make styler faster

[lorenzwalthert]

The following could make styler run faster:

• dropping unnecessary select() and arrange() statements.
• early termination (e.g. rules?)
• consider replacing some tidyverse calls with base R if substitution
  does not make the code "uggly"? What do you think @krlmlr ?
• which statements related to needs_indention()

microbenchmark::microbenchmark(dplyr = dplyr::mutate(mtcars, driver = 3), 
                               baseR = mtcars$driver <- 3)
#> Unit: microseconds
#>   expr      min       lq      mean   median        uq        max neval cld
#>  dplyr 1969.341 2009.056 3740.7995 2037.379 2091.4205 152117.420   100   b
#>  baseR   11.663   13.735   24.1628   20.456   22.9275    433.822   100  a

microbenchmark::microbenchmark(dplyr = dplyr::slice(mtcars, 1:3), 
                               baseR = mtcars[1:3, ])
#> Unit: microseconds
#>   expr      min       lq      mean    median       uq      max neval cld
#>  dplyr 2094.290 2148.480 2446.0150 2191.8545 2271.761 8700.220   100   b
#>  baseR  138.072  155.203  176.9532  165.3705  189.216  431.295   100  a

# pipe does probably not slow down things much.
library(magrittr)
microbenchmark::microbenchmark(dplyr = mtcars %>% summary(), 
                               baseR = summary(mtcars))
#> Unit: milliseconds
#>   expr      min       lq     mean   median       uq      max neval cld
#>  dplyr 4.550347 4.619500 5.089398 4.695070 4.833018 9.004815   100   a
#>  baseR 4.390407 4.497096 4.924098 4.548395 4.833101 8.284952   100   a

These are just some thoughts, profiling of our code is probably a better place to start anyways...

[krlmlr]

Let's identify our bottlenecks first, you could try profvis.

[lorenzwalthert]

With #90, I think we can close this, at least for now.