Blog post for tibble 2.0.1

[krlmlr]

I'm planning to add links to documentation for the funcitons mentioned in the text, unless there's some pkgdown magic which I could use.

[krlmlr]

@brodieG: Enabling colored HTML output using fansi::set_knit_hooks(knitr::knit_hooks, which = c("output", "error")) cripples some (but not all) tibbles in the output, see the preview at https://deploy-preview-266--tidyverse-org.netlify.com/articles/2019/01/tibble-2.0.0/. What am I doing wrong? Also, bold-faced ellipses are turned to a … character in some cases.

[brodieG]

I can look at it more later, but at a glance looks like your chunk options are not right. For example, you need results="asis" and comment="" when you set the hooks. If you can, try to match what vignette('sgr-in-rmd', package='fansi') does exactly. In particular this. If that doesn't work let me know and I'll try to debug, but probably can't get to it till this evening US East Coast time.

[krlmlr]

Thanks. This does change the output, but unfortunately doesn't help here:

[brodieG]

I tried just building the document with rmarkdown::render and it worked for me:

I'm guessing its a CSS conflict of some sort, unless maybe you don't have the latest version of fansi. I'll have to download and build your whole site so I can see why it works as a standalone RMD file for me, but not for you. The example above is a little modified to test the colors are actually being put in.

One other thing that would be helpful: can you report what crayon::num_colors() produces for you?

[krlmlr]

crayon reports 256 colors, I have fansi 0.4.0 on my system.

The site is rendered using blogdown::serve_site(), you may need to touch the .Rmarkdown file to initiate the process.

I don't understand why the problem happens with some but not all tibbles in the output. I'll turn off colors as a stop gap, perhaps we can find a solution and update the blog post later.

[krlmlr]

What's the timeline for merging and communicating this?

There's no need to rush, binaries for some platforms haven't been built yet. I'm planning a 2.0.1 release fairly soon, but I'd like to wait before all binaries are built.

[krlmlr]

OTOH, I think the binaries for R-release and R-devel will never be built due to the incompatibility with rlang 0.3.1. Let's wait for tibble 2.0.1 then.

[brodieG]

Rename your file *.Rmd, not *.Rmarkdown. I believe the latter triggers a blackfriday or some such markdown->html translation, which #?##! gratuitously substitutes HTML line breaks for new lines. If you use Rmd I think you get the normal pandoc conversion directly to HTML.

I am not 100% certain this is what's happening because debugging the blogdown -> knitr -> rmarkdown -> pandoc chain (or any permutations of them) is about as pleasant and easy as pulling my nails out with wooden spoons and since the change was enough to make it work for me I do not want to spend one additional second confirming the exact nature of the failure. However, this seems very likely because using Rmarkdown files leave .markdown artifacts, whereas using Rmd files leave html artifacts.

Please let me know if the change does not fix it for you. I really really hope it does since I was both able to reproduce the problem and fix it by making the change.

[hadley]

@krlmlr there's no rush, but at the same time, I don't think including coloured output is that important if it requires a very different setup to the rest of the blog.

@brodieG ah bummer — we deliberately use .Rmarkdown because if we use .Rmd we don't get an internal table of contents. Thanks for looking into it, and I 100% empathise with the pain of testing when you have this complicated combination of tools.

[brodieG]

I don't have the courage to mess with blackfriday, but it does look like there is an option that might be causing the problem:

What's concerning though is that it is off by default, so if it is indeed this it was presumably turned on for some reason.

That aside, it is possible to have TOCs in Rmd. If the positioning of the existing TOC in the sidebar is CSS, then it should be possible to do the same with the Rmd generated one. If not looks like Garrick has examined the problem.

But I agree with @hadley that this is not worth doing, especially when the colors in this case are so subtle.

[krlmlr]

I tweaked a bit for 2.0.1, should be ready to go.

Thanks again Brodie for looking into it, sorry for the pain. In the long run I'd love to see color in console output. Would it work if fansi::set_knit_hooks() generated HTML without newlines? A manual TOC or setting blackfriday options also seem worth investigating.

[krlmlr]

Coloring works now, examples look good to me.