Fixing things in drawing functions

[maximelucas]

• addressed sep argument not working multilayer plotting function #433 @thomasrobiglio can you check?
• fixed default settings in draw() ignored if one of them is passed as argument #443 (one function was missing)
• I tried to make draw_dihypergraph() more consistent with the other (updated) functions. I solved some bugs on the way, but making it fully consistent (and bug free) will require a bigger rewrite. I'll open an issue.
• Added some tests for draw_dihypergraph(). More needed.

[codecov]

Codecov Report

Attention: 21 lines in your changes are missing coverage. Please review.

Comparison is base (536f592) 92.03% compared to head (ac95a64) 92.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #476      +/-   ##
==========================================
- Coverage   92.03%   92.00%   -0.03%     
==========================================
  Files          60       60              
  Lines        4392     4429      +37     
==========================================
+ Hits         4042     4075      +33     
- Misses        350      354       +4     

Files	Coverage Δ
xgi/drawing/layout.py	99.18% <100.00%> (-0.02%)	⬇️
xgi/drawing/draw_utils.py	88.42% <50.00%> (-0.19%)	⬇️
xgi/stats/__init__.py	85.46% <50.00%> (-0.42%)	⬇️
xgi/drawing/draw.py	77.94% <81.18%> (+1.62%)	⬆️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[thomasrobiglio]

thanks max! I'll check about #433

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[maximelucas]

This is ready for review.

With the last commits, I rewrote draw_dihypergraph. The new implementation:

• uses draw_nodes() to draw both types of nodes

• uses individual FancyArrowPatch for the edges. This mimicks Networkx for directed graphs. I commented in the code why this seems to be the best solution (see this discussion). Arrows can be parametrised with arrowstyle, arrowsize, connectionstyle.

• solves bugs when setting colors/linewidths/etc.

• is now consistent with the other new functions, and drops dependency of _color_arg_to_dict and _scalar_arg_to_dict.

• has more tests

• has a new argument iterations to control the layout (related to Random edge position for dyads? #396). The previous layout was a spring layout computed on the "augmented projection", so that "edge"-nodes were sometimes on the outside, not at the barycenters. Now, the layout is computed as follows:

  1. compute a spring layout based solely on the original nodes.
  2. add the "edge"-nodes at the barycenters (at this point, nodes may overlap in larger networks)
  3. starting from this initial layout, run a spring layout on the whole thing, using iterations iterations. So 0 iterations will yield the layout computed in ii, while many iterations will yield the layout that we used to have.

  There is also a new pos argument so that user can play around. We could provide a special layout function for this in the future if we find a good way to do it.

• has a new In Depth tutorial

[thomasrobiglio]

Thanks @maximelucas this looks super good! I have added minor comments. One more small thing:

I have noticed that we were inconsistent in the naming of the in-depth tutos (see pic). Can you:
In depth 2 - Drawing hyperedges -> In depth: drawing hyperedges
In Depth - Drawing DiHypergraphs -> In depth: drawing DiHypergraphs
(or something else to make the three names consistent?)

[maximelucas]

Good points Thomas, thanks! All addressed now I think.

[nwlandry]

Great work, @maximelucas!! Much needed changes and thanks for pushing this through!