"Erase_water" function produces sf object that does not work with "mapview" operation in "mapview" package

[Steve-Koller]

I am using the excellent "erase_water" function (thank you for developing!) on a sf object called "MDC_tracts," which is comprised of census tract-level polygons of Miami-Dade County downloaded from tidycensus. The new sf object produced by erase_water, named "MDC_tracts_erased," does not render at all when running the operation

mapview(MDC_tracts_erased)

This stands in contrast to running mapview on the pre-erasure sf object (code below), which runs fine.

mapview(MDC_tracts)

Are sf object outputs from erase_water known to be incompatible with "mapview?" Is there a way to use sf objects that have been altered by "erase_water" with mapview? Thank you!

[walkerke]

@Steve-Koller it should work with mapview. This is working fine for me:

library(tidycensus)
library(tigris)
options(tigris_use_cache = TRUE)
sf::sf_use_s2(FALSE)

dade_income <- get_acs(
  geography = "tract",
  variables = "B19013_001",
  state = "FL",
  county = "Miami-Dade",
  geometry = TRUE,
  cb = FALSE,
  year = 2021
)

dade_income_erase <- erase_water(
  dade_income, 
  area_threshold = 0.95,
  year = 2021
)

mapview::mapview(dade_income_erase,
                 zcol = "estimate",
                 layer.name = "Median household<br/>income, 2017-2021 ACS")

A couple questions for you:

• Are you using a low area_threshold value? In water-rich areas like Miami, a low value (say, below 0.5) might create very complex polygons that mapview will struggle to draw.
• Are you using long-form tidycensus data with multiple variables, and in turn multiple rows per Census tract? This will give mapview a lot of polygons to draw and may crash it. If you need multiple variables, I'd recommend getting wide-form data from tidycensus with the argument output = "wide" then erasing the water. It'll make the erasing much faster as well!

[Steve-Koller]

Thank you so much! Bingo- I hadn't been including the below two lines. After adding those in, mapview worked perfectly.

options(tigris_use_cache = TRUE) sf::sf_use_s2(FALSE)

To your questions:

• I had been using a high area_threshold value (0.95). When using the default 0.75 it took my computer a while to execute, so I increased the threshold value.
• I had been using tidycensus data with multiple variables using wide format (17 columns), and the imported object had 707 rows. Given my setup this sf object is one that I previously pulled from tidycensus (another amazing package, thank you!) and have saved locally, so I used st_read to get it into RStudio instead of get_acs

Much appreciated!