- It is an R file that can specify tables, graphs, or statistical models.
- The code behind file can be viewed as a companion to the report file(s). We like to save the code behind file and the report file(s) in the same directory.
- It's the last file in the workflow before a
knitrreport file (typically an Rnw or Rmd). - Each logical section of code is separated into a 'chunk' and labelled so that the knitr code can hook into the right spots.
- A single code behind file can be called by multiple reports. For instance, an Rmd (ie, Markdown-based knitr file that can be rendered as HTML) can hook into the code behind file for a quick & dirty internal report that's intended just for your research team, while an Rnw (ie, a LaTeX-based knitr file that can be rendered as a PDF) can hook into the same code behind file for a polished external report that's intended for a wider audience.
- It's not always possible, but try to have the code behind file as device agnostic as possible. For graphs, this means that the code behind will specify a graph, but not actually draw it, or even declare its dimensions. The graphic format (eg, png, svg, pdf) should be determined by the Rmd or Rnw file.
- The file should be able to run independently of the report file. There are two advantages:
- The lack of dependency makes the file's relationships conceptually cleaner.
- Since the file can execute by itself, it's easier to develop and debug. The file can be run in it's entirety (Ctrl+Alt+R in RStudio), or several lines at a time (highlight the desired lines and Ctrl+Return in RStudio).
- Altough a report can execute the code behind's chunks in any order, we recommend running them in the same order as they appear in the code behind, because:
- It's easier understand and spot mistakes (such as inadvertantly skipping a section, or confusing the order of several sections).
- You can execute the code-behind file in its entirety (eg, Ctrl+Alt+R). This closely mimics how the report will execute, which makes development easier.