The profvis package is a powerful tool for line profiling in R.
This package is provided by the RStudio team, and its most appreciated feature is the interactive report that is automatically produced, representing a really effective way of visualizing and investigating time resources requested by each part of your code.
Since the lineprof package is not hosted on CRAN, but on GitHub, we first need it to install the devtools package in order to leverage the install_github function provided by this package.
Moreover, we will use the ggmap package to build the example to be profiled:
install.packages(c("devtools","ggmap"))
library(devtools)
install_github("rstudio/profvis")
library(profvis)
library(ggmap)- Define a
profvisobject containing the code to be profiled. Run the following piece of code, initializing thereportobject. - The following code is used from the drawing a route on a map with
ggmaprecipe in Chapter 3, Basic Visualization Techniques; refer to the recipe for further details:report <- profvis ({ library(ggmap) trip <- (route(from = "rome", to = "milan",structure = "route", output = "simple")) segment <- c() for(i in 1:nrow(trip)) { if(i == 1){segment[i] <- 1}else { if( i %% 2 != 0 ) { segment [i] <- i-segment[i-1]}else { segment [i] <- i/2 } } } } segment_couple <- c(0,segment[-length(segment)]) trip$segment <- segment trip$segment_couple <- segment_couple route_map <- get_map("italy",zoom = 6) ggmap(route_map) + geom_point(aes(x = lon, y = lat, size = 5, colour = hours),data = trip) + geom_line(data = trip,aes(group = segment)) + geom_line(data = trip, aes(group = segment_couple)) })Print the object runing print(report)This will open a web browser window, showing an interactive profiling report similar to the following:
In the preceding screenshot, on the left-hand side, you can find your code produced line by line, where each line has, on the right-hand side, a count of the required milliseconds for running and a count of the proportion of the time on the total running time. On the right-hand side part of the report, a barchart-like interactive plot is produced, where the bottom line represents the first function that was called and the upper bars represent the functions called by those first functions. You can hide lines of code having zero running time. On the top-right corner of the left-hand side part of your report, you can find a Settings control. Selecting it will open the following menu:
Selecting the checkbox next to Hide lines of code with zero time will result in hiding of all the lines of code that have a running time less than a millisecond:
- Focus on a single line of code. Selecting a single piece of plot on the right-hand side of the window will let you zoom in on that piece and all the bars that come from that first bar. Save the report as a web page (use the cmd + S or Ctrl + S shortcut).
Running cmd + S on Mac devices, or Ctrl + S on Windows devices, will let you save your report as a fully functional web page:
