Feature request: geom_sf_label_repel(), geom_sf_text_repel()

[yutannihilation]

Summary

ggplot2 will (hopefully) get geom_sf_label() and geom_sf_text() for labeling sf objects. (c.f. https://github.com/tidyverse/ggplot2/issues/2742#issuecomment-403315573) Are you interested in implementing the _repel version of these? Labelling sf objects is a very common task so I'm sure many people will find geom_sf_label_repel() useful.

If you feel ggrepel is not the right place for them, they will probably be implemented in ggforce.

Minimal code example

Proof of Concept is here: https://yutannihilation.github.io/ggsflabel/#geom_label_repel-for-sf

Suggestions

Note that this can be addressed after geom_sf_label() and geom_sf_text() are implemented in ggplot2. So, please stay tuned on tidyverse/ggplot2#2742! (I will do my best to finish it as soon as possible)

Version information

N/A

[slowkow]

Yes, I'd be interested to include those functions in ggrepel if you think it makes sense, and we're all in agreement across ggplot2, ggrepel, and ggforce. Looks like you have a great start already! Thanks for opening the issue!

I think I need to study the *_sf_* functions a bit to figure out exactly what is going on. I would have assumed that geom_text_repel() should be sufficient without an extra function for spatial data, but clearly this is not correct.

[yutannihilation]

Thanks for the quick reply. Glad to hear that!

I would have assumed that geom_text_repel() should be sufficient without an extra function for spatial data

I think you are right in some sense, but, you need to extract the coordinates by yourself (c.f. https://github.com/tidyverse/ggplot2/issues/2111) so we naturally want some shortcut for this. That is, *_sf_* functions, in my understanding :)

[yutannihilation]

Now geom_sf_text_repel() and geom_sf_label_repel() are basically possible with stat_sf_coordinates(). Example:

# github version of ggplot2
library(ggplot2)

nc <- sf::st_read(system.file("shape/nc.shp", package="sf"))
#> Reading layer `nc' from data source `/Library/Frameworks/R.framework/Versions/3.5/Resources/library/sf/shape/nc.shp' using driver `ESRI Shapefile'
#> Simple feature collection with 100 features and 14 fields
#> geometry type:  MULTIPOLYGON
#> dimension:      XY
#> bbox:           xmin: -84.32385 ymin: 33.88199 xmax: -75.45698 ymax: 36.58965
#> epsg (SRID):    4267
#> proj4string:    +proj=longlat +datum=NAD27 +no_defs
#> Reading layer `nc' from data source `/path/to/sf/shape/nc.shp' using driver `ESRI Shapefile'
#> Simple feature collection with 100 features and 14 fields
#> geometry type:  MULTIPOLYGON
#> dimension:      XY
#> bbox:           xmin: -84.32385 ymin: 33.88199 xmax: -75.45698 ymax: 36.58965
#> epsg (SRID):    4267
#> proj4string:    +proj=longlat +datum=NAD27 +no_defs

ggplot(nc) +
  geom_sf() +
  ggrepel::geom_label_repel(
    data = head(nc),
    aes(label = NAME, geometry = geometry),
    stat = "sf_coordinates",
    min.segment.length = 0
  )
#> Warning in st_point_on_surface.sfc(sf::st_zm(x)): st_point_on_surface may
#> not give correct results for longitude/latitude data

Created on 2018-09-04 by the reprex package (v0.2.0).

[slowkow]

@yutannihilation Do you think we should create two new functions?

• geom_sf_text_repel()
• geom_sf_label_repel()

I suppose you could say they are not needed anymore, since you've shown that we can use stat = "sf_coordinates" successfully for many types of figures. On the other hand, it would be convenient to have a wrapper.

What do you think? I think it makes sense to add them.

Whether or not we add two more functions, I think the vignette for ggrepel needs an example showing how to put repulsive labels on a sf plot.

If you're interested, I'd be very happy to review your pull request and merge it in. If you prefer, I can try to do it myself, but I'd appreciate your comments to make sure I'm writing the functions the way you'd expect.

---

As an aside, you're also making me wonder if I should think about a stat like stat_repel() to enable repulsive positioning for other types of elements besides text boxes... I think I need to spend more time studying the ggplot2 code before I'm ready for this.

[yutannihilation]

Yes, I think these functions are worth adding, of course! I want to implement them if you are ok (Sorry for not yet starting on this...).

While it seems easy to implement them as simple wrappers, I'm wondering if I can add an option to prevent labels to overwrap the geometry. Maybe is this the same thing as your concept of stat_repel()? I too have to study the ggplot2 code. 🤔

[slowkow]

Ok, feel free to get started whenever you have time. Thanks for your help!

I like the idea of avoiding collision with the shape of the geometry. This might be a challenge to implement, but I have an inkling that there might be library we can use to do that.

[yutannihilation]

Thanks, I will try :)

[kendonB]

I am facing problems trying to get geom_label_repel with a geom_sf. One neat feature would be to also have it repel from sf features themselves. In my case, I'm trying to map NZ regions with the labels falling in the ocean.

[bjsmith]

Working on the same problem as @kendonB .

[CeresBarros]

Any news on this front? have geom_sf_label/text_repel() been implemented? Thanks :)

[slowkow]

@CeresBarros Did you try the code example by @yutannihilation ?

Did it meet your needs?

[CeresBarros]

I did not - I wasn't sure if that was the current "solution". I'll try it :)

[CeresBarros]

Just tried it, works great, thanks!

[francisbarton]

Just come across this issue - getting "Error: geom_label_repel requires the following missing aesthetics: x and y" message when using geom_label_repel with a geom_sf plot. I was really confused because there's nothing in the ggrepel docs saying anything about geom_label_repel requiring x and y aesthetics! Searching the web for that error message didn't get me any useful info, so I came here to search.

@yutannihilation 's solution works great, but I do think this should be covered in the ggrepel examples vignette. So that future users don't have to hunt around for the fix. Would you welcome a PR to your vignette to include an example of this?

[slowkow]

@francisbarton Contributions to the documentation are welcome! Please feel free to make a PR with your suggested changes or additions.

[fariadamasceno]

Hi, just to let you know that I would be one more who would appreciate the wrapper functions. =) The code from @yutannihilation worked nicely, thanks!

[slowkow]

I'd be happy to review and merge a pull request that resolves this issue.

[Steve-Koller]

Not sure if this is the right place to bring this up, but I'm wondering if it's possible to selectively apply geom_sf_label_repel() to certain features in an sf object and geom_sf_label to other features in the same object? I'm currently trying to do this but unable. Can share photos or a reprex if that would help.

Below is sample code, where "geom_repel_states" is a vector with a few states' FIPS codes. These are the states whose labels I'd like to repel, since their labels overlap when including in geom_sf_label().

Thank you!

example_plot = ggplot(example_sf) + 
    geom_sf(color = "black", aes(fill = example_variable_name)) +
    geom_sf_label_repel(aes(label = ifelse(example_sf$state_fips %in% geom_repel_states, example_variable_name, ""))) + 
    geom_sf_label(aes(label=ifelse(!(example_sf$state_fips %in% geom_repel_states), example_variable_name, ""))) 

  example_plot

[slowkow]

@Steve-Koller This isn't really the best place to ask (better would be a new issue, best would be a new question on Stackoverflow), but since you asked...

Each geom_*() function takes a data argument. You can use geom_text_repel(data = ...) to plot a subset of the original data. The ... might be another dataframe like my_subdata or it might also be a function like function(d) d %>% filter(my_condition).

[slowkow]

We added an example to the ggrepel documentation. Feel free to comment below with your thoughts, and let me know if you believe we need to keep this issue open.

https://ggrepel.slowkow.com/articles/examples.html#label-sf-objects

# thanks to Hiroaki Yutani 
# https://github.com/slowkow/ggrepel/issues/111#issuecomment-416853013

library(ggplot2)
library(sf)

nc <- sf::st_read(system.file("shape/nc.shp", package="sf"), quiet = TRUE)

ggplot(nc) +
  geom_sf() +
  ggrepel::geom_label_repel(
    data = head(nc),
    aes(label = NAME, geometry = geometry),
    stat = "sf_coordinates",
    min.segment.length = 0
  )