default_aesthetics in geom_sf does not handle NULL properly

[wesnm]

When plotting a map using a discrete fill scale, I get the error below. The source seems to be default_aesthetics not handling a NULL value for "type" (the line numbers may be off, I was adding some debugging prints to check out internal values):

ggplot(nyt$state.maps[!(STATE %in% c("Alaska", "Hawaii", "Puerto Rico"))]) +
    geom_sf(aes(geometry=GEOMETRY, fill=INFECTIONS))

 Error in if (type == "point") { : argument is of length zero 
18. default_aesthetics(params$legend) at geom-sf.R#113
17. modify_list(default_aesthetics(params$legend), data) at geom-sf.R#113
16. f(...) at ggproto.r#178
15. g$draw_key(g$data[i, ], g$params, key_size) at guide-legend.r#635
14. FUN(X[[i]], ...) 
13. lapply(guide$geoms, function(g) {
    g$draw_key(g$data[i, ], g$params, key_size)
}) at guide-legend.r#634
12. FUN(X[[i]], ...) 
11. lapply(seq_len(nbreak), draw_key) at guide-legend.r#639
10. unlist(lapply(seq_len(nbreak), draw_key), recursive = FALSE) at guide-legend.r#639
9. guide_gengrob.legend(X[[i]], ...) at guides-.r#356
8. FUN(X[[i]], ...) 
7. lapply(gdefs, guide_gengrob, theme) at guides-.r#249
6. guides_gengrob(gdefs, theme) at guides-.r#137
5. build_guides(plot$scales, plot$layers, plot$mapping, position, 
    theme, plot$guides, plot$labels) at plot-build.r#179
4. ggplot_gtable.ggplot_built(data) at plot-build.r#158
3. ggplot_gtable(data) at plot.r#173
2. print.ggplot(x) 
1. (function (x, ...) 
UseMethod("print"))(x) 

Passing a value for legend (or character value for show.legend) works fine:

ggplot(nyt$state.maps[!(STATE %in% c("Alaska", "Hawaii", "Puerto Rico"))]) +
    geom_sf(aes(geometry=GEOMETRY, fill=INFECTIONS_CUT), legend="meh")

You can repeat this using the data from this stack overflow post this stack overflow post and some small changes. Moving the fill aesthetic inside the aes mapping gives the error:

Error:
ggplot2::ggplot() + 
    geom_sf(data = gainsville_df, aes(geometry=Geomtry, fill=`Cluster Group`), alpha=0.2) +
    coord_sf(crs = "+init=epsg:4326")+
    scale_fill_manual(values = c("red", "grey", "seagreen3","gold", "green","orange"), name= "Cluster Group")

No Error:
ggplot2::ggplot() + 
    geom_sf(data = gainsville_df, aes(geometry=Geomtry, fill=`Cluster Group`), alpha=0.2, legend="meh") +
    coord_sf(crs = "+init=epsg:4326")+
    scale_fill_manual(values = c("red", "grey", "seagreen3","gold", "green","orange"), name= "Cluster Group")

[yutannihilation]

Confirmed. Here's a minimal reprex. The error happens when the data is NOT an sf and the scale of colour or fill is discrete.

library(ggplot2)

nc <- sf::read_sf(system.file("shape/nc.shp", package="sf"))
nc$BIR74_bin <- cut_number(nc$BIR74, 3)
nc_tibble <- tibble::as_tibble(nc)

ggplot(nc) +
  geom_sf(aes(fill = BIR74_bin))

ggplot(nc_tibble) +
  geom_sf(aes(geometry = geometry, fill = BIR74_bin))
#> Error in if (type == "point") {: argument is of length zero

^{Created on 2020-04-26 by the reprex package (v0.3.0)}

[clauswilke]

Specifically, things work the moment show.legend is specified (see below). This suggests the problem is in the legend detection code. I suspect this if branch needs an else:

ggplot2/R/layer-sf.R

Lines 51 to 59 in d3d47be

	if (is_sf(data)) {
	sf_type <- detect_sf_type(data)
	if (sf_type == "point") {
	self$geom_params$legend <- "point"
	} else if (sf_type == "line") {
	self$geom_params$legend <- "line"
	} else {
	self$geom_params$legend <- "polygon"
	}

library(ggplot2)

nc <- sf::read_sf(system.file("shape/nc.shp", package="sf"))
nc$BIR74_bin <- cut_number(nc$BIR74, 3)
nc_tibble <- tibble::as_tibble(nc)

# doesn't work
ggplot(nc_tibble) +
  geom_sf(aes(geometry = geometry, fill = BIR74_bin))
#> Error in if (type == "point") {: argument is of length zero

# works
ggplot(nc_tibble) +
  geom_sf(aes(geometry = geometry, fill = BIR74_bin), show.legend = "polygon")

^{Created on 2020-04-26 by the reprex package (v0.3.0)}

[clauswilke]

I've created a PR to fix this: #3964

[yutannihilation]

Related problem. Maybe p2 should show a line legend?

library(ggplot2)
library(patchwork)

p1 <- rbind(c(1,1), c(2,2), c(3,3))
s1 <- rbind(c(0,3), c(0,4), c(1,5), c(2,5))
s2 <- rbind(c(0.2,3), c(0.2,4), c(1,4.8), c(2,4.8))
s3 <- rbind(c(0,4.4), c(0.6,5))

d_sf <- sf::st_sf(
  g_point = sf::st_sfc(sf::st_multipoint(p1)),
  g_line = sf::st_sfc(sf::st_multilinestring(list(s1,s2,s3))),
  v = "a"
)

p1 <- ggplot(d_sf) +
  geom_sf(aes(geometry = g_point, colour = "a"))

p2 <- ggplot(d_sf) +
  geom_sf(aes(geometry = g_line, colour = "a"))

p1 / p2

^{Created on 2020-04-26 by the reprex package (v0.3.0)}

[yutannihilation]

I mean, the absence of else branch is a problem, but it seems we should use geometry aesthetic to determine detect_sf_type() in general.

[clauswilke]

Good point. I'll look into it tomorrow.

[clauswilke]

It's getting a bit convoluted, but here you go.

library(ggplot2)
library(patchwork)

p1 <- rbind(c(1,1), c(2,2), c(3,3))
s1 <- rbind(c(0,3), c(0,4), c(1,5), c(2,5))
s2 <- rbind(c(0.2,3), c(0.2,4), c(1,4.8), c(2,4.8))
s3 <- rbind(c(0,4.4), c(0.6,5))

d_sf <- sf::st_sf(
  g_point = sf::st_sfc(sf::st_multipoint(p1)),
  g_line = sf::st_sfc(sf::st_multilinestring(list(s1,s2,s3))),
  v = "a"
)

p1 <- ggplot(d_sf) +
  geom_sf(aes(geometry = g_point, colour = "a"))

p2 <- ggplot(d_sf) +
  geom_sf(aes(geometry = g_line, colour = "a"))

p1 / p2

^{Created on 2020-04-26 by the reprex package (v0.3.0)}