Using purrr::map to get multiple columns in the output

Question

Data

I have 2 data sets:
* segments dataset represents the road segments ( lhrs .)
* hwys dataset represents the highways that contain individual lhrs .

> segments
# A tibble: 1 x 5
   lhrs mto_collision_ref_number latitude longitude highway_number
  <dbl>                    <dbl>    <dbl>     <dbl>          <dbl>
1 10004                  1549630     42.9     -78.9              1 


> hwys
# A tibble: 5 x 3
  STREET          longitude latitude
  <fct>               <dbl>    <dbl>
1 HIGHWAY 3           -80.0     42.9
2 ADELAIDE AVE E      -78.9     43.9
3 HOWARD AVE          -83.0     42.2
4 HIGHWAY 12          -79.7     44.7
5 CORONATION BLVD     -80.3     43.4

The problem

As you can see, the STREET column is missing in the segments dataset. I want to create this column in the segments dataset by finding the distance between a given lhrs and a STREET based on the longitude and latitude values. This means that I need to compare one set of long, lat of lhrs to all 5 STREET locations and find the one that has the minimum distance. I think this can be done using purrr package.

My code

I can find the distances between each lhrs and STREET using the geosphere::distVincentyEllipsoid() distance as follows:

library(tidyverse)



segments_nested <- segments %>% group_by(mto_collision_ref_number) %>% nest()


segments_nested %>% 
  mutate(diztances = purrr::map(
    data, ~ distVincentyEllipsoid(hwys %>% select(longitude, latitude),
                                             c(.$longitude, .$latitude)))) %>% 
  unnest(.preserve = data)


# A tibble: 5 x 3
  mto_collision_ref_number data             diztances
                     <dbl> <list>               <dbl>
1                  1549630 <tibble [1 x 4]>    85316.
2                  1549630 <tibble [1 x 4]>   110700.
3                  1549630 <tibble [1 x 4]>   342921.
4                  1549630 <tibble [1 x 4]>   213961.
5                  1549630 <tibble [1 x 4]>   125547.  

HOWEVER , I can't still figure out how to connect these distances with the STREET . Please guide me how I can use purrr::map to calculate the distances AS WELL AS the corresponding STREET . Once I have that, I can just group_by(mto_collision_ref_number) and get the summarize(min(diztances)) .

Answer 1

One way home is to take advantage of the flexibility in the anonymous function and use it to return an object that is already to spec. I used a combination of group_by() and transmute() .

# this is setup for transmute() so we keep 'STREET' around
hwys <- group_by(hwys, STREET) 

segments_nested %>%
  mutate(results = purrr::map(
    data, ~ transmute(hwys, diztances = geosphere::distVincentyEllipsoid(c(longitude, latitude),
                                             c(.$longitude, .$latitude))))) %>% 
  unnest(results)

And bingo 'STREET' is back on the menu boys !

  mto_collision_ref_number STREET         diztances
                     <int> <chr>              <dbl>
1                  1549630 HIGHWAY3          89840.
2                  1549630 ADELAIDEAVEE     111101.
3                  1549630 HOWARDAVE        345569.
4                  1549630 HIGHWAY12        210099.
5                  1549630 CORONATIONBLVD   126702.

In the future try to share your data in a easier to reproduce format, I prefer read.table(text = ) but dput() is also fine as suggested above. I had to copy, paste and manipulate your output chunk to get it into R:

segments <- read.table(
  text = "lhrs mto_collision_ref_number latitude longitude highway_number
  1 10004 1549630 42.9 -78.9 1",
  header = T,
  stringsAsFactors = F
)
hwys <- read.table(
  text = "  STREET longitude latitude
  1 HIGHWAY3 -80.0 42.9
  2 ADELAIDEAVEE  -78.9 43.9
  3 HOWARDAVE -83.0 42.2
  4 HIGHWAY12 -79.7 44.7
  5 CORONATIONBLVD -80.3 43.4",
  header = T, 
  stringsAsFactors = F
)