[Networkit] Reading graph segfaults

Staudt, Christian (ITI) christian.staudt at kit.edu
Thu Apr 24 16:43:51 CEST 2014


Yes. It is not enough to rebuild the C++ core only, the Python bindings have to be rebuilt as well. So I always run
	python3 setup.py develop
which builds everything and installs the package into the system in development mode, i.e. just links it rather than copying it.

Best,
Christian


Am 24.04.2014 um 16:34 schrieb Lars Beckmann <lars.beckmann at gmail.com>:

> Hi Christian,
> 
> yes, I did a rebuild after switching to Dev. To rebuild, I ran "sudo scons --optimize=O --target=Core", which worked fine as far as I can see. Only when I try to run the python import statement, I get that error.
> 
> Any other ideas?
> 
> Best,
> 
> Lars
> 
> 
> On Thu, Apr 24, 2014 at 4:26 PM, Staudt, Christian (ITI) <christian.staudt at kit.edu> wrote:
> Hi Lars,
> I cannot reproduce your error, „from NetworKit import *“ works fine for me.
> 
> Are you sure the package is properly built? Did you rebuild when you switched branches? That could be the reason why PageRank is not there.
> 
> Best,
> Christian
> 
> Am 24.04.2014 um 16:18 schrieb Lars Beckmann <lars.beckmann at gmail.com>:
> 
>> Hi Christian,
>> 
>> thanks for the info! I pulled from the repo and updated to the dev branch. I then built the core, which worked without errors. When I then tried to do "from NetworKit import *", I get this error message:
>> 
>> Traceback (most recent call last):
>> File "<stdin>", line 1, in <module>
>> File "/home/adminuser/Utilities/NetworKit/src/python/NetworKit.py", line 24, in <module> import centrality
>> File "/home/adminuser/Utilities/NetworKit/src/python/centrality.py", line 1, in <module> from _NetworKit import Betweenness, PageRank, EigenvectorCentrality, DegreeCentrality, ApproxBetweenness
>> ImportError: cannot import name PageRank
>> 
>> When I switched back to the default branch, rebuilt and ran the statement again, I got no errors. But ApproxBetweenness is only in the Dev branch as far as I could see. I am not a Python expert, so maybe there is an obvious solution to only import ApproxBetweenness. For now I would be happy to test that part (PageRank is interesting too, but maybe later). Or are we doing something completely wrong here?
>> 
>> Best,
>> 
>> Lars
>> 
>> 
>> On Tue, Apr 22, 2014 at 1:24 PM, Staudt, Christian (ITI) <christian.staudt at kit.edu> wrote:
>> Hi Lars,
>> in addition to the paper mentioned by Henning (http://arxiv.org/abs/1403.3005), there a two main sources  for documentation on NetworKit’s classes and functions:
>> 
>> a) the user guide in the form of an IPython Notebook: Doc/Notebooks/NetworKit_UserGuide.ipynb
>> b) the Python docstrings for classes and functions, as well as Doxygen comments in the C++ source code
>> 
>> I admit that both sources are not in perfect shape, so please inform us if and maybe how they could be improved.
>> 
>> Especially for centrality.ApproxBetweenness, which is a very recent addition, we are interested in your practical experiences, e.g. Did it work for your data? How long did it take? Which parameters did you chose for your purposes? etc. 
>> 
>> Kind regards
>> Christian
>> 
>> christian.staudt at kit.edu
>> http://parco.iti.kit.edu/staudt/index-en.shtml
>> Institut of Theoretical Computer Science - Parallel Computing Group 
>> Building 50.34 Room 249
>> Karlsruhe Institute of Technology (KIT)
>> 
>> 
>> 
>> 
>> 
>> Am 22.04.2014 um 12:54 schrieb Lars Beckmann <lars.beckmann at gmail.com>:
>> 
>>> Hi Henning,
>>> 
>>> thanks for the quick reply, that fixed the issue.
>>> 
>>> Christian said that there is an approximation algorithm for betweenness centrality in the dev-branch. Is there an overview for all the available algorithms and their usage? In particular we are interested in centrality measures for very large graphs (millions of nodes, (hundreds of) millions of edges). If you could point me to docs or any other resource that shows the usage of these algorithms, that would be great. Note: I did go trough the doc folder in the repo, but maybe there is more that I haven't found yet?
>>> 
>>> Thanks in advance,
>>> 
>>> Lars
>>> 
>>> 
>>> On Fri, Apr 18, 2014 at 5:49 PM, Henning Meyerhenke <meyerhenke at kit.edu> wrote:
>>> Dear Lars,
>>> 
>>> Thanks for your request. The reason for the error is the input graph's slight, yet important deviation from the input format requirements. The Metis format requires graphs to be 1-indexed, i.e. vertex indices start with 1, not 0.
>>> 
>>> The unfulfilled assertion happens when an edge with index 0 occurs. Then NetworKit's range check fails.
>>> 
>>> Could you please change your data set according to the format and try again?
>>> 
>>> Thanks,
>>> Henning
>>> 
>>> 
>>> Am 18.04.14 15:03, schrieb Lars Beckmann:
>>> 
>>> Hi,
>>> 
>>> we are trying to use NetworKit to compute centrality measures, but
>>> NetworKit segfaults when we try to read the graph. We are using a clean
>>> VM with Ubuntu 13.10
>>> (http://virtualboximages.com/Ubuntu+13.10+amd64+VirtualBox+VDI+Virtual+Appliance)
>>> where we installed the prerequisites mentioned in the NetworKit readme.
>>> We successfully built NetworKit on the VM, but when we try to read our
>>> graph using: G = graphio.METISGraphReader().read("xxx.graph") NetworKit
>>> segfaults. We have built NetworKit without debug flags, but we can do
>>> so, if that is important for the devs to help us out.
>>> 
>>> You can find the graph in question here:
>>> https://drive.google.com/file/d/0B7vRIovbLojkSHpTUy1hR3dFRHM/edit?usp=sharing
>>> 
>>> Thanks for your help,
>>> 
>>> Lars
>>> 
>>> --
>>> Lars Beckmann
>>> 
>>> Uhlenstraße 10
>>> 33098 Paderborn
>>> Germany
>>> 
>>> Cell (D2): +49 (0) 178 59 69 150
>>> Phone: +49 (0) 89 380 123 90
>>> Fax: +49 (0) 180 102 113 4628
>>> 
>>> -- 
>>> 
>>> =======================================================
>>> Karlsruhe Institute of Technology (KIT)
>>> Institute of Theoretical Informatics (ITI)
>>> 
>>> Juniorprof. Dr. Henning Meyerhenke
>>> Theoret. Informatics / Parallel Computing
>>> 
>>> Phone: +49-721-608-41876
>>> Web: http://parco.iti.kit.edu/henningm/
>>> 
>>> KIT - University of the State of Baden-Wuerttemberg and
>>> National Research Center of the Helmholtz Association
>>> =======================================================
>>> 
>>> 
>>> 
>>> 
>>> -- 
>>> Lars Beckmann
>>> 
>>> Uhlenstraße 10
>>> 33098 Paderborn
>>> Germany
>>> 
>>> Cell (D2): +49 (0) 178 59 69 150
>>> Phone: +49 (0) 89 380 123 90
>>> Fax: +49 (0) 180 102 113 4628
>>> <ATT00001.c>
>> 
>> 
>> _______________________________________________
>> NetworKit mailing list
>> NetworKit at ira.uni-karlsruhe.de
>> https://lists.ira.uni-karlsruhe.de/mailman/listinfo/networkit
>> 
>> 
>> 
>> 
>> -- 
>> Lars Beckmann
>> 
>> Uhlenstraße 10
>> 33098 Paderborn
>> Germany
>> 
>> Cell (D2): +49 (0) 178 59 69 150
>> Phone: +49 (0) 89 380 123 90
>> Fax: +49 (0) 180 102 113 4628
> 
> 
> 
> 
> -- 
> Lars Beckmann
> 
> Uhlenstraße 10
> 33098 Paderborn
> Germany
> 
> Cell (D2): +49 (0) 178 59 69 150
> Phone: +49 (0) 89 380 123 90
> Fax: +49 (0) 180 102 113 4628

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20140424/acaf270d/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 495 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20140424/acaf270d/attachment.sig>


More information about the NetworKit mailing list