Alternative lambda expression mechanism

[renkun-ken]

do.call(Map,c(function(...) list(...), lst, .i=list(seq_along(lst)), .name=list(names(lst))))

An alternative method of constructing lambda expression symbols is to construct the argument list for Map in the first place instead of constructing environment in each iteration.

[renkun-ken]

list.map2 <- function(.data,expr) {
  expr <- substitute(expr)
  l <- lambda(expr)
  envir <- new.env(FALSE,parent.frame())
  args <- c(function(...) {
    args <- list(...)
    list2env(args,envir)
    eval(l$expr,args$.,envir)
  },list(.data),list(seq_along(.data)),list(names(.data)))
  names(args) <- c("f",l$symbols)
  do.call(Map, args)
}

Above is an example of the method but benchmark does not show significant performance improvement.

> microbenchmark::microbenchmark(a=list.map(devs,age+.i),b=list.map2(devs,age+.i),times = 10000)
Unit: microseconds
 expr    min     lq median     uq      max neval
    a 46.390 50.085 51.727 53.370  1071.07 10000
    b 42.285 45.980 47.621 49.264 41228.14 10000

[renkun-ken]

> microbenchmark::microbenchmark(a=list.map(users,age+.i),b=list.map2(users,age+.i),times = 10)
Unit: milliseconds
 expr      min       lq   median       uq      max neval
    a 515.8601 559.6479 568.8228 577.3335 611.1610    10
    b 425.9984 451.1198 460.6385 481.0625 669.6938    10

where users is a list of 50,000 records. The memory usage is very efficient.

The performance seems to improve 20%.