Speedup predict.dummyVars

[khotilov]

predict.dummyVars was very slow with large datasets with many factors due to the colnames(x) <- assignments within nested loops. Here's an illustrative example:

#create a local version of predict.dummyVars that is fixed to be the same as in this PR
predict_dummyVars <- caret:::predict.dummyVars
fix(predict_dummyVars)

n <- 400000
p <- 100
x <- data.frame(rep(list(x=rep(c('a','b'), n/2)), p/4)) # nominal
x <- cbind(x, data.frame(rep(list(x=rep(c('c','d','e','f'), n/4)), p/4)))
x <- cbind(x, data.frame(rep(list(x=rep(1, n)), p/2)))  # some numeric
colnames(x) <- paste0('x', 1:p)
dim(x)

dumm <- dummyVars("~ .", data = x, fullRank = T)

# before
ptm <- proc.time()
x1 <- predict(dumm, x)
print(proc.time() - ptm)

# after
ptm <- proc.time()
x2 <- predict_dummyVars(dumm, x)
print(proc.time() - ptm)

all.equal(x1,x2)

that results in:

> # before
   user  system elapsed 
  33.60   44.07   77.86 
> # after
   user  system elapsed 
   2.05    0.78    2.81 
> all.equal(x1,x2)
[1] TRUE

[codecov-io]

Codecov Report

Merging #727 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #727      +/-   ##
==========================================
- Coverage   16.97%   16.97%   -0.01%     
==========================================
  Files          90       90              
  Lines       13183    13185       +2     
==========================================
  Hits         2238     2238              
- Misses      10945    10947       +2

Impacted Files	Coverage Δ
R/dummyVar.R	0% <0%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f27c68b...69269f8. Read the comment docs.

[topepo]

Looks good. Thanks