Diagnostics: No symbol in scope when assigning variables in function calls

[juliasilge]

Discussed in https://github.com/posit-dev/positron-beta/discussions/175

^{Originally posted by jeffkeller-einc May 7, 2024}

I don't know what this pattern is called, but I do it a lot to modularize my non-package R projects.

source("helper_functions.R", local = helpers <- new.env())
helpers$myfun(...)
```</div>

[juliasilge]

This is valid code that we currently do not correctly lint:

sum(x <- 1:10)
#> [1] 55
x * 2
#>  [1]  2  4  6  8 10 12 14 16 18 20

^{Created on 2024-05-07 with reprex v2.1.0}

It says "no symbol named 'x' in scope" for the second x there.

[lionel-]

In theory, we should analyse the semantics of argument evaluation and only consider <- when passed to "strict" arguments (non-lazy/NSE ones). In practice, we can just consider all arguments to be strict, at least in the short and medium terms.

Such a pragmatic approach will have lots of false positives though, in particular local(x <- foo) and test_that("", { x <- 1 }) which we should special-case.

For the longer term, the annotations discussed last year at the R sprint would be useful here: https://bugs.r-project.org/show_bug.cgi?id=14354

[jennybc]

I just came across a variant of this in cli, except it's a variable assignment in an if() condition.

The general form that's problematic:

if (nzchar(ASDF <- Sys.getenv("ASDF"))) {
  if (ASDF == "whatever") {
    message("hi")
  }
}

In Positron, the ASDF inside the if body gets the diagnostic "No symbol named 'ASDF' in scope."

The real world example: https://github.com/r-lib/cli/blob/7ca06ff9c1c881359b562ecb98a1414e44d78fb8/R/ansi-hyperlink.R#L347-L355

[lionel-]

I'm now thinking we should be overly optimistic about potential assignments instead of overly cautious.

[testlabauto]

Verified Fixed

Positron Version(s) : 2025.01.0-2
OS Version          : OSX

Test scenario(s)

Verified with Julia's example

Link(s) to TestRail test cases run or created: