Vignettes use absolute paths

[zkamvar]

After updating, I notice that the vignettes/articles are using absolute paths to images instead of relative paths.

case in point: grunwaldlab/poppr@5697ce1#diff-f627caa64acaddf31682a506ff09110dR353

[batpigandme]

Hmm, there's something inconsistent happening, it seems, because the logo in your article below is definitely relative.
https://grunwaldlab.github.io/poppr/articles/mlg.html

[zkamvar]

Ignore the one hosted at grunwaldlab.github.io. That was built with the previous iteration of pkgdown.

This is the one you want:

https://htmlpreview.github.io/?https://github.com/grunwaldlab/poppr/blob/5697ce13af6a0931c82cbc739e4d8c6135fd1844/docs/articles/mlg.html

[batpigandme]

Does the view appear the same for you locally (i.e. without htmlpreview)?

[zkamvar]

Locally, I will get the correct images showing up, but I suspect if I change computers, this will not be true.

[hadley]

I changed how vignettes are built and didn’t have an example of embedding a plot. Oops!

I’ll reconsider my approach again. Might be able to fix by setting working directory instead of using output dir, or could try building in vignette dir with temp name and then copying to articles.

[hadley]

Setting the working directory doesn't seem to help. Now trying the copying approach.

[hadley]

Actually, I can't reproduce the problem. I added a new test vignette containing a figure, but I always get a relative path. How are you building the site?

[hadley]

And can you please provide a pointer to the source of the vignette?

[jayhesselberth]

I still see abs paths in README and vignettes, see #554

I get the same result calling pkgdown::build_site() or using RStudio menu item.

[hadley]

Oh the problem is images, not figures.

[hadley]

Hmmm setting output_dir to anything (even .) appears to trigger absolute urls in the output.

[hadley]

And we can't set output_file because that affects the path used for R figures.

I think that implies that we have to build in the directory in which the article lives. We can't even copy that directory because we also have to copy all children, and in the case of README.Rmd that's the whole package.

To figure out which files we need to copy, maybe we can dir_ls() before and after, and then move all new files?

[hadley]

Hmmm, that strategy will fail if you're previewing locally.

Another idea: copy .Rmd to output directory, build in that directory, then delete the .Rmd. Would also need to copy files found by rmarkdown::find_external_resources(), and document that if a neede file doesn't get imported, you'll need to add it to resource_files in the yaml header.

[hadley]

Alternatively, since the problem is only images, maybe it's easier to simply add another layer of html tweaking?

[hadley]

Ok, I've resolved the absolute path problem but it reveals a new problem - images in the vignettes directory are not currently copied to the documentation website. Should they be? R does not (unless you list in .install_extras). But this behaviour was probably designed in the era of pdf vignettes where there was no need to copy inputs. It seems like it would be more convenient if pkgdown copied these files; but precisely what should it copy?

Maybe a reasonable heuristic is to copy any "image" files found by rmarkdown::find_external_resources()? That should get the files you need, without copying over
potentially large files that are needed to generate the article, but not to view it.

...

Oh better, rmarkdown::find_external_resources include a field "web" indicating if the file is needed for rendering on the web.