Knitr and ReportingTools

Jessica L. Larson and Gabriel Becker

We can use knitr and ReportingTools together, as shown in this very simple example from our basic vignette.

Here we will create a simple report that integrates ReportingTools report capabilities with knitr.

To build this report, simply call the knit2html function from the knitr package.

By default, knitr will write the destination file within the current working directory.
We thus call knit2html from within the reportDirectory.
Alternatively, we can call knit2html with an output argument that matches the reportDirectory argument passed to HTMLReport.

Note: the output html file must be in the directory indicated by the location argument. If this is not the case the ReportingTools functionality will break. By default, knitr will write the destination file within the current working directory, and ReportingTools will assume that. When calling knit2html with an alternate output argument, it should match the reportDirectory argument passed to HTMLReport. The one exception to this rule is package vignettes, where it is necessary to place the vignettes in the vignettes directory and pass “../inst/doc” as the reportDirectory. If the output lands in the vignettes directory, R would clean it up after building the vignette, and all images, javascript, etc would be lost.

my.df <- data.frame(EGID = c("103", "104", "105", "106", "107"), RPKM = c(4, 
    5, 3, 100, 75), DE = c("Yes", "Yes", "No", "No", "No"))

library(ReportingTools)

We now can publish our data frame as table through the use of the knitrHandlers. When using knitrHandlers we must specify the destination location as an argument either to the knitrHandlers constructor function directly or as an additional argument to HTMLReport when knitrHandlers is specified in the handlers argument.

We do not call finish since we will still add output to our knitr report.

library(XML)
instDoc <- file.path("..", "inst", "doc")
htmlRep <- HTMLReport(shortName = "knitrReport", handlers = knitrHandlers, reportDirectory = instDoc)

# publish(my.df, htmlRep)
htmlRep[["mydf"]] <- my.df  ##this line is equivalent to the line above

EGID	RPKM	DE
EGID	RPKM	DE
103	4	Yes
104	5	Yes
105	3	No
106	100	No
107	75	No

EGIDRPKMDE

Thus we are able to include tables generated by ReportingTools based directly within knitr documents.

Now we will try to put in an image:

library(lattice)
set.seed(123)
htmlRep[["pic"]] = xyplot(y ~ x, data = data.frame(y = rnorm(25), x = rnorm(25)))

figuresknitrReport/knitrReport-59999.png

## We could also add a picture like this:
y <- rnorm(500)
pic2 <- histogram(y, main = "Sample of 500 observations from a Normal (0,1)")
publish(pic2, htmlRep)

figuresknitrReport/knitrReport-72776.png

To build the report, simply call the knit2html function from the knitr package.

As we have seen, using ReportingTools in conjunction with knitr is as easy as selecting a different set of ReportHandlers (in this case those generated by the knitrHandlers constructor function) and ensuring that the location of the final HTML file matches (up to the directory) that specified by the location field of that ReportHandlers object.

This allows us to create rich knitr-based reports and vignettes while leveraging the power of ReportingTools to format and display generated R objects.