Non default plot size messes with "legend"

[knokknok]

I would expect the plots produced by the 2 samples below to be identical but they are not:

library(knitr)
knit_patterns$set(all_patterns$rnw)
opts_knit$set(out.format = 'latex')
render_latex()
knit(output = 'legend.tex', text = '
<<test, fig.width=12, fig.height=12>>=
x <- 0:64/64
matplot(x, outer(x, 1:7, function(x, k) sin(k * pi * x)), type = "o", col = 1:7, ylim = c(-1, 1.5), pch = "*")
legend("topleft", legend=paste("sin(", 1:7, "pi * x)"), col=1:7, lty=1:7, pch = "*", cex = 0.8)
@
')

pdf(width=12, height=12)
x <- 0:64/64
matplot(x, outer(x, 1:7, function(x, k) sin(k * pi * x)), type = "o", col = 1:7, ylim = c(-1, 1.5), pch = "*")
legend("topleft", legend=paste("sin(", 1:7, "pi * x)"), col=1:7, lty=1:7, pch = "*", cex = 0.8)
dev.off()

Notice that the legend spacing is incorrect (the result from knitr looks like what happens when you manually resize your device after plotting).

[yihui]

should be fixed now; you must be using an interactive R session, which is not good unless you have special reasons; normally you should call knitr in a non-interactive R session via R -e 'library(knitr);...' or Rscript -e ...

[knokknok]

It is fixed, thanks!

I do use knitr interactively because I like being able to play with my objects and try things out before actually putting code in my Rnw.

Why is that bad?

[yihui]

because finally you will have to run it non-interactively

when you run it interactively, the objects are generated in the global environment, and you may screw up some randomly without noticing it (e.g. you created a function in the console but forgot to do it in the source document)

if you want to do exploratory work, you can use RStudio to run code chunks; it is pretty convenient. When it comes to compile the whole document, just hit the button to compile it in a separate R session to guarantee everything works fine

[knokknok]

Ok, thanks. I'm pretty entrenched in emacs+terminal though ;-)

[github-actions]

This old thread has been automatically locked. If you think you have found something related to this, please open a new issue by following the issue guide (https://yihui.org/issue/), and link to this old issue if necessary.