min_dist_to_points.Rmd

---
output: github_document
---
# Fast min distance calculator

Calculate the min distance between a point and another set of points quickly.

Original use case:
<br>Had 1e6 points in grid.
<br>Needed min distance for evry point on grid out to set of data collection points.

Result below performed fastest.
<br>Other tests included
<br>1. Straight dplyr without multiprocessing
<br>2. Multidplyr, created memory issues
<br>3. Converting to matrix and using matrix operations
<br>4. Other flavors of lapply or purrr::map
```{r eval=FALSE}
library(tidyverse)
library(parallel)

raster_df_list = raster_df %>%
  split((seq(nrow(.)) - 1) %/% 1000)

no_cores = detectCores()
cl = makeCluster(no_cores)

clusterExport(cl, "raster_df_list") %>% invisible()
clusterExport(cl, "source_locations_df") %>% invisible()
clusterEvalQ(cl, library(tidyverse)) %>% invisible()

raster_df_w_min_distance = parLapply(cl,
          raster_df_list,
          function(raster_df_iter){
            raster_df_iter %>%
              mutate(row_num = 1:nrow(.)) %>%
              group_by(row_num) %>%
              mutate(min_distance_to_source = source_locations_df %>%
                       mutate(distance_iter = sqrt((raster_LONGITUDE - source_LONGITUDE)^2 + 
                                                     (raster_LATITUDE - source_LATITUDE)^2)) %>%
                       select(distance_iter) %>%
                       min()) %>% 
              ungroup() %>%
              select(-row_num)
}) %>%
  bind_rows()
stopCluster(cl)
```