tuyenhn / dart Goto Github PK

If pixel(s) with the null sentinel value (e.g. -9999 for CHIRPS) inside the buffer around communes/wards, this will greatly affect the weighted average.
We may need to eliminate these nulls out of the buffer before calculate the weighted average. This is not implemented at the moment.

Why not writing

geo_feature %>%
   sf::st_cast("POINT", warn = FALSE) %>%
   sf::st_distance(.$centroid) %>%
   max()

instead of

dist <- geo_feature %>%
    sf::st_cast("POINT", warn = FALSE) %>%
    sf::st_distance(.$centroid)

max(dist)

I would remove the if test in the dart_max_dist() function has it has a huge computational cost. Furthermore I would replace the argument with_centroid = FALSE by centroid forcing the user to specify a column name of the geographical feature that contains centroid data.

I don't think we want sf::st_transform(sf::st_crs(4326)) at the end of dart_centroid()

I don't think we need sf::st_geometry() before calling sf::st_centroid().

Why not simply

if (!class(sf)[1] == "sf") {

instead of

if (!(is.data.frame(sf) && class(sf)[1] == "sf")) {

Isn't sf necessarily a data frame anyway?

The projection 9210 in

centroids <- dart_centroid(sf, 9210)

shouldn't be hard coded but passed as an argument instead.

remotes::install_github("s3818311/dart") typed locally failed. I supposed it's because the repository is private. Which bring the question: maybe we don't want to keep it private actually. To be discussed.

I think that a function like dart_centroid() should contain only the following code:

dart_centroid <- function(sf, target_crs = 4326) {
    sf %>%
       sf::st_transform(sf::st_crs(target_crs)) %>%
       sf::st_centroid()
}

Should be a parameter where the user can specify their id column.
Case in point:
buffers_sf <- buffers %>% sf::st_as_sf() %>% mutate(centroid = sf$centroid) %>% mutate(id = sf$ID_3)
Other sf might not have ID_3 as their chosen id column