Skip to contents

Changelog

Source: NEWS.md

rxode2ll (development version)

  • llikNbinom() and llikNbinomMu() now accept a continuous (non-integer) size. In the negative binomial’s mean/dispersion parameterisation size is a real dispersion parameter, not a count, and stats::dnbinom() has always allowed it. Previously size was stored in an integer vector and silently truncated, so a size above 1 returned the log-likelihood at trunc(size) and a size between 0 and 1 truncated to 0 and aborted; the R-level assertIntegerish(size) rejected such values outright. Truncation also made the log-likelihood a step function of size, so the dispersion could not be estimated even when its true value was an integer. This blocked nbinomMu() models with continuous overdispersion in nlmixr2. size must now be strictly positive; non-positive values return NA. The INT_MAX bound added in 2.0.15 still applies to x everywhere, and to size for rxLlikBinom where size is a number of trials.

  • llikNbinom() and llikNbinomMu() now return NA instead of aborting the R process when the mean handed to the underlying likelihood is out of domain. This covers a prob of 0 (and, through the C API, a prob outside [0, 1] or a negative mu), and – because size is no longer bounded by INT_MAX – a large size with a small prob whose product overflows. Previously these threw a C++ exception that escaped the rxLlikNbinom() / rxLlikNbinomMu() entry points used by rxode2 solves and focei, killing the session.

  • llikNbinom() at prob == 1 and llikNbinomMu() at mu == 0 now match stats::dnbinom(). The distribution is degenerate at zero there, so fx is 0 (a likelihood of 1) at x == 0 and -Inf otherwise. These are the only points where stats::dnbinom() defines a value that the underlying Stan likelihood cannot compute – it requires a strictly positive mean. No derivative exists at a point mass, so dProb / dMu are NA; an unusable gradient beats a NA log-likelihood, and beats the abort these used to cause. prob == 0 remains NA, matching the NaN R returns there.

rxode2ll 2.0.16

CRAN release: 2026-07-24

  • Fix ABI issues in rxode2 and the nlmixr2 ecosystem by requiring RcppParallel 6.0.0

rxode2ll 2.0.15

CRAN release: 2026-07-17

  • Fix signed integer overflow: loop indices in llikXxxInternal() Rcpp functions changed from int to R_xlen_t; previously vectors longer than ~500 million elements (>4 GB) could cause undefined behavior due to int overflow when assigned from x.size().
  • Fix undefined behavior from unchecked double-to-int casts for discrete distribution count arguments (x, size) exceeding INT_MAX (2147483647); the five affected C-API functions (rxLlikPois, rxLlikBinom, rxLlikGeom, rxLlikNbinom, rxLlikNbinomMu) now return NA instead of invoking undefined behavior for out-of-range inputs.
  • Fix type correctness for loop indices in Stan autodiff functor operator() methods across all 15 distributions: changed from int to Eigen::Index to match the return type of Eigen::VectorXd::size().
  • Add comprehensive derivative validation tests using central finite differences for all 14 differentiable distributions.
  • Add integer overflow bounds-checking tests for all 5 discrete distributions, calling the *Internal() C++ functions directly to bypass R-level input validation.
  • Add a skip()-guarded large-vector test that documents the R_xlen_t fix and can be run manually on systems with >4 GB free RAM.
  • Changed _log() to _lpmf()/_lpdf() because these functions changed in StanHeaders

rxode2ll 2.0.14

CRAN release: 2026-02-24

  • Make all distribution calculations thread safe

rxode2ll 2.0.13

CRAN release: 2024-12-15

  • Update compile flags in anticipation of BH 1.87.0 (#10)

rxode2ll 2.0.12

CRAN release: 2024-11-21

  • Update compile flags for alpine linux (#7)

rxode2ll 2.0.11

CRAN release: 2023-03-17

  • Update compile flags based on CRAN C++17 requirement (Issue #5)

rxode2ll 2.0.10

CRAN release: 2023-03-16

  • Update compile flags based on StanHeaders change.

rxode2ll 2.0.9

CRAN release: 2022-09-29

  • Initial release, exports log-likelihood functions from ‘stan’ and their derivatives to R and C. This allows linkage in ‘rxode2’ to allow generalized log-likelihood in ‘nlmixr2’.
  • This is a data like package and is likely not to change very much over time.
  • Added a NEWS.md file to track changes to the package.