Re: [boost] [Boost Graph Library] Having trouble constructing necessary parameters for the isomorphism algorithm

Subject: Re: [boost] [Boost Graph Library] Having trouble constructing necessary parameters for the isomorphism algorithm
From: Jeremiah Willcock (jewillco_at_[hidden])
Date: 2011-01-17 14:12:49

On Sun, 16 Jan 2011, Alexander Bock wrote:

>> Date: Mon, 10 Jan 2011 10:44:21 -0500
>> From: jewillco_at_[hidden]
>> To: boost_at_[hidden]
>> Subject: Re: [boost] [Boost Graph Library] Having trouble constructing necessary parameters for the isomorphism algorithm
>>
>> On Sun, 9 Jan 2011, Alexander Bock wrote:
>>
>>>
>>> Ok. I don't see the appropriate function in the header. I see one that looks like one on line 387 in isomorphism.hpp though.
>>>
>>> Through trial and error, I managed to get the code to compile by changing the container for vertices from listS to vecS. I ran some tests and sometimes it works, but on strange occasions the program crashes with what looks like an index out of range message...I tracked down the crash to line 205 of isomorphism.hpp in the function bool match(edge_iter iter, int dfs_num_k). I have no idea if this is an actual bug (I doubt it though) or what I'm doing wrong.
>>>
>>> Here's the full error message:
>>>
>>> Expression: ("_Myptr + _Off <= ((_Myvec *)(this->_Getmycont()))->_Mylast && _Myptr + _Off >= ((_Myvec *)(this->_Getmycont()))->_Myfirst", 0)
>>>
>>> The error occurs on line 163 in the STL vector header.
>>
>> Do you have a simple, complete example that I could try? Does your STL
>> implementation have a debugging mode that turns on bounds checking, or do
>> you have a tool similar to Valgrind?

Are you initializing the index bundled property in your graphs?  You are 
passing it in as the vertex_invariant* arguments to isomorphism(), and 
that algorithm requires that the invariant maps both be initialized. 
Valgrind reports a use of an uninitialized value, which is likely from one 
of the invariant maps.  Also, what are you passing in as max_invariant? 
It appears that you are leaving it as the default, which appears to only 
work if your invariant is the vertex degree.

-- Jeremiah Willcock