Quarto complains about inexistent duplicate chunk labels when using tar_quarto() within workflow [help]

[chihayakenji]

Help

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

Description

(Also asked on Stack Overflow)
I'm using tar_quarto() to create a report within my targets workflow and either quarto or knitr error saying that there is a duplicate chunk label in my .qmd file. I have only one label, so it is definitely not duplicated. How do I fix this? Is this a bug?

Below is a reproducible example with a simpler project.

I'm using r 4.60, targets 1.12.0, modelsummary 2.6.0, tarchetypes 0.14.1 and quarto 1.5.1 on a an apple computer running macOS Tahoe 26.5.1 with an ARM apple silicon chip for my reproducible example. The original problem occurred in a Remote Desktop server running Windows with the same r and package versions.

The targets file:

# Load packages
library(targets)
library(tarchetypes)
# Target options:
tar_option_set(
  packages = c("tibble", "modelsummary")
)
# Target list
list(
  tar_target(
    name = data,
    command = tibble(x = rnorm(100), y = rnorm(100))
  ),
  tar_target(
    name = model,
    command = lm(y ~ x, data = data)
  ),
  tar_quarto(
    name = report,
    path = "report.qmd",
    quiet = FALSE
  )
)

The .qmd file:

     --- 
     title: "Duplicate chunk label where there is none" 
     format: html
     --- 
     The following creates an error when embedded within a targets project.
     
     ```{r}
     
     library(targets) 
     tar_load_globals() 
     tar_load(model) 
     ```
     
     Table @tbl-modelsummary shows a model summary
     
     ```{r}
     #| label: tbl-modelsummary
     #| tbl-cap: "Model summary" modelsummary(model) 
     ```

The error message with traceback:

     tar_source(): these files do not exist: R
     + report dispatched                         
     
     
     processing file: report.qmd
     1/4                   
     2/4 [unnamed-chunk-1] 
     Error:
     ! Could not parse knitr report report.qmd to detect dependencies: Duplicate chunk label 'tbl-modelsummary', which has been used for the
 chunk:
     modelsummary(model)
     
     Quitting from report.qmd:7-12 [unnamed-chunk-1]
     Execution halted
     ✖ report errored
     ✖ errored pipeline [1.5s, 0 completed, 2 skipped]
     Error:
     ! Error in tar_make():
       ! Error running quarto CLI from R.
     Caused by error:
     ✖ Error returned by quarto CLI.
       -----------------------------
 
   processing file: report.qmd   1/4   2/4 [unnamed-chunk-1]   Error:  
 ! Could not parse knitr report report.qmd to detect dependencies:
 Duplicate   chunk label 'tbl-modelsummary', which has been used for
 the chunk:   modelsummary(model)
      Quitting from report.qmd:7-12 [unnamed-chunk-1]   Execution halted    Caused by error in `processx::run()`: ! System command
 'quarto' failed   See
 https://books.ropensci.org/targets/debugging.html
 
 Hide Traceback
     ▆
  1. └─targets::tar_make()
  2.   └─targets:::callr_outer(...)
  3.     ├─targets:::if_any(...)
  4.     └─targets:::callr_error(traced_condition = out, fun = fun)
  5.       └─targets::tar_throw_run(message, class = class(traced_condition$condition))
  6.         └─targets::tar_error(...)
  7.           └─rlang::abort(message = message, class = class, call = tar_envir_base)

[ginolhac]

Hello,

using similar packages versions it works on my Linux machine:

The report.qmd I changed a few things:

--- 
title: "Duplicate chunk label where there is none" 
format: html
--- 

The following creates an error when embedded within a targets project.

```{r}

library(targets) 
library(modelsummary)
tar_load(model) 
```

Table @tbl-modelsummary shows a model summary

```{r}
#| label: tbl-modelsummary
#| tbl-cap: "Model summary" 

modelsummary(model) 
```

like not load globals are they are none, add library(modelsummary) since we need it and make it clear the we call it not next to the chunk option for tbl caption

[chihayakenji]

Hi
Thank you for the reply. The toy example works after making the changes you suggested. Playing around with it a bit more, it seems that it is the tar_load_globals() call that is causing the problem. This is of concern because, while there are no globals in the reproducible example, the real project has globals. What should I do in this case?

Also, do you know why I have to library(modelsummary) in the quarto document when it is already specified within the _targets.r file within tar_option_set()? Is it the case that quarto documents behave differently?

[ginolhac]

I guess because tar_quarto() only calls quarto render and this process is independent so you need what you use. For example you also need library(targets).

Can you try source() in the qmd for the individual R scripts? And avoid using tar_load_globals().

[chihayakenji]

This worked! I'm still curious about what was going on, though. Does tar_load_globals() duplicates the chunk leading to quarto thinking that there are two chunks with the same label?

[ginolhac]

I actually never used tar_load_globals(). Reading the help page:

Load user-defined packages, functions, global objects, and settings defined in the target script file (default: ⁠_targets.R⁠). This function is for debugging, testing, and prototyping only. It is not recommended for use inside a serious pipeline or to report the results of a serious pipeline.

So I guess we should not plan on using that. Load the packages you needs in targets script and report.qmd.
For loading your helper R scripts in R/*R you can also tar_source() if you have more than one instead of individual source().

PS: maybe answer yourself on stackoverfow.

[rkrug]

It went through the first chunks - the chunk names as far as I remember evaluated fr duplication BEFORE the chunk processing starts.

So my guess it,is is something in the part loaded dynamically when the code is rendered - in the target?

[chihayakenji]

So could it be that calling tar_load_globals() was making quarto see the same chunk twice by loading it?