Create gf_ functions to support choropleth maps? about ggformula

[rpruim]

@dtkaplan proposed something along the lines of

gf_choropleth(income ~ state, data = Crimes, map = US)

This looks clean, but still leaves some things to be determined:

1. What aesthetic is income mapped to?

• fill most likely, but at least alpha and fill make sense here. Are we going to force one? Allow either? Have two functions, one for each?

2. What data structure is map?

• must users come up with a map? are we only supporting certain maps?

3. How are data and map merged?

• presumably there must be a key that is used to merge the data together. how is it specified?

4. merging the data can be slow. For repeated maps that differ only in how the data are displayed and not the data themselves, it would be nice to avoid repeatedly doing the merge. Perhaps it is better to think of map creation as a two step process. In step one, the data are prepared and merged. Once that has happened, the plot itself can be created and modified.
5. income is not "positional" and so doesn't fit as well with our current formula approach. (At least it is unusual in this regard.)
6. Given some of the other mapping options out there, is this the best place to devote ggformula resources?

[rpruim]

Now that ggplot2 will have geom_sf(), perhaps we can build gf_sf() to mimic. Need to explore.

[rpruim]

First attempt is functional. Improvements/examples/documentation may be added/refined later.

library(ggformula)
theme_set(theme_bw())
library(maps)
MI <- sf::st_as_sf(map("county", "michigan", plot = FALSE, fill = TRUE))
gf_sf(fill = ~factor(nchar(ID)), data = MI) %>% gf_refine(coord_sf())

[dtkaplan]

This is great!

[nicholasjhorton]

I concur. But I now worry that the documentation/overview is out of date for both ggformula and mosaic and wonder if a separate issue should be created about how to help instructors get a comprehensive list of the functionality of these packages.

[rpruim]

I've added an example to the vignette.

if (require(maps) && require(mosaic) && require(dplyr) && require(sf) && require(purrr)) {
  USshape <- 
    sf::st_as_sf(maps::map('state', plot = FALSE, fill = TRUE)) %>%
    dplyr::mutate(
      name_length = nchar(ID),
      centroid = purrr::map(geometry, st_centroid),
      coords_x = purrr::map_dbl(centroid, 1),
      coords_y = purrr::map_dbl(centroid, 2)
    ) 
    
  gf_sf(fill = ~ factor(name_length), color = "white", data = USshape, alpha = 0.5) %>%
    gf_sf(data = sf::st_centroid(USshape), color = "white", alpha = 0.5, size = 3) %>%
    gf_text(coords_y ~ coords_x, label = ~ ID, color = "gray20", size = 2) %>%
    gf_labs(x = "", y = "") %>%
    gf_refine(mosaic::theme_map(), theme_bw()) %>%
    gf_labs(fill = "name length")
}

Created on 2018-06-01 by the reprex package (v0.2.0).