fix: automatically arrange loops in plot()

[schochastics]

fix #407
fix #556

devtools::load_all("~/git/R_packages/rigraph/")
#> ℹ Loading igraph
g1 <- make_graph(c(1, 1))
g2 <- make_graph(c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1))
plot(g1)

plot(g2, margin = 0.1)

^{Created on 2025-06-24 with reprex v2.1.1}

[szhorvat]

This is definitely an improvement over the current situation, but I'd consider it a makeshift solution until something like this can be implemented:

This show the issues better:

g2 <- make_graph(c(1,2, 2,3, 3,1, 1,4, 4,5, 5,1, 3,4, 5,2, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1))
plot(g2)

Problems:

• By adding more loops, the loop sizes grow, eventually intersecting edges.
• The loop direction seems fixed, so it even intersects edges that are incident to the loop's own vertex. Thus for a high quality solution, it is unavoidable to determine "where the loops fits" between the other incident edges of a vertex. At that point, having different angles for different loops is a small addition.

Compare:

More plots:

g2 <- make_graph(c(1,2, 2,3, 3,1, 1,4, 4,5, 5,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1))
plot(g2)

[szhorvat]

In short: I would merge this, as it is an improvement, but I disagree with closing #407, as it's not a complete fix.

[schochastics]

Thanks. I will have a look if I can implement an angle version instead and otherwise I would merge as is

[szhorvat]

I will have a look if I can implement an angle version

The examples I showed are from Mathematica, but the angle version is implemented in Python as well, so that might be a useful reference.

I think doing this might be quite a bit of work. IMO it'd make sense to merge the improvement in this PR, as it's already much better than the current situation. Further improvements can always come later.

[schochastics]

You are right that these seems complicated. But the least that I will try is to place the loops more smartly in free space. This does not look very nice

library(igraph)
g <- make_graph(c(1,2,2,3,3,1,1,1,1,1,1,1,1,1),directed = FALSE)
V(g)$x <- c(1,1.5,2)
V(g)$y <- c(0,1,0)
plot(g,margin = 0.1)

devtools::load_all("~/git/R_packages/rigraph")
#> ℹ Loading igraph
plot(g,margin = 0.1)

^{Created on 2025-06-24 with reprex v2.1.1}

I am aware of edge.loop.angle, but this should probably be possible to calculate dynamically

[schochastics]

Getting closer. loops are now placed in the largest gap available. Need to adjust the handling of edge.loop.angle. Idea: default to NA to use dynamic calculation.

devtools::load_all("~/git/R_packages/rigraph")
#> ℹ Loading igraph

g2 <- make_graph(c(1,2, 2,3, 3,1, 1,4, 4,5, 5,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1))
plot(g2)

g2 <- make_graph(c(1,2, 2,3, 3,1, 1,4, 4,5, 5,1, 3,4, 5,2, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1))
plot(g2)

^{Created on 2025-06-24 with reprex v2.1.1}

[schochastics]

@szhorvat how is this? Might need some parameter tweaking but I think overall it seems fine

devtools::load_all("~/git/R_packages/rigraph")
#> ℹ Loading igraph

g <- make_graph(c(1,2,2,3,3,1,1,1,1,1,1,1,1,1),directed = FALSE)
V(g)$x <- c(1,1.5,2)
V(g)$y <- c(0,1,0)
plot(g,margin = 0.2,loop.size=2)

g2 <- make_graph(c(1,2, 2,3, 3,1, 1,4, 4,5, 5,1, 3,4, 5,2, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1),directed = FALSE)
plot(g2,loop.size = 2)

^{Created on 2025-06-25 with reprex v2.1.1}

[schochastics]

need to fix label placement

[szhorvat]

Awesome! Does the loop angular width depend on the loop count and available angular space? Or it if fixed? (No time to compile and test until evening today.)

[schochastics]

Currently depending only on the loop count. I still have to make it adapt to the available space too

[szhorvat]

A nitpick (don't take it too seriously):

Ideally, the lines would converge on the centre of the node. They should not touch on the boundary of the node.

Do test with square node shapes if you haven't yet! There were bugs in the past with this shape.

[szhorvat]

Do you think it's a good idea to keep the original version (loops within loops) as an option? I haven't seen this style before. It's not actually bad, it's just different. Once the loop is fit in the largest space (which you implemented), it could be an interesting option.

[schochastics]

Now its implemented. Will check on the nitpick and rectangular nodes.
You mean you havent seen the "loop in loop" before? I actually have never seen the "flower pedals" style before! Personally, I think we should decide on one style. I am fine with the "pedals". Keeping both makes the already convoluted plotting code even worse.

[schochastics]

Ok I think we might be done now with the main tasks (I will try to tweak the needed parameters a bit to make it a little nicer)

devtools::load_all("~/git/R_packages/rigraph")
#> ℹ Loading igraph

g <- make_graph(c(1,2,2,3,3,1,1,1,1,1,1,1,1,1),directed = FALSE)
V(g)$x <- c(1,1.5,2)
V(g)$y <- c(0,1,0)
plot(g,margin = 0.4,loop.size = 2, vertex.size=10)

g2 <- make_graph(c(1,2, 2,3, 3,1, 1,4, 4,5, 5,1, 3,4, 5,2, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1, 1,1),directed = FALSE)
plot(g2,loop.size = 2,vertex.shape="rectangle")

^{Created on 2025-06-25 with reprex v2.1.1}

[schochastics]

I dont understand where these errors come from 😞
Error is legit, need to fix

[krlmlr]

Thanks, let's strive for consistency.