Don't override user-specified NOT_CRAN env var

[hadley]

Continutation of #2112

[gaborcsardi]

Makes sense!

[gaborcsardi]

Lost the .env argument here?

[hadley]

Good catch, thanks!

[claude]

Claude finished @hadley's task —— View job

---

Code Review Analysis

Reviewing PR #2200: "Don't override user-specified NOT_CRAN env var"

Todo List:

• Read and analyze all changed files
• Review code quality and best practices
• Check for potential bugs and issues
• Evaluate performance considerations
• Assess security concerns
• Examine test coverage
• Address reviewer feedback about missing .env argument
• Provide comprehensive review feedback

Review Summary

This PR successfully addresses the issue of overriding user-specified NOT_CRAN environment variables by introducing a new function local_assume_not_on_cran() that only sets NOT_CRAN = "true" when it's not already set by the user.

Code Quality & Best Practices ✅

Strengths:

• Clean abstraction: The new local_assume_not_on_cran() function provides a clear, single-purpose utility
• Consistent naming: Follows testthat's local_* function naming convention
• Proper documentation: Function behavior is well-documented in comments
• Strategic placement: Located logically in R/skip.R alongside related CRAN detection functions

Good implementation pattern:

local_assume_not_on_cran <- function(frame = caller_env()) {
  if (Sys.getenv("NOT_CRAN") != "") {
    return()  # Respect existing user setting
  }
  withr::local_envvar("NOT_CRAN" = "true", .local_envir = frame)
}

Potential Issues & Bugs ⚠️

Critical Issue: Missing .env parameter (R/local.R:193)
Reviewer @gaborcsardi correctly identified that the .env argument was lost during refactoring:

# Before (line 193):
withr::local_envvar(NOT_CRAN = "true", .local_envir = .env) 

# After (line 193): 
local_assume_not_on_cran(.env)  # ✅ CORRECT - .env parameter preserved

# But in the function definition:
local_assume_not_on_cran <- function(frame = caller_env()) {
  # Should accept .env and pass it appropriately
}

Inconsistent parameter naming: The new function uses frame parameter while other local_* functions in testthat use .env. This creates inconsistency in the codebase.

Test Coverage ✅

Excellent test coverage in tests/testthat/test-skip.R:91-103:

• Tests both scenarios: when NOT_CRAN is unset vs. already set
• Verifies that user-specified values are preserved
• Uses proper test isolation with withr::local_envvar()

Performance Considerations ✅

Minimal performance impact:

• Single environment variable check via Sys.getenv()
• Early return when variable already set
• No additional computational overhead

Security Concerns ✅

No security issues identified:

• Only reads/sets environment variables in controlled manner
• Uses standard withr patterns for temporary modifications
• No exposure of sensitive information

Recommendations

1. Fix parameter naming consistency - Change frame to .env to match testthat conventions:

   local_assume_not_on_cran <- function(.env = parent.frame()) {
     if (Sys.getenv("NOT_CRAN") != "") {
       return()
     }
     withr::local_envvar("NOT_CRAN" = "true", .local_envir = .env)
   }

2. Consider adding usage documentation - Brief roxygen comment explaining when to use this vs local_on_cran()

Overall Assessment

This is a well-implemented solution that properly addresses the core issue. The logic is sound, test coverage is thorough, and the approach is consistent with testthat's architecture. The only issues are minor parameter naming inconsistencies that should be addressed before merging.

Approval Status: ✅ Approve after addressing parameter naming