Preserving folded blocks when rewriting YAML

[arnaudgallou]

I'm developing a package that can insert author metadata in YAML header of Quarto files. I do this by rewriting YAML blocks using the following steps:

# extract YAML from file
yaml.load()
# add new data
as.yaml()
# insert YAML back into file

The problem is that some information are lost in the process, including folded blocks:

lines <- "
  bar: >
    Lorem ipsum
    Vivamus quis
"

yaml::yaml.load(lines) |> 
  yaml::as.yaml() |> 
  cat()
#> bar: |
#>   Lorem ipsum Vivamus quis

My question is: is there any way to preserve folded blocks in that case? I'm not aware of a special type I can use to target the > with a handler.

I'm considering to insert author data in a separate YAML header when my insert function detects complex cases (e.g. the use of custom tags, repeated blocks, comments being dropped due to the libyaml dependency, etc.). I would prefer to do that on rare occasions though.

[spgarbet]

This would require preserving meta-information about the storage format in the converted objects. R could do this utilizing attributes. However, this library utilizes the libyaml library. This meta-information is stripped before it's tokenized. Without diving into the next library down it's not possible.

> lines <- "
+   bar: >
+     Lorem ipsum
+     Vivamus quis
+ "
> 
> invisible(yaml::yaml.load(lines, handlers=list(str=function(x) {print(x); x})))
[1] "bar"
[1] "Lorem ipsum Vivamus quis\n"

The str handler is what passed from the yaml tokenizer. If I get down into that level I'll be finishing yaml 1.2 compliance.