build_dir() renders Rmd although output is newer due to other file with the same basename

[solarchemist]

I think I might have stumbled over an unexpected quirk of the "Build Website" button in Rstudio server (actually, the build pipeline in general). I think I have nailed down the issue to these particular lines in build_dir():

bases = with_ext(files, '')
files = files[!i & bases == bases[i]]  # files with same basename as f (Rmd)
if (length(files) == 0 || any(older_than(files, f))) render_it()

It just so happens, in my case, that my blog post folders (coming from Jekyll) are named like 2018-06-28-cool-project, with both the Rmd and Rproj file always using the same basename (2018-06-28-cool-project.Rmd, 2018-06-28-cool-project.Rproj). I don't remember where I picked up this convention, but in any case, it succesfully manages to trick the older_than check to always detect an older "output" file, since (in this case) the Rproj file is indeed older and always shares the same basename with the Rmd file.

A simple demonstration:

> files
[1] "/hugo/content/post/2018-06-28-cool-project/2018-06-28-cool-project.html" 
[2] "/hugo/content/post/2018-06-28-cool-project/2018-06-28-cool-project.Rproj"
> f 
"/hugo/content/post/2018-06-28-cool-project/2018-06-28-cool-project.Rmd"
> blogdown:::older_than(files, f)
[1] FALSE  TRUE

Which naturally causes any(older_than(files, f)) to always return TRUE.

This causes much longer rebuilds than necessary (although I must say the Hugo step really is a bullet train compared to Jekyll's horsecart!).

A simple solution for me would be to just change the names of my Rproj files to something other than the basename of my Rmd files. But I thought perhaps this situation might affect others and warrant a different approach?

[yihui]

Should be fixed now. Thanks for the report!