Do mathematical operation on multiple columns with comma-separated strings of numbers of unequal length

user_8039978 0
Viewed 96

I have a large dataframe in which some columns contain long strings of comma-separated numerical data of unequal lengths, here the columns A, B, and C:

df <- data.frame(
  id = 1:3,
  A = c("200, 100, 80, 100","120, 210, 220", "170, 200"),
  B = c("0.1, 0.2, 0.3","0.2, 0.3, 1.0, 0.4, 0.9", "0.55, 0.77, 0.99, 0.35"),
  C = c("700.1, 701.0, 699.2", "702.5, 702.9", "705.4, 705.4, 706.0")
)

I need to calculate for this numerical data in A, B, and C the percentage change. I take it that to facilitate this, I need to separate each number into its own row using separate_rows. But how can I do this step for all three columns A, B, and C in one go?

All I can do is do it column by column - first for A, then for B, and finally for C:

library(tidyverse)
df %>%
  # Step 1 - column `A`:
  separate_rows(A, sep = ",", convert = TRUE) %>%
  mutate(A_0 = lag((lead(A)-A)/A*100)) %>%
  group_by(id) %>%
  summarise(across(c(B,C), first),
    A = paste0(A, collapse = ", "),       
    A_0 = paste0(A_0, collapse = ", ")
    ) %>%
  ungroup() %>%
  # Step 2 - column `B`:
  separate_rows(B, sep = ",", convert = TRUE) %>%
  mutate(B_0 = lag((lead(B)-B)/B*100)) %>%
  group_by(id) %>%
  summarise(across(c(A,A_0,C), first),
            B = paste0(B, collapse = ", "),       
            B_0 = paste0(B_0, collapse = ", ")
            ) %>%
  ungroup() %>%
  # Step 3 - column `C`:
  separate_rows(C, sep = ",", convert = TRUE) %>%
  mutate(C_0 = lag((lead(C)-C)/C*100)) %>%
  group_by(id) %>%
  summarise(across(c(A,A_0,B,B_0), first),
            C = paste0(C, collapse = ", "),       
            C_0 = paste0(C_0, collapse = ", ")
  )
# A tibble: 3 × 7
     id A                 A_0                                 B        B_0           C      C_0      
  <int> <chr>             <chr>                               <chr>    <chr>         <chr>  <chr>    
1     1 200, 100, 80, 100 NA, -50, -20, 25                    0.1, 0.… NA, 100, 50   700.1… NA, 0.12…
2     2 120, 210, 220     20, 75, 4.76190476190476            0.2, 0.… -33.33333333… 702.5… 0.471967…
3     3 170, 200          -22.7272727272727, 17.6470588235294 0.55, 0… -38.88888888… 705.4… 0.355669…

Is there a better way?

tidyr dplyr r
2 Answers

We may loop across the columns, split at the , followed by one or more spaces (\\s+), loop over the list with map, convert to numeric, get the lag of the difference between lead and the current value proportion, pasted (toString) return as a character vector (_chr), and if needed order the columns within select

library(dplyr)
library(purrr)
df %>% 
  mutate(across(A:C, ~ {
         map_chr(strsplit(.x, ",\\s+"), ~ {
            tmp <- as.numeric(.x)
            toString(lag((lead(tmp)- tmp)/tmp *100))})
      }, .names = "{.col}_0")) %>% 
  select(id, gtools::mixedsort(names(.)[-1]))

-output

id                 A                      A_0                       B                                         B_0                   C
1  1 200, 100, 80, 100         NA, -50, -20, 25           0.1, 0.2, 0.3                                 NA, 100, 50 700.1, 701.0, 699.2
2  2     120, 210, 220 NA, 75, 4.76190476190476 0.2, 0.3, 1.0, 0.4, 0.9          NA, 50, 233.333333333333, -60, 125        702.5, 702.9
3  3          170, 200     NA, 17.6470588235294  0.55, 0.77, 0.99, 0.35 NA, 40, 28.5714285714286, -64.6464646464647 705.4, 705.4, 706.0
                                        C_0
1 NA, 0.128553063848018, -0.256776034236798
2                    NA, 0.0569395017793562
3                 NA, 0, 0.0850581230507546
user_37322710

Something like this can get you started

library(tidyverse)
#> Warning: package 'tidyr' was built under R version 4.1.3
#> Warning: package 'readr' was built under R version 4.1.3
#> Warning: package 'dplyr' was built under R version 4.1.3

df_example <- data.frame(
  id = 1:3,
  A = c("200, 100, 80, 100","120, 210, 220", "170, 200"),
  B = c("0.1, 0.2, 0.3","0.2, 0.3, 1.0, 0.4, 0.9", "0.55, 0.77, 0.99, 0.35"),
  C = c("700.1, 701.0, 699.2", "702.5, 702.9", "705.4, 705.4, 706.0")
) 

df_example |>
  as_tibble() |> 
  mutate(across(-id,
                .names =  "{.col}_0",
                \(x) x |> 
                  str_split(',') |> 
                  map(as.numeric) |> 
                  map(\(x) lag((lead(x)-x)/x*100)) |> 
                  map_chr(toString))
         )
#> # A tibble: 3 x 7
#>      id A                 B                       C            A_0   B_0   C_0  
#>   <int> <chr>             <chr>                   <chr>        <chr> <chr> <chr>
#> 1     1 200, 100, 80, 100 0.1, 0.2, 0.3           700.1, 701.~ NA, ~ NA, ~ NA, ~
#> 2     2 120, 210, 220     0.2, 0.3, 1.0, 0.4, 0.9 702.5, 702.9 NA, ~ NA, ~ NA, ~
#> 3     3 170, 200          0.55, 0.77, 0.99, 0.35  705.4, 705.~ NA, ~ NA, ~ NA, ~

Created on 2022-03-28 by the reprex package (v2.0.1)

user_71999660
Related