[help] Inconsistent rendering of tar_mermaid() in targets quarto markdown

[dcaseykc]

Help

• I understand and agree to https://books.ropensci.org/targets/help.html.

Description

I have a targets pipeline specified in quarto document (https://github.com/PHSKC-APDE/hyrule/blob/mermaid/example_workflow/_targets.qmd). I want to render the DAG via tar_mermaid so it works with github markdown.

tar_mermaid(targets_only = F) seems to work, but tar_mermaid(T) does not (i.e. this document is blank).

When executed locally/interactively, tar_mermaid(T) behaves just fine.

I'm trying to make a more simple reprex that reproduces the issue, but so far I haven't been able to figure out why my specific case is being wonky.

To recreate my specific issue (at least on my two machines):

1. Open https://github.com/PHSKC-APDE/hyrule/blob/mermaid/example_workflow/_targets.qmd in Rstudio
2. Download/install the packages that are required to run the pipeline (install.packages(c('remotes', 'tarchetypes', 'targets', 'glue', 'data.table'))
3. Knit/render the document
4. Check tm_true.txt and confirm that it is empty. A mermaid chart should render, but it's the targets_only = FALSE one.
5. Change ~line 114 from 'tm_false.txt' to 'tm_true.txt' and rerender. The place where the memaid chart should be will show an error.

[wlandau]

In the "ML Record Linkage Pipeline" section, if I see:

print(tar_network(targets_only = TRUE)$vertices$type)
#> character(0)

print(tar_network(targets_only = FALSE)$vertices$type)
#>  [1] "function" "function" "function" "function" "function" "function"
#>  [7] "function" "object"   "function" "function" "object"   "function"
#> [13] "object"   "function" "function" "function" "function" "function"
#> [19] "function" "function" "function" "function" "function" "function"
#> [25] "function" "function" "function" "function" "function"

The pipeline has no targets in it, only imported objects and functions. So it makes sense that the first graph is empty.

[dcaseykc]

And the pipeline has nothing in it because the tar_unscript call clears everything out. And that would explain why I couldn't recreate it -- because my test pipeline was calling tar_mermaid at the end of the document. I feel very dense.

Anyways -- thank you very much for the help!