Filters resulting in empty data frames throw unexpected error in ggforce 0.3.4

[Robrecht Cannoodt]

ggforce 0.3.4 seems to break a plotting function in dyngen. Note that I might implement a workaround in dyngen to avoid the package being removed from CRAN.

Edit: the example below now works because I implemented the workaround. See below for a smaller example that still errors.

Original example

Example:

install.packages("ggforce") # install ggforce 0.3.4
library(dyngen)
data("example_model")
plot_backbone_modulenet(example_model)
# Error in `check_required_aesthetics()`:
#   ! stat_edge_loop requires the following missing aesthetics: x, y, from, to, span, direction and strength
# Run `rlang::last_error()` to see where the error occurred.

When I install ggforce 0.3.3, the error is gone:

devtools::install_url("https://cran.r-project.org/src/contrib/Archive/ggforce/ggforce_0.3.3.tar.gz")
library(dyngen)
data("example_model")
plot_backbone_modulenet(example_model)

---

I managed to trim down the example. It appears the issue arises when the specified filter removes all of the rows from the data frame.

For example:

library(tidyverse)
library(ggraph)

nodes <- tribble(
  ~name,
  "A"
)
edges <- tribble(
  ~from, ~to, ~effect,
  "A", "A", 1
)
gr <- tidygraph::tbl_graph(nodes = nodes, edges = edges)
layout <- tribble(
  ~rowname, ~x, ~y,
  "A", 0, 0
) %>%
  as.data.frame() %>% 
  column_to_rownames()

# this works
ggraph(gr, layout = "manual", x = layout$x, y = layout$y) +
  geom_edge_loop(aes(filter = effect >= 0))

# this errors
ggraph(gr, layout = "manual", x = layout$x, y = layout$y) +
  geom_edge_loop(aes(filter = effect >= 0)) +
  geom_edge_loop(aes(filter = effect < 0))

[Malcolm Barrett]

I seem to be having a related issue in ggdag, also on a timeline for being removed (also working on a workaround!)

[Robrecht Cannoodt]

Feel free to take a look at my workaround: https://github.com/dynverse/dyngen/blob/master/R/ggforce_issue_273_workaround.R . It's definitely not the cleanest solution but it's only a temporary workaround so it'll do for now.

…

--- Darn it, I just noticed I posted the issue in the wrong repository. In all fairness, ggraph is the dependency that dyngen uses and not ggforce, which is why this bug was not caught by doing a revdepcheck.

On Fri, 26 Aug 2022, 22:45 Malcolm Barrett, ***@***.***> wrote: I seem to be having a related issue in ggdag, also on a timeline for being removed (also working on a workaround!) — Reply to this email directly, view it on GitHub <#324 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAEHFKX7CXPG5TWRE6DD75LV3EUGPANCNFSM57R4LQ7A> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[Thomas Lin Pedersen]

Can I get you to test it with the devel version of ggforce? I can't reproduce it right now (there has been a lot of upkeep so can't say what might have fixed the issue)

[David Schoch]

just tried it with the dev version of ggforce and got the following error with the code above

Error in `check_required_aesthetics()`:
! stat_edge_loop requires the following missing aesthetics: x, y, from, to, span, direction and strength
Run `rlang::last_error()` to see where the error occurred.

[Thomas Lin Pedersen]

oh yeah - I just ran the first example that seems to now work... I can reproduce the second example error

[Robrecht Cannoodt]

Indeed, I implemented a workaround for the original example. I updated the post to reflect this. It's a pity that the dev version of ggforce didn't magically solve the issue ;)

[Thomas Lin Pedersen]

It's an issue in ggraph not ggforce. Will be fixed soon

[Robrecht Cannoodt]

🙇 Thanks!

[Thomas Lin Pedersen]

Should work now with dev ggforce and dev graph (soon to be released both of them)... sorry about that I have no idea why it worked before 😬

[Robrecht Cannoodt]

Thanks a lot Thomas! I appreciate your efforts in finding and solving the problem so quickly.