v5.0.1

• Change random number generation to always return doubles internally as well as no longer take a rxode2 individual structure, this is inferred by the thread number.

• Change string representation of model variables to internal binary C code (to avoid macOS M1 sanitizer issues with strings).

• Allow user to change the internal serialization type with options("rxode2.serialize.type"); Currently can be one of “qs2”, “qdata”, “base”, “bzip2” and “xz”. This option must be set before rxode2 is loaded, once loaded it keeps the option initially set. This is set to xz which is from base R, but could be sped up with either "qs2" (more future proof) or "qdata" (a bit faster).

• Removed lsoda CDIR$ IVDEP directive, as requested by CRAN.

What's Changed

• Fix CRAN issues by @mattfidler in #963
• Use (void*) references for rxbinom() and other random number gen by @mattfidler in #964
• Change random number generators to not include the rxode2 individual patient structure. by @mattfidler in #967
• Add/Use bzip2 by @mattfidler in #968
• Use xz as default serialization/compression by @mattfidler in #969

Full Changelog: v5.0.0...v5.0.1

v5.0.0

rxode2 5.0.0

• Better error for tad(depot) when linCmt() doesn't include a
  depot compartment.

• Remove qs dependency; For rxode2 ui objects, use lists instead of
  serialized objects. The internal C++ code still generates qs2
  sterilization objects (#950)

• Fixed translation for censoring/limit to account for a possible
  CMT variable before the CENS / LIMIT column (#951, #952)

• Added dmexpit() for getting the diagonal Jacobian.

• Added special handling of mixest and mixunif.

What's Changed

• CRAN release by @mattfidler in #941
• Add special handling of mixest and mixunif by @mattfidler in #944
• add dmexpit() by @mattfidler in #946
• Fix grammar in error messages: "in not in" → "is not in" by @Copilot in #953
• Checks for censoring based on name by @mattfidler in #952
• Use qs2 instead of qs by @mattfidler in #957

New Contributors

• @Copilot made their first contribution in #953

Full Changelog: v4.1.1...v5.0.0

v4.1.1

rxode2 4.1.1

• Stacking for multiple-endpoint ipredSim now matches
  multiple-endpoint sim; Issue #929

• Fix occasional $props that threw an error with empty properties
  (when using properties like tad0()); Issue #924

• Allow mixture models mix() to be loaded with rxS() as a step to
  support mixtures in nlmixr2's focei; Issue #933.

• Identify the correct transformation type for iov variables (#936)

• Fix multiple compartment simulation edge cases where simulations
  were not being performed (#939)

• When referencing cmt in models, the variable is forced to be CMT
  (related to #939)

v4.1.0

• Change rxui $ evaluation when completing in rstudio, fixes strange calculations popping up in rstudio (#909)

• Add orphan rxode2 model unloading when using rxUnloadAll(), and change the return type to always be a boolean.

• Add assertRxUiIovNoCor to assert IOVs have no correlations in them.

• Handle the levels for inter-occasion variability in the ui better (#614)

• Create a new function mix() that will allow mixture models to be simulated in preparation of mixture support in nlmixr2. This allows mixture models to be specified as: v = mix(v1, p1, v2, p2, v3) where the probability of having v=v1 is modeled by p1, v=v2 is modeled by p2, and v=v3 is modeled by probability 1-p1-p2.

• Created new functions mlogit() and mexpit() to convert probabilities used in mixture models to log-scaled values. mlogit() converts the probabilities to log-scaled values (using root-finding) and mexpit() converts the log values into probabilities. The equation for the conversion of log to probabilities is $p_i = \frac{exp(x_i)}{1+\sum_{j=1}^{N-1}exp(x_j)}$

• Added new assertion assertRxUiNoMix which throws an error when a mixture model is present (ie mix())

• Fix for label processing when calling rxode2(uiModel)

What's Changed

• 909 rstudio 2025 by @mattfidler in #910
• Try unloading orphan rxode2 dlls by @mattfidler in #911
• Add assertion for new nlmixr2est by @mattfidler in #913
• iov bad cov by @mattfidler in #915
• Fix and test for #919 by @mattfidler in #920
• 918 mix by @mattfidler in #921
• 922 probit probit inv by @mattfidler in #923

Full Changelog: v4.0.3...v4.1.0

v4.0.3

rxode2 4.0.3

• For CRAN's m1 ASAN checks of nlmixr2est, loading and unloading the
  same dll or by deleting the dll and recreating the exact same code,
  and then loading the dll will cause the ASAN check to flag an odr
  violation. Because of this, a mechanism to not unload dlls has been
  added. This allows the next version of nlmixr2est to not have
  issues with Mac m1 san checks.

v4.0.2

rxode2 4.0.2

• At the request of CRAN, be a bit more careful so that names are not
  duplicated. Now include the md5 hash, a global counter and random 4
  digit and number combination. In addition add the name of the
  original function so it will be easier to debug in the future.

• Fall back to data.frame rbind when rbind.rxSolve() fails

What's Changed

• New ODR, as indirectly request by CRAN. by @mattfidler in #904

Full Changelog: v4.0.1...v4.0.2

v4.0.1

rxode2 4.0.1

• Add the ability to use rbind for solved rxode2 frames.

• Fix LTO issue for _rxode2_calcDerived

What's Changed

• Remove more tests for CRAN by @mattfidler in #898
• fix typo by @leovuong in #900
• Another rbind instance by @mattfidler in #899
• Cran update by @mattfidler in #902

Full Changelog: v4.0.0...v4.0.1

v4.0.0

rxode2 4.0.0

• Add more information errors about NAs during solving.

• Fix rxDerived() for mixed vector and non-vector input.

• Fix model variables for alag(cmt) when they are defined before
  d/dt() or linCmt()

• Just in time use of state.ignore in the model variables, fixes
  negative length error observed in #857.

• Fix steady state bug with time-varying covariates. Now the
  covariates are inferred at the time of the steady state (instead of
  searching through the subject based on the projected time).

• Rework the linear solved systems to use the wnl solutions, and
  threaded linear systems solve (for non-gradient solutions). This new
  method closes a variety of linear compartment model bugs (#261,
  #272, #441, #504, #564, #717, #728, #827, and #855)

• Added new types of bounds for event tables:

  • 3 point bounds et(list(c(low, mid, high))) when specified this way,
    they will not change. Perfect for use with babelmixr2's PopED (#862,
    #863, #854)

  • Intervals simulated by normal values instead of uniform. In this
    case the first seen interval will be 3 elements with NA at the end
    et(list(c(mean, sd, NA), c(mean, sd))), and the other elements
    can simply be 2 declaring the c(mean, sd)

  • Of course the uniform windows of et(list(c(low, high))) still work

  • Currently these different types of windows cannot be mixed.

• Add ability to pipe a list or named numeric as an eta with
  %>% ini(~eta)

• Added a fix for event tables where expanding IDs in non-sequential
  order. In particular if the first ID is not the minimum ID when expanding
  the first event table, the smallest ID was not in the output table. Now
  the smallest ID is in the event table. (Fixes #878, #869, #870)

• Added ability to pipe ini() or lotri(), or any other expression
  that can be converted to an ini with as.ini(). Also allows ini()
  expressions to be converted to lotri with as.lotri(). Fixes #871

• Added new type of variability expression for simulation and
  estimation with focei and likelihood related methods: +var(). This
  changes standard deviation parameters to variance parameters.

• Added new type of endpoint expression for focei estimation
  +dv(). This only transforms the data and not the predictions. I
  can only see it being useful in model linearization.

• Bug fix for parameters that are in both input ($params) and output
  ($lhs) that respects the order of the $lhs declaration (Fixes
  #876)

• Add rxFixRes to literally fix the residual estimates in a model (#889)

• Now modeled duration of 0 is treated as a bolus dose (#892)

What's Changed

• Add piping tests and piping for eta names by @mattfidler in #860
• Create 3 sampling window examples by @mattfidler in #863
• Update continuous integration by @billdenney in #864
• Copilot suggested fix for et() by @mattfidler in #869
• Add piping for lotri and ini expressions by @mattfidler in #872
• 873 var by @mattfidler in #874
• Fix rxode2 lhs order when a covariate and output match by @mattfidler in #877
• 880 dv by @mattfidler in #881
• 827 linear compartment solutions by @mattfidler in #856
• Pre-declare theta for linCmtB by @mattfidler in #883
• Better error; difficult to test by @mattfidler in #888
• add/test rxFixRes() by @mattfidler in #890
• Handle dur=0 as a bolus by @mattfidler in #893
• fix typo by @leovuong in #896

New Contributors

• @leovuong made their first contribution in #896

Full Changelog: v3.0.4...v4.0.0

v3.0.4

rxode2 3.0.4

• Add stable hashes for rxUi objects (#838, #689)

• Fix for iov simulation (#842)

• Fix for rxnbinom() called directly from R (#847) and expand it to
  match more close with R's rnbinom() including allowing named mu=
  calls. In rxode2 ui, these are also now allowed.

What's Changed

• rebuild documentation by @billdenney in #830
• Create a rxUi method for digest by @mattfidler in #838
• Fix nesting by @mattfidler in #844
• Fixes for rxnbinom by @mattfidler in #849

Full Changelog: v3.0.3...v3.0.4

v3.0.3

rxode2 3.0.3

• Add logit/expit named expressions, that is logit(x, high=20)
  becomes logit(x, 0, 20) in ui models.

• Updated random ui models like rxnorm(sd=10) to accept complex
  numeric expressions like rxnorm(sd=10+1).

• Updated random ui models to accept complex non-numeric expressions
  like rxnorm(sd=a+b)

• Rework the tad() and related functions so they use the same
  interface as compartments (this way they do not depend on the order
  of compartments); See #815. For mu-referencing, Also allow dummy
  variables to ignore state requirements (ie podo(depot) in a single
  line will not error when parsing mu-referenced equations).

• Add getRxNpars to api. This allows the development version of
  babelmixr2 to better check what model is loaded and unload/reload
  as necessary.

• Add rxUdfUiControl() to rxode2 user function to get control
  information from something like nlmixr2

• Bug fix for tracking time after dose when dosing to 2 compartments
  occur at the exact same time (#804, #819)

• Change transit() model so that it uses tad0(), podo0() and
  related functions for a bit more stable simulation and estimation

• Fix compile flags to work with BH 1.87 (#826)

What's Changed

• Add getRxNpars to api by @mattfidler in #802
• add rxUdfUiControl() by @mattfidler in #807
• White list PopOS with isystem by @mattfidler in #809
• simplify repeated ifelse by @billdenney in #800
• Add NN activation fns by @mattfidler in #811
• Create R NN activation interface by @mattfidler in #813
• Fix parsing for tad() and podo() by @mattfidler in #817
• Change the tlast handling to be based on tlast per cmt by @mattfidler in #819
• Named logit/expit by @mattfidler in #824
• Fix BH flags to work with BH 1.87.0 by @mattfidler in #826

Full Changelog: v3.0.2...v3.0.3