Yesterday I updated blogdown, from v. 0.1 to 0.4. This broke build.site(), because when it tries to compile an Rmd file that calls projecttemplate, the working directory is no longer being set to the project root, which presently is “/Users/kendall/Documents/Github/InvasionHet”. This was acheived by setting the knitr parameter root.dir in .Rprofile in the post directory.
Here’s what the knitr option currently is:
knitr::opts_knit$get()$root.dir[1] "/Users/kendall/Documents/Github/InvasionHet"# Originally gave NULLand here’s what is being viewed as the working directory:
getwd()[1] "/Users/kendall/Documents/Github/InvasionHet"# Originally gave post subdirectorySo it appears that the local .Rprofile is no longer being read in.
When I move the .Rprofile to the project root (and update the relative path to point to that same directory) then it does get read in (see updated values above).
Well, that didn’t fix the problem. The call to ProjectTemplate::load.project() (which is first encountered in content/post/2017-09-19-fit-ler-dispersal-kernels.Rmd) still sees the current directory as post. In fact, it’s not even clear why the chunk was executing, as cache was set to TRUE!
Maybe the cache location has been moved?
What’s even wierder is that, having modified 2017-09-19-fit-ler-dispersal-kernels.Rmd (just turning caching on and off), it re-rendered under serve.site just fine!
I don’t know why things aren’t recompling fine under build_site. But the workaround for creating a publishable version without rebuilding the Rmd files is to use blogdown::hugo_build() instead.
But the problem with this is that the new pages are not being added to the main index!