NEWS

cpp4r 0.7.0

I re-added some conditionals to use newer C++ features when the compiler supports them. These can add a small speed gain in some cases.
The codebase default is C++11, with compile time newer standards detection to use newer features and optimizations.
Adds a check for all supported C++ standards in R >= 4.0.0 (>= CXX17).
Replaces 'testthat' with 'tinytest' after I detected an incompatibility with C++23.

register() now captures the C++ namespace of decorated functions and emits fully-qualified declarations and calls, so functions defined inside a namespace { ... } block (or written as ns::fun) can be registered without further setup.
Removed the run-time dependency on the decor, tibble, and vctrs packages; the small subset of helpers that register() relies on is now bundled internally and uses base data.frame.
Adds fast paths that provide slight speed improvements (e.g., avoiding branching)
Follows, to the best possible extent, the R's C API description from https://github.com/hadley/r-internals.
Slightly expanded cpp4r::function.
Fixed two empty tests.

Avoids non-API methods when handling data frames (thanks to Luke Tierney for reporting this)

Clearer documentation about the C++ workflow (i.e., how to use anticonf to specify a C++ standard)
Allows for default values like my_fun(int x = 100) to call my_fun() with the same result as my_fun(100L) from R

This is the first release on CRAN
Added as_logicals() and as_strings() in the same style of as_doubles() and as_integers()
Improved memory management for r_vector iterators
Slightly faster than cpp11
vendor() and unvendor() use path = NULL as default to adhere to CRAN policies.

Reduced dependencies on R side
The vignettes are rendered only for the package site to keep the CRAN build minimal