R和Knitr / HTML - 处理需要很长时间才能运行的脚本

Question

I'm documenting some code that will be read by other students, and knitr seems a good way to do that. One point that is bothering me is that, for scripts that require a long time to run, my approach is not efficient.

Suppose I have something like

<!--begin.rcode example1, fig.width=8, fig.height=10
input <- data[,c("K12","K23","delta")]
output <- data[,"Class"]
startTime <- proc.time()
result <- C5.0(input, output)
totalTime <- proc.time()-startTime
cat("Execution time: ", totalTime[3], "\n")
plot(result)
result
end.rcode-->

This creates, plots and prints a decision tree based on a data frame called data. I would like to have several chunks like this in a single .Rhtml document, and in each chunk I'd change the data set or the parameters of the algorithm.

If the data set is large, the call to C5.0 will take some time. If I add some other examples to the same .Rhtml file, I'd have to rerun it to create the .html file and figures. I've been doing this repeatedly, since I want to comment on the results of the execution, and in order to create the .html file I need to knitr the .Rhtml file again, which means rerunning the code.

What I am looking for is a way to either tell knitr that it can reuse the previous results of executing a chunk (don't see a way to do that, besides, seems risky and manual) or break the .Rhtml in pieces that can be knitr'd separately but still make a whole .html when I need to do it -- even better, something like a "make" for knitr that would rerun only the changed .Rhtmls and create a single .html file. By the way, if anyone knows how to knitr a .Rhtml file from the command line that could also be useful -- I'm using RStudio for convenience, but a single command line command would also be helpful.

I know this seems subjective, but I am not looking for a better (subjectively, "my approach is better than yours") way to do a task -- any approach that works without the need to rerun the whole .Rhtml page will do.

thanks

Answer 1

One very easy way of combining long running code with knitr is to use the cache option. To use caching, just add cache=TRUE .

My personal experience is that every so often you have to delete the directories that cache makes and do a clean build.