dplyr mutate/replace several columns on a subset of rows

These solutions (1) maintain the pipeline, (2) do not overwrite the input and (3) only require that the condition be specified once:

1a) mutate_cond Create a simple function for data frames or data tables that can be incorporated into pipelines. This function is like mutate but only acts on the rows satisfying the condition:

mutate_cond <- function(.data, condition, ..., envir = parent.frame()) {
  condition <- eval(substitute(condition), .data, envir)
  .data[condition, ] <- .data[condition, ] %>% mutate(...)
  .data
}

DF %>% mutate_cond(measure == 'exit', qty.exit = qty, cf = 0, delta.watts = 13)

1b) mutate_last This is an alternative function for data frames or data tables which again is like mutate but is only used within group_by (as in the example below) and only operates on the last group rather than every group. Note that TRUE > FALSE so if group_by specifies a condition then mutate_last will only operate on rows satisfying that condition.

mutate_last <- function(.data, ...) {
  n <- n_groups(.data)
  indices <- attr(.data, "indices")[[n]] + 1
  .data[indices, ] <- .data[indices, ] %>% mutate(...)
  .data
}


DF %>% 
   group_by(is.exit = measure == 'exit') %>%
   mutate_last(qty.exit = qty, cf = 0, delta.watts = 13) %>%
   ungroup() %>%
   select(-is.exit)

2) factor out condition Factor out the condition by making it an extra column which is later removed. Then use ifelse, replace or arithmetic with logicals as illustrated. This also works for data tables.

library(dplyr)

DF %>% mutate(is.exit = measure == 'exit',
              qty.exit = ifelse(is.exit, qty, qty.exit),
              cf = (!is.exit) * cf,
              delta.watts = replace(delta.watts, is.exit, 13)) %>%
       select(-is.exit)

3) sqldf We could use SQL update via the sqldf package in the pipeline for data frames (but not data tables unless we convert them -- this may represent a bug in dplyr. See dplyr issue 1579). It may seem that we are undesirably modifying the input in this code due to the existence of the update but in fact the update is acting on a copy of the input in the temporarily generated database and not on the actual input.

library(sqldf)

DF %>% 
   do(sqldf(c("update '.' 
                 set 'qty.exit' = qty, cf = 0, 'delta.watts' = 13 
                 where measure = 'exit'", 
              "select * from '.'")))

4) row_case_when Also check out row_case_when defined in Returning a tibble: how to vectorize with case_when? . It uses a syntax similar to case_when but applies to rows.

library(dplyr)

DF %>%
  row_case_when(
    measure == "exit" ~ data.frame(qty.exit = qty, cf = 0, delta.watts = 13),
    TRUE ~ data.frame(qty.exit, cf, delta.watts)
  )

Note 1: We used this as DF

set.seed(1)
DF <- data.frame(site = sample(1:6, 50, replace=T),
                 space = sample(1:4, 50, replace=T),
                 measure = sample(c('cfl', 'led', 'linear', 'exit'), 50, 
                               replace=T),
                 qty = round(runif(50) * 30),
                 qty.exit = 0,
                 delta.watts = sample(10.5:100.5, 50, replace=T),
                 cf = runif(50))

Note 2: The problem of how to easily specify updating a subset of rows is also discussed in dplyr issues 134, 631, 1518 and 1573 with 631 being the main thread and 1573 being a review of the answers here.

You can do this with magrittr's two-way pipe %<>%:

library(dplyr)
library(magrittr)

dt[dt$measure=="exit",] %<>% mutate(qty.exit = qty,
                                    cf = 0,  
                                    delta.watts = 13)

This reduces the amount of typing, but is still much slower than data.table.

Here's a solution I like:

mutate_when <- function(data, ...) {
  dots <- eval(substitute(alist(...)))
  for (i in seq(1, length(dots), by = 2)) {
    condition <- eval(dots[[i]], envir = data)
    mutations <- eval(dots[[i + 1]], envir = data[condition, , drop = FALSE])
    data[condition, names(mutations)] <- mutations
  }
  data
}

It lets you write things like e.g.

mtcars %>% mutate_when(
  mpg > 22,    list(cyl = 100),
  disp == 160, list(cyl = 200)
)

which is quite readable -- although it may not be as performant as it could be.

As eipi10 shows above, there's not a simple way to do a subset replacement in dplyr because DT uses pass-by-reference semantics vs dplyr using pass-by-value. dplyr requires the use of ifelse() on the whole vector, whereas DT will do the subset and update by reference (returning the whole DT). So, for this exercise, DT will be substantially faster.

You could alternatively subset first, then update, and finally recombine:

dt.sub <- dt[dt$measure == "exit",] %>%
  mutate(qty.exit= qty, cf= 0, delta.watts= 13)

dt.new <- rbind(dt.sub, dt[dt$measure != "exit",])

But DT is gonna be substantially faster: (editted to use eipi10's new answer)

library(data.table)
library(dplyr)
library(microbenchmark)
microbenchmark(dt= {dt <- dt[measure == 'exit', 
                            `:=`(qty.exit = qty,
                                 cf = 0,
                                 delta.watts = 13)]},
               eipi10= {dt[dt$measure=="exit",] %<>% mutate(qty.exit = qty,
                                cf = 0,  
                                delta.watts = 13)},
               alex= {dt.sub <- dt[dt$measure == "exit",] %>%
                 mutate(qty.exit= qty, cf= 0, delta.watts= 13)

               dt.new <- rbind(dt.sub, dt[dt$measure != "exit",])})


Unit: microseconds
expr      min        lq      mean   median       uq      max neval cld
     dt  591.480  672.2565  747.0771  743.341  780.973 1837.539   100  a 
 eipi10 3481.212 3677.1685 4008.0314 3796.909 3936.796 6857.509   100   b
   alex 3412.029 3637.6350 3867.0649 3726.204 3936.985 5424.427   100   b

I just stumbled across this and really like mutate_cond() by @G. Grothendieck, but thought it might come in handy to also handle new variables. So, below has two additions:

Unrelated: Second last line made a bit more dplyr by using filter()

Three new lines at the beginning get variable names for use in mutate(), and initializes any new variables in the data frame before mutate() occurs. New variables are initialized for the remainder of the data.frame using new_init, which is set to missing (NA) as a default.

mutate_cond <- function(.data, condition, ..., new_init = NA, envir = parent.frame()) {
  # Initialize any new variables as new_init
  new_vars <- substitute(list(...))[-1]
  new_vars %<>% sapply(deparse) %>% names %>% setdiff(names(.data))
  .data[, new_vars] <- new_init

  condition <- eval(substitute(condition), .data, envir)
  .data[condition, ] <- .data %>% filter(condition) %>% mutate(...)
  .data
}

Here are some examples using the iris data:

Change Petal.Length to 88 where Species == "setosa". This will work in the original function as well as this new version.

iris %>% mutate_cond(Species == "setosa", Petal.Length = 88)

Same as above, but also create a new variable x (NA in rows not included in the condition). Not possible before.

iris %>% mutate_cond(Species == "setosa", Petal.Length = 88, x = TRUE)

Same as above, but rows not included in the condition for x are set to FALSE.

iris %>% mutate_cond(Species == "setosa", Petal.Length = 88, x = TRUE, new_init = FALSE)

This example shows how new_init can be set to a list to initialize multiple new variables with different values. Here, two new variables are created with excluded rows being initialized using different values (x initialised as FALSE, y as NA)

iris %>% mutate_cond(Species == "setosa" & Sepal.Length < 5,
                  x = TRUE, y = Sepal.Length ^ 2,
                  new_init = list(FALSE, NA))

dplyr mutate/replace several columns on a subset of rows

Related

Recent Posts