Get the latest non-NA value based on date column by group

library(dplyr)
library(tidyr)

df1 %>% 
  mutate(date = as.Date(date)) %>% 
  group_by(country_name) %>%
  arrange(date) %>%
  select(-date) %>% 
  fill(everything()) %>% 
  slice(n())

#> # A tibble: 2 x 4
#> # Groups:   country_name [2]
#>   country_name column_1 column_2 column_3
#>   <chr>           <dbl>    <int>    <dbl>
#> 1 UK                0.5        3     NA  
#> 2 US               10          3      7.3

Data:

read.table(text = "country_name  date         column_1 column_2 column_3
                   US            2016-11-02   7.5      NA       NA      
                   US            2017-09-12   NA       NA       9       
                   US            2017-09-19   NA       8        10      
                   US            2020-02-10   10       NA       NA      
                   US            2021-03-10   NA       NA       7.3     
                   US            2021-05-02   NA       3        NA      
                   UK            2016-11-02   NA       2        NA      
                   UK            2017-09-12   0.5      3        NA", 
           header = T, stringsAsFactors = F) -> df1

Update:

Thanks to @Darren Tsai handling the warning:

Warning: Problem while computing `..1 = across(-country_name, ~parse_number(.)).
i 1 parsing failure. row col expected actual 1 -- a number NA NA

Adding this line of code:

 mutate(across(-country_name, ~str_trim(str_replace_all(., 'NA', ''))))

library(tidyverse)
library(lubridate)

df1 %>% 
  mutate(date = ymd(date)) %>% 
  group_by(country_name) %>%
  arrange(date, .by_group = TRUE) %>% 
  summarise(across(starts_with("column"), ~paste(rev(.), collapse = ' '))) %>% 
  mutate(across(-country_name, ~str_trim(str_replace_all(., 'NA', '')))) %>% 
  mutate(across(-country_name, ~parse_number(.)))

  country_name column_1 column_2 column_3
  <chr>           <dbl>    <dbl>    <dbl>
1 UK                0.5        3     NA  
2 US               10          3      7.3

First answer:

Here is how we could do it:

1. If necessary transform date column to date class with ymd() function from lubridate.
2. group by country_name
3. Now comes the trick we use across for col1 col2... etc. and collapse in reverse with paste(rev(.).... to get the last value to first place. This is important for the next step.
4. Use parse_number() from readr package that will extract the first number!

library(dplyr)
library(lubridate)
library(readr)

df %>% 
  mutate(date = ymd(date)) %>% 
  group_by(country_name) %>%
  arrange(date, .by_group = TRUE) %>% 
  summarise(across(starts_with("column"), ~paste(rev(.), collapse = ' '))) %>% 
  mutate(across(-country_name, parse_number))

 country_name column_1 column_2 column_3
  <chr>           <dbl>    <dbl>    <dbl>
1 UK                0.5        3     NA  
2 US               10          3      7.3

You could na.omit and reverse each column and take first element. Then rbind. Take care of the right order and if it's as.Date formatted.

by(transform(dat, date=as.Date(date)), dat$country_name, \(x) {
  cbind(x[1, 1, drop=FALSE], 
        lapply(x[order(x$date), 3:5], \(z) {
          z <- el(rev(na.omit(z)))
          ifelse(length(z) == 1, z, NA_real_)
        }))
}) |> c(make.row.names=FALSE) |> do.call(what=rbind)
#   country_name column_1 column_2 column_3
# 1           UK      0.5        3       NA
# 2           US     10.0        3      7.3

Data:

dat <- structure(list(country_name = c("US", "US", "US", "US", "US", 
"US", "UK", "UK"), date = c("2016-11-02", "2017-09-12", "2017-09-19", 
"2020-02-10", "2021-03-10", "2021-05-02", "2016-11-02", "2017-09-12"
), column_1 = c(7.5, NA, NA, 10, NA, NA, NA, 0.5), column_2 = c(NA, 
NA, 8L, NA, NA, 3L, 2L, 3L), column_3 = c(NA, 9, 10, NA, 7.3, 
NA, NA, NA)), class = "data.frame", row.names = c(NA, -8L))

You can summarise each country across multiple columns with across(). The latest non-NA value can be subsetted by .x[date == max(date[!is.na(.x)])].

library(dplyr)

df %>%
  group_by(country_name) %>%
  summarise(across(starts_with("column"),
                   ~ if(all(is.na(.x))) NA else .x[date == max(date[!is.na(.x)])])) %>%
  ungroup()

# # A tibble: 2 × 4
#   country_name column_1 column_2 column_3
#   <chr>           <dbl>    <int>    <dbl>
# 1 UK                0.5        3     NA  
# 2 US               10          3      7.3

Another idea:

df %>% 
  group_by(country_name) %>%
  arrange(desc(date), .by_group = TRUE) %>% 
  summarise(across(starts_with("column"), ~ .x[!is.na(.x)][1])) %>% 
  ungroup()

Here is a base R solution. It uses two sapply calls: one for country and one for column.

foo <- structure(list(country_name = c("US", "US", "US", "US", "US", 
"US", "UK", "UK"), date = c("2016-11-02", "2017-09-12", "2017-09-19", 
"2020-02-10", "2021-03-10", "2021-05-02", "2016-11-02", "2017-09-12"
), column_1 = c(7.5, NA, NA, 10, NA, NA, NA, 0.5), column_2 = c(NA, 
NA, 8L, NA, NA, 3L, 2L, 3L), column_3 = c(NA, 9, 10, NA, 7.3, 
NA, NA, NA)), class = "data.frame", row.names = c(NA, -8L))


split(foo, foo$country_name)|>
  sapply( function(s) {
    s = s[order(s$date),]
    sapply(s[,3:5], function(x) {
       y = na.omit(x)
       ifelse(length(y)> 0, y[length(y)], NA) })}) |>
  t()

#   column_1 column_2 column_3
#UK      0.5        3       NA
#US     10.0        3      7.3