knitr
has a few pretty straightforward ways of handling this.
Option 1: Using knit_child()
with inline R code
Say your setup is like the following. In the same directory, you have:
graph.R
## ---- graph
library(ggplot2)
CarPlot <- ggplot() +
stat_summary(data= mtcars,
aes(x = factor(gear),
y = mpg
),
fun.y = "mean",
geom = "bar"
)
CarPlot
chapter1.Rnw
Hey, look, a graph (Figure~\ref{fig:graph})!
<<graph, echo=FALSE, message=FALSE, fig.lp='fig:', out.width='.5\\linewidth', fig.align='center', fig.cap="A graph", fig.pos='h!'>>=
@
main.Rnw
\documentclass{article}
\begin{document}
<<external-code, echo=FALSE, cache=FALSE>>=
read_chunk('./graph.R')
@
\Sexpr{knit_child('chapter1.Rnw')}
\end{document}
Then, you can knit
the main.Rnw
file and compile the resulting .tex
file with either pdflatex
or xelatex
.
The output is:

Note that you can also read the external .R
file from the child .Rnw
file.
So, the following would have worked just as well.
chapter1-mod.Rnw
<<external-code, echo=FALSE, cache=FALSE>>=
read_chunk('./graph.R')
@
Hey, look, a graph (Figure~\ref{fig:graph})!
<<graph, echo=FALSE, message=FALSE, fig.lp='fig:', out.width='.5\\linewidth', fig.align='center', fig.cap="A graph", fig.pos='h!'>>=
@
main-mod.Rnw
\documentclass{article}
\begin{document}
\Sexpr{knit_child('chapter1-mod.Rnw')}
\end{document}
Option 2: Using chunk option child
Assuming you have graph.R
and chapter1.Rnw
from above in the same directory, then your main.Rnw
should be:
\documentclass{article}
\begin{document}
<<external-code, echo=FALSE, cache=FALSE>>=
read_chunk('./graph.R')
@
<<child-demo, child='chapter1.Rnw'>>=
@
\end{document}
Note that you can also read the external .R
file from within the child document in this case, too.
So, assuming you had graph.R
and chapter1-mod.Rnw
from above in the same directory, then your main-mod.Rnw
file should be:
\documentclass{article}
\begin{document}
<<child-demo, child='chapter1-mod.Rnw'>>=
@
\end{document}
Best Answer
Personally, I like emacs+ess+auctex. I think it hits all your points, but unfortunately it has an epic learning curve.