GitLab

This makes the code more efficient if the input parameter combinations
are much larger than the graph itself.

This fixes a bug in the "correlated" mode in random_rewire().

This implements a general "blockmodel" generation / rewiring algorithm,
using Gibbs acceptence / rejection sampling (a.k.a Metropolis-Hastings).

This also implements some optimizations in the rewiring code, which
makes it more efficient on filtered graphs.

This introduces several simplifications and corrections to the graph
rewire algorithm, to guarantee unbiased sampling.

Now a move is outright rejected if it produces a
self-loop/parallel-edge, instead of retried. This also adds a
"non-sweep" mode, where edges are rewired randomly, possibly with
repetition.

The edge moves are now simplified to the target of the edges only,
since swaping sources is redundant.

The number of iterations can now be explicitly modified, so it is not
necessary to call the function more than once, and it is emphasized in
the documentation that only after sufficiently many iterations can the
graph be guaranteed to be fully mixed.

This enables support for older versions of GCC (>=4.2), by falling back
to the boost tr1 implementation when the GCC version is less than 4.4.

This includes also some internal refactoring, and bug fixes.

This extensively modifies the random_rewire() algorithm, so that only
either the source or the edge of each edge is rewired (not both, as
previously), and no parallel edges are created during the algorithm (if
desired).

The new version is much faster, and never gets stuck. However, more than
one run may be necessary in order to obtain a uniform shuffling.

This fixes a implicit directionality bug in random_rewire() with
undirected graphs.

This fixes a problem, where the underlying edge directionality would
cause correlations to arise when rewiring undirected graphs. Now each
edge is correctly considered in both possible orientations, which are
chosen randomly.

This allows compilation with older versions of boost (>=1.38), which in
some sistems are the only option. This commit also removes the linking
of the boost_graph shared library, since none of its symbols are
actually required.

This implements a simpler rewire strategy where the edges are randomly
placed, without preserving the degrees of the vertices.

Fix compilation warnings with boost 1.40, remove unnecessary files, fix
autoconf macros and link with "boost_graph".

This also raises the minimum boost version to 1.40.

No longer only thrown GraphError upon any error, but instead throw
specific exceptions which are more meaninful and are mapped to standard
python exceptions, such as IOError, ValueError and RuntimeError.

When using undirected graphs, the implicit direction of each edge was
not properly handled, and the correlated case would not work. Now, in
this case, the direction is always checked, and the edge is inverted if
necessary.

The generators from boost::random seem to have a bug which causes them
to be biased. The generators from tr1::random seem to be in better
shape.

See: http://thread.gmane.org/gmane.comp.lib.boost.user/48006

This finally fixes in the bug addressed by commit 309ddbbd, where
parallel edges could be erroneously created. In fact, the bug was more
serious: The source and target edge lists always pointed to the same
list (in the uncorrelated case, but could occasionally happen for the
correlated case also) which got shuffled during iteration. Since the
shuffling of one list interfered with the shuffling of the other, some
combinations of source and target edges could simply never be
considered... This commit forces both lists to always be independent.

Thins changes the graph filtering code slightly to wrap graph types with
GraphWrap, which automatically updates the edge index list when edges
are removed and added to the graph.

This also changes how graphs are passed to algorithms, which is now by
reference instead of pointer. (hence this touches lots of code, but
changes are trivial)

In some circumstances, the test for parallel edges would fail to detect
one, if it involved another "new" edge and one of the current ones being
rewired.

This commit also removes the requirement that edge indexes be continuous
in the range [0, num_edges(g)), which is not in general the case.

This includes a new vector property map type (fast_vector_property_map)
which has optional disabling of bounds checking, through its associate
map type (unchecked_fast_vector_property_map). This should improve
performance on algorithms which depend on tight loops which access
property maps.

Bounds checking is only disabled locally just before the algorithms run,
and proper care is taken for bounds checking _beforehand_. The property
maps exposed to python still have internal bounds checking.

Unnecessary dynamic polymorphism is replaced by static polymorphism, in
the definitions of the rewiring strategies.

This also creates the misc sub-module. This may be re-organized in the
future.

Restructure the rewiring code, introducing further abstraction through
class inheritance.

Both uncorrelated and correlated cases draw edges directly.
This has actually proven faster than drawing vertices for the correlated
case, since realizing that indexes could be stored instead of edges.
Doing so avoids changes in the pool of candidate edges, which in turn
removes the need to rebuild it for each edge to rewire.
Consequently, it also makes the uncorrelated case a lot quicker.

In the uncorrelated undirected case, the new code also fixes a serious
bug: when building the edge pool, only one end of each edge was looked
at, because the "edges" vector is not equivalent to drawing all
out_edges from all vertices, as is done now.

This commit splits libraph_tool into different libraries:
 
   - libgraph_tool_core
   - libgraph_tool_clustering (*)
   - libgraph_tool_community (*)
   - libgraph_tool_correlations (*)
   - libgraph_tool_distance (*)
   - libgraph_tool_generation (*)
   - libgraph_tool_layout (*)
   - libgraph_tool_misc (*)
   - libgraph_tool_stats (*)

It also adds the python sub-module 'test', which provides extensive unit
testing of the core functionality. The core library is fully functional
and all test pass successfully.

(*) -> module needs to be ported to new refactoring, and does not yet build

This is a huge commit which completely refactors the metaprogramming
engine which generates and selects (at run time) the graph view type and
the desired algorithm implementation (template instantiation) that runs
on it.

Things are laid out now as following. There exists a main underlying
graph type (GraphInterface::multigraph_t) and several other template
classes that mask it some way or another, in a hierarchic fashion:

     multigraph_t -> filtered_graph (edges only, vertices only, both)
         |                               |           |           |
         |                               |           |           |
         |-------(reversed_graph)--------|-----------|-----------|
         |                               |           |           |
         \------(UndirectedAdaptor)------------------------------/

The filtered_graph filters out edges and/or vertices from the graph
based on some scalar boolean property. The reversed_graph reversed the
direction of the edges and, finally, the UndirectedAdaptor treats the
original directed graphs as undirected, transversing the in- and
out-edges of each vertex indifferently. Thus, the total number of graph
view types is 12. (The option --disable-graph-filtering can be passed to
the configure script, which will disable graph filtering altogether and
bring the total number down to 3, to reduce compile time and memory
usage)

In general, some specific algorithm, implemented as a template function
object, needs to be instantiated for each of those types. Furthermore,
the algorithm may also depend on other types, such as specific
property_maps. Thus, the following scheme is used:

    struct my_algorithm // algorithm to be implemented
    {
        template <class Graph, class PropertyMap>
        void operator()(Graph *g, PropertyMap p, double& result) const
        {
            // ...
        }
    };

    // in order for the above code to be instantiated at compile time
    // and selected at run time, the run_action template function object
    // is used from a member function of the GraphInterface class:

    double GraphInterface::MyAlgorithm(string prop_name)
    {
        double result;
        boost::any vprop = prop(property, _vertex_index, _properties);
        run_action<>()(*this, bind<void>(my_algorithm(), _1, _2,
                                         var(result)),
                       vertex_scalar_properties())(vprop);
        return result;
    }

The whole code was changed to reflect this scheme, but now things are
more centralized and less ad-hoc code needed to be
written. Unfortunately, due to GCC's high memory usage during template
instantiations, some of the code (namely all the degree correlation
things) had to be split in multiple compilation units... Maybe this will
change in the future if GCC gets optimized.

This commit also touches other parts of code. More specifically, the way
filtering gets done is very different. Now we only filter on boolean
properties, and with the above scheme, the desired implementation runs
with the correct chosen type, and no implicit type conversions should
ever happen, which would have a bad impact on performance.

Move remaining parallelism check out of the outer/inner loops and into the parallel_check() function.

3) missing parallel edge check in uncorrelated strategy
2) "_in()" for target and source of "target swap edge"
1) order of arguments of swap_edge_triad::parallel_check

melodrama time...
"premature optimization is the root of all evil"
"careless abstraction are the fruits of its tree"

The check for parallel edges was cleaned up and abstracted into a
function, and other minor things were changed. Code comments were added
to places where it might have been hard to understand.

Major changes to both correlated and uncorrelated rewiring code.
Both cases are now functional and use iterative shuffling.
Uncorrelated picks edges to rewire directly, correlated picks
vertices in order to choose edges.