Profiling in the tidyverse

[wlandau]

I find it difficult to read profvis output when I use tidyverse functions. Is there a way to hide superfluous parts of the call stack?

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

# dplyr
profvis::profvis({
  d <- data.frame(x = rnorm(1e7))
  d <- mutate(d, y = rnorm(1e7))
  d <- mutate(d, z = rnorm(1e7))
})

# magrittr
profvis::profvis({
  data.frame(x = rnorm(1e7)) %>%
    cbind(y = rnorm(1e7)) %>%
    cbind(z = rnorm(1e7))
})

# base
profvis::profvis({
  d <- data.frame(x = rnorm(1e7))
  d$y <- rnorm(1e7)
  d$z <- rnorm(1e7)
})

^{Created on 2018-11-04 by the reprex package (v0.2.1)}

[wch]

cc @hadley @jimhester @lionel-

[hadley]

The dplyr issue should be resolved in the dev version; the magrittr version might one day be resolved if we rework how the pipe works.

[sanjmeh]

Gentlemen, was there any movement on this issue? There is a similar issue on stack overflow where profvis() is not able to remove internal calls from shinyapp profiling as well.
https://stackoverflow.com/questions/57776408/how-to-hide-internal-function-calls-in-profvis-flame-graph-running-on-a-shinya
I am presuming these two issues are related.
Looking forward to a solution.

[lionel-]

The patch is pending review https://bugs.r-project.org/bugzilla/show_bug.cgi?id=17595