modify, a function that wraps data.table for quick replacement of values in selected rows

modify

library(dplyr)
library(data.table)

EmptySymbol = function() (quote(f(,)))[[2]]

get.data.table.modify.call = function(args=NULL, filter.call=NULL, by=NULL, dat.quote=quote(dt)) {
  if (length(args)==1) {
    com = call(":=",names(args)[1],args[[1]])
  } else {
    com = as.call(c(list(quote(`:=`)),args))
  }
  
  if (is.null(filter.call)) {
    ca = call('[',dat.quote, EmptySymbol(),com )
  } else {
    ca = call('[',dat.quote, filter.call, com)
  }
  if (!is.null(by)) {
    ca$by = by
  }
  
  ca
} 

# A verb to replace data
modify = function(.data,...,.filter,.by=NULL, .envir=parent.frame()) {

  .data = substitute(.data)
  args =  eval(substitute(alist(...)))

  data.var = as.character(.data)
  dat = get(data.var,.envir)

  if (!is.data.table(dat)) {
    dt = as.data.table(dat)
  } else {
    dt = dat
  }

  if (missing(.filter)) {
    filter.call=NULL
  } else {
    filter.call=substitute(.filter)
  }
  ca = get.data.table.modify.call(args=args, by=.by, filter.call=filter.call)
  
  eval(ca)
  if (!is.data.table(dat)) {
    if (is.tbl(dat)) {
      dat = as.tbl(dt)
    } else {
      dat = as(dt, class(dat))
    }
    assign(data.var, dat, .envir)
  }
  invisible(dat)
}

examples.modify = function() {
  library(dplyr)
  library(data.table)

  n = 1e6
  df = data.frame(a= sample(1:5,n,replace=TRUE),
                   b= sample(1:100,n,replace=TRUE),
                   x=rnorm(n))
  dt = as.data.table(df)
  tbl = as.tbl(df)  
  #modify(tbl, x = x+100,.filter=a==2)
  microbenchmark(times = 5L,
    modify(tbl, x = x+100,.filter=a==2),
    modify(df, x = x+100,.filter=a==2),
    modify(dt, x = x+100,.filter=a==2),
    dt[a==2,x:=x+100],
    mutate.df = mutate(df, x=ifelse(a==2,x+1,x)),
    mutate.tbl = mutate(tbl, x=ifelse(a==2,x+1,x))
  )
## Results  
## Unit: milliseconds
##                                       expr       min        lq    median        uq       max neval
## modify(tbl, x = x + 100, .filter = a == 2)  54.96051  56.38434  61.31385  61.50925  67.71102     5
##  modify(df, x = x + 100, .filter = a == 2)  65.90863  85.99066  86.04277  86.13829  91.89006     5
##  modify(dt, x = x + 100, .filter = a == 2)  50.61876  56.41078  61.40109  61.45043  61.79544     5
##               dt[a == 2, `:=`(x, x + 100)]  50.85836  56.50513  60.78332  61.05451  67.58312     5
##                                  mutate.df 794.84943 821.20072 827.20788 837.43679 849.53164     5
##                                 mutate.tbl 788.86398 802.19081 835.90954 843.26593 899.45605     5
}