"Compiling" non-R chunks in Positron quarto docs

PeterG · November 20, 2025, 6:24am

Hi everyone,

I'm testing out Positron for the first time and looking to streamline my workflow. Specifically, I am trying to create quarto documents that combine R code with Stan models.

The cmdstanr::register_knitr_engine function makes it possible to define Stan-specific chunks that get compiled into a object that is available to all subsequent R chunks. So, I can write something like:

```{stan, output.var="model"}
stan_stuff { ... }
```

```{r}
model$sample()
```

This works great when I compile the full quarto documents I'm building.

Where I'm hitting friction is in interactive editing. When I'm working exclusively with R in Positron, I can trigger a "Run Above" command that will execute all chunks above the current one, making all objects assigned available to the current session.

Obviously, this doesn't happen with Stan chunks, but that causes my code to fail because model isn't available.

Has anyone found a good solution? I can create the session object manually in the console, but that's tedious and prone to error. Is there a better way? Is it possible, perhaps, to tell Positron how to "run" these blocks?

Thanks in advance for any advice.

julia · November 23, 2025, 11:41pm

This is an interesting point; thank you for raising it! It looks somewhat to this issue about params:

github.com/posit-dev/positron

RMD Documents with params set do not create the `params` object when run interactively

opened 07:40PM - 13 Aug 24 UTC

csnydermph

enhancement area: quarto lang: r support

## System details: #### Positron and OS details: Positron Version: 2024.08….0 (system setup) build 31 Code - OSS Version: 1.91.0 Commit: 7ee0596b81d578059e63c9af7c535c6b6040b947 Date: 2024-08-10T13:58:22.910Z Electron: 29.4.0 Chromium: 122.0.6261.156 Node.js: 20.9.0 V8: 12.2.281.27-electron.0 OS: Windows_NT x64 10.0.19045 #### Interpreter details: R 4.3.3 ## Describe the issue: In RStudio, using ctrl+enter or running a code chunk inside of an RMD file would populate the `params` R object with the default params set in the YAML.. In Positron, the `params` object is not created when running code or code chunks interactively from the RMD file. This issue prevents manually running / testing parametric reports; the whole report must be rendered with defined params each time. ## Steps to reproduce the issue: 1. Create RMD file and define `params` in the YAML 2. Create an R code chunk that references the `params` object, and manually run the code chunk 3. ` Error: ! object 'params' not found` ## Expected or desired behavior: The desired behavior is for Positron to populate the `params` object for RMD files as RStudio, so RMD files can be tested / iterated on when they are parameterized. ## Were there any error messages in the UI, Output panel, or Developer Tools console? ``` Error: ! object 'params' not found ```

Do you know how cmdstanr::register_knitr_engine works under the hood? I am not very familiar with the details there, but we would be very happy to work with the cmdstanr folks to find a good path forward.

PeterG · November 24, 2025, 6:38am

Thanks for your reply! The issue you link is definitely related conceptually (differences between interactive and build modes).

As far as I understand, cmdstanr::register_knitr_engine just passes a cmdstanr-specific processing command over to knitr. This is a core capacity of knitr itself, which defines different engines for different types of chunks.

> names(knitr::knit_engines$get())
 [1] "awk"       "bash"      "coffee"    "gawk"      "groovy"    "haskell"   "lein"      "mysql"     "node"     
[10] "octave"    "perl"      "php"       "psql"      "Rscript"   "ruby"      "sas"       "scala"     "sed"      
[19] "sh"        "stata"     "zsh"       "asis"      "asy"       "block"     "block2"    "bslib"     "c"        
[28] "cat"       "cc"        "comment"   "css"       "ditaa"     "dot"       "embed"     "eviews"    "exec"     
[37] "fortran"   "fortran95" "go"        "highlight" "js"        "julia"     "python"    "R"         "Rcpp"     
[46] "sass"      "scss"      "sql"       "stan"      "targets"   "tikz"      "verbatim"  "glue"      "glue_sql" 
[55] "gluesql"

What I don't fully understand is how block handling gets interpreted during interactive editing. For example, in something like this:

...is the |> Run Cell | Run Next Cell | Run Above bit getting added by positron or by the Quarto vscode plugin? Does knitr get involved when these buttons are pushed, or is it all handled in parallel?