zkamvar
commented
9 months ago Hi @trhallam,
This is coming from {pegboard} and it's failing during the step that attempts to validate the lesson while gathering child documents.
Situation
The following R Markdown context cannot be parsed by {pegboard} without evaluating any of the code.
```{r load_config, include=FALSE}
snippets <- paste('files/snippets/', 'TEST1', sep='')Working include:
Not-working include:
One of the things to know about pegboard is that it parses the structure of the document, but it does not evaluate any of the code, which is why the error reads:
Error in initialize(...) :
the file '/home/runner/work/workbench-mwe/workbench-mwe/paste(snippets, /test-include.Rmd, sep = )' does not exist
Calls:
That being said, even if we _did_ evaluate the code supplied to the child parameter, we would still find ourselves in the same situation because the `snippets` object is defined in a separate chunk, which {pegboard} does not know about because it does not evaluate code inside the chunks:
``` r
tmp <- tempfile()
writeLines("```{r snip, child=file.path(snippet, 'test.txt')}\n```", tmp)
ep <- pegboard::Episode$new(tmp)
# this is the document we are working with ----------
ep$show()
#> ```{r snip, child=file.path(snippet, "test.txt")}
#> ```
# grab the child document expression ----------------
print(child <- xml2::xml_attr(ep$code, "child"))
#> [1] "file.path(snippet, \"test.txt\")"
# if we parse and evaluate it, it will fail ---------
print(parse(text = child))
#> expression(file.path(snippet, "test.txt"))
eval(parse(text = child))
#> Error in eval(parse(text = child)): object 'snippet' not foundCreated on 2023-10-03 with reprex v2.0.2
Potential solutions
IIRC, you are looking to find a way to address https://github.com/carpentries/sandpaper/issues/81, which is important for the HPC group. I'm thinking there may be two ways of going about it, but knowing that we need to engineer the solution and open the pathways for contribution in {sandpaper} or {pegboard} before it's possible. I do want to set expectations right now and say that I am currently working part time on medical leave and may not be able to address this quickly, but I will do my best to present all the information I know.
Option One: parameterize and eval
One thing you could do at the moment, which is a bit wonky, but it will work in any situation is to duplicate snippets and use eval = snippet == <THING>. This way, {pegboard} will be able to easily track the child documents.
```{r setup, include = FALSE}
snippet <- "TEST1"This will only load one snippet file depending on the value of snippet:
The drawback, of course is that the snippets are inherently tied to the parent documents and are a bit difficult to wrangle and add content
### Option Two: Parameterized reports
This is much closer to the spirit of https://github.com/carpentries/sandpaper/issues/81 where you could specify a file called `params.yaml`
```yaml
snippet: "TEST1"and then in your R Markdown, you would be able to use this pattern without the need for the setup chunk.
```{r, child=file.path(params$snippets, 'test-include.Rmd'), eval=TRUE}
The caveat to this approach is that it will take a bit of engineering inside of both {sandpaper} and {pegboard} to make the params available since each R Markdown document is considered to run independently:
``` r
tmp <- tempfile()
writeLines("```{r snip, child=file.path(params$snippet, 'test.txt')}\n```", tmp)
ep <- pegboard::Episode$new(tmp)
ep$show()
#> ```{r snip, child=file.path(params$snippet, "test.txt")}
#> ```
# If we assume that the params are available, the child document will evaluate correctly
params <- list(snippet = "files/TEST1")
# grab the child document expression ----------------
print(child <- xml2::xml_attr(ep$code, "child"))
#> [1] "file.path(snippet, \"test.txt\")"
eval(parse(text = child))
#> [1] "files/TEST1/test.txt"Created on 2023-10-03 with reprex v2.0.2
I believe this is the correct way to go because it means that you can continue to use the same pattern you have been using (you can also replace the {{ site.param }} syntax with `r params$param`). The only thing I would be cautious about would be allowing the fields in params.yaml to evaluate as R code (which is happily avoided).
Setting Expectations
I want to reiterate, with several factors going on, adding this feature is going to take some time to implement. I will do my best to coach @carpentries/workbench-maintainers to tackle this, but please be patient.
trhallam
froggleston
Hi,
I'm having some issues with a more recent version of sandpaper where the deploy step is failing in Github.
Local builds of the site build without issue.
The problem:
I use r eval statemets within Rmd documents to dynamically include content depending on a configuration for the lesson I want to publish. This was working previously but there is now an issue I think in the way the deploy function finds and copies content. There don't appear to be any logs or useful output in the Github actions build process because the R output is limited in the deploy step.
I have created a MWE that demonstrates the failing build process.
https://github.com/trhallam/workbench-mwe/actions/runs/6392622732/job/17350256259