition for any given graph,
this is called the Graph Partition Problem, and is in general NP-hard. A
good starting-point heuristic is the Kernighan-Lin algorithm:

      http://en.wikipedia.org/wiki/Kernighan%E2%80%93Lin_algorithm

This is not yet implemented in graph-tool.

Notice however that, unfortunately, there is no way to use GAiN arrays
with graph-tool's algorithms, since they all expect simple (local)
arrays. So I'm not sure you will be able to accomplish much after
ordering your graphs...

Cheers,
Tiago

--
Tiago de Paula Peixoto <tiago <at> skewed.de>

_______________________________________________
graph-tool mailing list
graph-tool <at> skewed.de
http://lists.skewed.de/mailman/listinfo/graph-tool
Hector | 18 Jan 20:32 2012
Picon

Installing in a virtual environment

Hi all,
I've tried to install graph-tool inside a virtual environment in
multiple ways now, e.g.

1. pip install graph-tool
2. pip install http://downloads.skewed.de/graph-tool/graph-tool-2.2.15.tar.bz2
3. easy_install graph-tool
4. Running "configure-make-make install" from source

The first three failed basically because there's no setup.py and the
fourth one failed because it couldn't find numpy (which is already
installed in the virtual environment). I could do a system-wide
installation, but that's not what I want. Have you guys successfully
installed graph-tool inside a virtual environment?
Best,
--

-- 
 Hector
_______________________________________________
graph-tool mailing list
graph-tool <at> skewed.de
http://lists.skewed.de/mailman/listinfo/graph-tool
Tiago de Paula Peixoto | 20 Jan 09:21 2012
Picon

Re: Installing in a virtual environment

Hi Hector,

Sorry for the delay.

On 01/18/2012 08:32 PM, Hector wrote:
> Hi all,
> I've tried to install graph-tool inside a virtual environment in
> multiple ways now, e.g.
> 
> 1. pip install graph-tool
> 2. pip install http://downloads.skewed.de/graph-tool/graph-tool-2.2.15.tar.bz2
> 3. easy_install graph-tool
> 4. Running "configure-make-make install" from source
> 
> The first three failed basically because there's no setup.py and the
> fourth one failed because it couldn't find numpy (which is already
> installed in the virtual environment). I could do a system-wide
> installation, but that's not what I want. Have you guys successfully
> installed graph-tool inside a virtual environment?

The first three approaches failed because graph-tool uses autotools
instead of python's distutils. The fourth approach should work
automatically, if all the libraries are found in the standard places. If
the script did not find numpy, this means you probably need to specify
exactly where it is. You should provide the exact output of the
configure script, together with the file config.log, so we can see
exactly what is happening.

Cheers,
Tiago

--
Tiago de Paula Peixoto <tiago <at> skewed.de>

_______________________________________________
graph-tool mailing list
graph-tool <at> skewed.de
http://lists.skewed.de/mailman/listinfo/graph-tool
Tue V Jensen | 27 Jan 14:25 2012
Picon

Apt installation fails on Debian Sid

Hi!

I tried installing the graph_tool package as per the instructions on the
download page, but apt-get fails with the following error:

Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
 graph-tool : Depends: libcgal7 but it is not installable

It seems libcgal7 was recently renamed to libcgal8, as this is the only
available package in both testing and Wheezy.
Can I install using libcgal8, or will  have to wait for a version compatible
with this?

Thanks!
Tue

--
View this message in context: http://main-discussion-list-for-the-graph-tool-project.982480.n3.nabble.com/Apt-installation-fails-on-Debian-Sid-tp3693256p3693256.html
Sent from the Main discussion list for the graph-tool project mailing list archive at Nabble.com.

Gmane