ggplot not plotting linestring points from sf object - r

Question

Given the following data:
library(tidyverse)
library(sf)
df <- structure(list(geometry = c("LINESTRING (-85.76 38.34, -85.72 38.38)",
"LINESTRING (-85.46 38.76, -85.42 38.76)",
"LINESTRING (-85.89 38.31, -85.89 38.31)"
), var1 = c(4, 5, 6
), var2 = c(1, 2, 3
)), row.names = c(NA, -3L), class = c("tbl_df", "tbl", "data.frame"
))
df
df_sf <- sf::st_as_sf( df, wkt = "geometry" )
# Simple feature collection with 3 features and 2 fields
# geometry type: LINESTRING
# dimension: XY
# bbox: xmin: -85.89 ymin: 38.31 xmax: -85.42 ymax: 38.76
# CRS: NA
# # A tibble: 3 x 3
# geometry var1 var2
# <LINESTRING> <dbl> <dbl>
# 1 (-85.76 38.34, -85.72 38.38) 4 1
# 2 (-85.46 38.76, -85.42 38.76) 5 2
# 3 (-85.89 38.31, -85.89 38.31) 6 3
We can use plot to plot the data including the LINESTRING that has two points at the same location (row = 3):
plot(st_geometry(df_sf), lwd = 10)
giving:
but when we plot it using ggplot the point is dropped:
ggplot() +
geom_sf(data = df_sf, lwd = 8)
Without manually extracting locations that only contain a point, is there a quick way to tell ggplot to plot these? I can see that these points are technically not a line as theres no distance between them but plot is able to pick them up. This question seems related but slightly different, my LINESTRINGs are already created.
thanks

Answer 1

This is one of those situations where the headdesk emoji might come in handy:
ggplot() +
geom_sf(data = df_sf, lwd = 8, lineend = "round")
From the package's vignette on aesthetic specs, the default lineend is "butt", which stops precisely at the end point of a line (so a line of 0 length won't have anything to show), while the "round" alternative extends beyond the end point:
TL;DR: The 0-length linestring wasn't dropped. We just couldn't see it.

Answer 2

I think that you can solve that problem if you modify the LINESTRING geometries whose length is equal to 0 and cast them as POINTS. For example:
# packages
library(ggplot2)
library(sf)
#> Linking to GEOS 3.8.0, GDAL 3.0.4, PROJ 6.3.1
# data
df <- structure(
list(
geometry = c(
"LINESTRING (-85.76 38.34, -85.72 38.38)",
"LINESTRING (-85.46 38.76, -85.42 38.76)",
"LINESTRING (-85.89 38.31, -85.89 38.31)"
),
var1 = c(4, 5, 6),
var2 = c(1, 2, 3)
),
row.names = c(NA, -3L),
class = c("tbl_df", "tbl", "data.frame")
)
df_sf <- st_as_sf( df, wkt = "geometry" )
# Rebuild the geometry column in such a way that the zero-length LINESTRINGS are
# actually POINTS:
new_df_sf_geometry <- st_geometry(df_sf)
idx <- which(st_length(new_df_sf_geometry) == 0)
for (i in idx) {
new_df_sf_geometry[i] <- unique(st_cast(new_df_sf_geometry[i], "POINT"))
}
# This is the result
new_df_sf_geometry
#> Geometry set for 3 features
#> geometry type: GEOMETRY
#> dimension: XY
#> bbox: xmin: -85.89 ymin: 38.31 xmax: -85.42 ymax: 38.76
#> CRS: NA
#> LINESTRING (-85.76 38.34, -85.72 38.38)
#> LINESTRING (-85.46 38.76, -85.42 38.76)
#> POINT (-85.89 38.31)
# Replace the geometry
st_geometry(df_sf) <- new_df_sf_geometry
# Plot
ggplot(df_sf) +
geom_sf(size = 3)
Created on 2020-05-25 by the reprex package (v0.3.0)
If you need, you can also adopt more sophisticated approaches than a for-loop such as purrr::map_if.