Discussions

Infos

• Node count: 667
• Edge count: 4888

A node represents a website. The size of a node depends on his inbound links. A hyperlink between two websites create a edge. Edge are not weighted; if there is more than one hyperlink between two websites, only the first one is counted.

Legend

• black: companies
• blue: CPAN’s authors pages
• red: opensource
• green: bloggers
• purple: offical community

Downloads

These documents are all licenced under the creative commons licence. You can open .gexf graph files with the opensource editor Gephi.

• The distributions map in PDF format (July 2009)
• The distributions map in SVG format (July 2009)

A new version of the authors map has been created for the YAPC::EU. This time, instead of CPANTS’ data, we used CPANDB and Task::Kensho to create a seed of modules.

Task::Kensho is a project from the EPO, which lists some Perl modules recommended for modern Perl development. CPANDB is a project from Adam Kennedy, and it’s an unified database for the CPAN index.

Discussions

Infos

• Node count: 840
• Edge count: 2248

The authors from this map are the one extracted from the new distributions map. There is a lot of informations on this map, so I will explain how to read it.

• The size of a node represents the number of modules an author has released on the CPAN
• An edge is created when an author uses a module from another author
• The size of the label represents the number of inbound links. The more uses of this author’s modules by other authors, the bigger the label
• The color represent the average kwalitee of an author
• The thickness of an edge depends on the number of modules by an author used by another author

So if we look at the map, we can see that:

• Adam Kennedy releases a lot of modules on the CPAN, but few of them are used by other CPAN developpers on the CPAN, so he gets a big green node, with a small label
• Giles Aas has few modules released on the CPAN, but they are extremely used (like LWP and URI), so the size of the node is small, but the label is much bigger
• The size of the edge between Rocco Caputo and Chris Williams is important, because Rocco releases the POE distribution, and Chris Williams releases a lot of POE extensions

And to conclude, we still have our Modern Perl authors cluster, on the top of the map, with developpers like Stevan Little, Yuval Kogman, Marcus Ramberg, Florian Ragwitz, …

As for the previous maps, Gephi was used to do the spatialization.

Downloads

These documents are all licenced under the creative commons licence. You can open .gexf graph files with the opensource editor Gephi.

• The authors map in PDF format (July 2009)
• The authors map in SVG format (July 2009)
• The authors map in GEXF format (July 2009)

The distributions map has also been upgraded for the YAPC::EU. As for the authors map, we used CPANDB and Task::Kensho. We used modules recommendend by Task::Kensho as the seed. Then we extracted from CPANDB all the modules that depend on the modules listed in the seed. So this map is rather oriented towards the modern Perl. The TK group doesn’t appear anymore.

Discussions

Infos

• Node count: 2724
• Edge count: 5426

Nodes

A node represents a distribution. The color and the size represents the count of other distributions using it. Some green nodes have the same size as whites nodes. The color then distinguishes them because we didn’t want to use a smaller size for nodes.

Edges

An edge is created when a module from a distribution is used by another distribution. The edges are not weighted.

Study

Even if we don’t have the same data as for the previous map, we can do some comparaisons. We still see a Moose cluster, and Catalyst is now closer (Catalyst has switched to Moose in June). The Web cluster still exists (with LWP, URI, …).

Some other interesting cluster are the Test::*, with Test::Exception, Test::Simple, Test::Class, Test::POD, … If anyone still doubts that tests are important in the Perl community, I hope this map will convince them.

Downloads

These documents are all licenced under the creative commons licence. You can open .gexf graph files with the opensource editor Gephi.

• The distributions map in PDF format (July 2009)
• The distributions map in SVG format (July 2009)
• The distributions map in GEXF format (July 2009)

We used data available from the CPANTS database to reconstruct the relationships between developers inside the CPAN.

Discussions

Nodes

A graph node is an author : a Perl developer having commited a distribution (a software library) to CPAN.
An edge is a relationship between two authors : one (or both) of these two authors relies on the package developed by the other.

Layout

We used a force-directed based layout algorithm on the graph. Each node is affected by a repulsion force by all other nodes, except for nodes having a connection with him (attraction force).
As a result, the stronger the relationship is (ie. more packages involved), the nearest the two authors will be.
Some special authors

• Gisle Aas (main developer of LWP and URI)
• Dave Rolsky (DateTime, Moose, …)
• Adam Kennedy (> 200 dists)
• Michael Schwern (MakeMaker, Test::*)
• Ingy (YAML, Kwiki, IO::All, nearly 100 modules)
• Stevan Little (Moose, Class::MOP, …)
• Matt Sergeant (XML::*, SQLite, …)

The Modern Perl party

• Developers from Moose, Catalyst, DBIx::Class..
• In the upper left part – one can see Steven, Sartak, perigin, jrockway, mstrout, nothingmuch, marcus ramberg
• jrockway / steven / nothingmuch form a small cluster

Downloads

These documents are all under the creative commons licence. You can open .gexf graph files with the opensource editor Gephi 0.6. These graphs contains only the “core” of CPAN.

• The authors map in PDF format (march 2009)
• The authors map in SVG format (march 2009)
• The authors map in GDF format (march 2009)

A static view of the cpan community. You can zoom-in with the mouse scroll. Click on the “fullscreen” icon.

Dynamic Flash map

This map allow searches and interactive exploration.

Downloads

• CPAN Web Map – may 2009 (special version for impression, acrobat reader only)

We used data available from the CPANTS database to reconstruct the network of dependencies between packages inside the CPAN. After removing small independent sets, emerge the “core” of CPAN.

Discussions

We initialy used Graphviz Perl module to generate an hierarchical, butterfly-like view of the graph.

With this kind of representation, the package on which we focus is a trunk, dependencies are the roots, and other libraries which use the package are the leaves.
This is interesting, because Perl modules (like everything having a lifecycle) grow, evolve, and may die, or be recycled into another library.

But at the macro level, the full community appears as a complete ecosystem: a dynamic system in which each entity evolve with other species in order to adapt to their environment (other packages, industry tendencies, developers lifecycle..).
We believe that visualization and exploration of software repositories as large-scale dynamical networks is, and we hope to see similar initiatives in other softwares communities. We invite anybody interested in such network to download Gephi and the CPAN graph to create nice visualizations and discover interesting patterns!

Downloads

• The cpan core (march 2009) (pdf)
• The cpan core (march 2009) (svg)
• The cpan core (march 2009)