[ideas] Warn when name conflicts introduced by static branching

[shiyw]

Help

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

Description

Currently, {tarchetypes} does not warn name conflict of names from values of tar_map and names in command of tar_target.

# _targets.R
f <- \() cat("Hi")

values <- tibble::tibble(
  f = 1:10
)

list(
  tar_map(
    values = values,
    tar_target(target, f())
  )
)

tar_make()

▶ dispatched target target_6
✖ errored target target_6
▶ completed pipeline [0.198 seconds]
Error:
! Error running targets::tar_make()
Error messages: targets::tar_meta(fields = error, complete_only = TRUE)
Debugging guide: https://books.ropensci.org/targets/debugging.html
How to ask for help: https://books.ropensci.org/targets/help.html
Last error message:
    attempt to apply non-function
Last error traceback:
    .handleSimpleError(function (condition)  {     state$error <- build_mess...
    h(simpleError(msg, call))

... although tar_manifest gives some hints.

tar_manifest()

# A tibble: 10 × 2
   name      command
   <chr>     <chr>  
 1 target_6  6L()   
 2 target_10 10L()  
 3 target_7  7L()   
 4 target_8  8L()   
 5 target_9  9L()   
 6 target_1  1L()   
 7 target_2  2L()   
 8 target_3  3L()   
 9 target_4  4L()   
10 target_5  5L()   

A repo storing this reprex

[wlandau]

Interesting edge case. I still think the general behavior is correct because tar_map() operates on static code analysis, and your use of values is telling tar_map() to treat f as a placeholder. The static code analysis and meta programming happen before the pipeline runs, and the former is by design not aware of any objects or functions in memory that the expression may reference.

[shiyw]

Yeah, it's unusual usage which no one wish to use that. And current behavior is correct in the sense that commands in tar_map run after the global variables are set, so values introduced by tar_map should precede the global variables.

My suggestion is should there be a warning in such cases? Since it did take me sometime for me to diagnosis the problem when I unintentionally introduced the conflict. Or, maybe we can mention it in the manual or reference page.

I learned to add some special prefix (like arg_) for fields in values to avoid unintended conflicts.