mrgsolve 2.0.0

Breaking Changes

• data_set() and idata_set() no longer accept .subset, .select,
  object, or need arguments. These arguments allowed filtering and
  column selection inside the call; that processing should now be done on the
  data frame before passing it to data_set() or idata_set(). (#1374)

• The n argument to simeta() and simeps() has been discouraged for a
  while and is now removed. Calling simeta(n) or simeps(n) to resimulate
  a single ETA or EPS value is no longer supported. Use simeta() or simeps()
  (no argument) to resimulate all values. (#1373)

• knobs(), wf_sweep(), and render() have been removed. These
  functions were previously deprecated. (#1369)

• The default value of root in $NMXML and $NMEXT has changed from
  "working" to "cppfile". The root directory for locating NONMEM
  output files now defaults to the directory containing the model .cpp file
  rather than the R working directory. Pass root = "working" to restore the
  previous behavior; but users are highly encouraged to use the new default.
  (#1368)

• summarise.each() is removed; it has not been practically reachable by the
  user for several years (#1352).

• Several modlib model library models have had parameter and compartment
  names standardized (#1361):

  • The first extravascular compartment is now named EV (was EV1) in
    pk1cmt, pk2cmt, irm1-irm4, and emax.
  • The first absorption rate constant is now named KA (was KA1) in
    models that previously used numbered names.
  • Code that references these compartments or parameters by name (e.g.,
    init(mod, EV1 = 0) or param(mod, KA1 = 1)) will need to be updated.
  • We continue to discourage users from using these models in production
    code.

New Features

• Models can now use a ** b syntax to represent pow(a, b); this syntax is
  always available, without need to invoke a plugin (#1360).

• With the nm-vars plugin, mrgsolve now recognizes NONMEM / Fortran
  IF / ELSE / THEN statements; this code can be left in your model and
  mrgsolve will convert to appropriate C++ under the hood (#1360).

• ERR(n) is now recognized as an alias for EPS(n) when using the nm-vars
  plugin (#1367).

• New DSL preprocessing functions convert NONMEM/Fortran model code to
  mrgsolve / C++ syntax (#1360, #1362, #1366):

  • convert_pow() converts Fortran ** exponentiation to pow().
  • convert_fort_if() converts Fortran IF/THEN/ELSE/END IF blocks to C++
    if/else syntax.
  • convert_semicolons() inserts missing trailing semicolons on assignment
    statements.
  • RStudio addins are provided to allow conversion of certain NONMEM source
    code blocks to mrgsolve format in the RStudio editor window.

• Closed-form three-compartment linear models with or without depot compartments
  can now be implemented via $PKMODEL (#1345).

• $PKMODEL gains an advan input for selecting 1, 2, or 3 compartment
  models with analytical solutions; for example, setting advan to 2, 4,
  or 12 will give you 1, 2, or 3 compartment model in closed form with a
  depot compartment. When advan is specified, compartments with standard
  names are automatically registered if neither $INIT nor $CMT are
  found in the model file and no compartments are registered at the time
  that $PKMODEL is processed (#1345).

• Two new modlib models, pk3 and pk3iv, provide the corresponding
  pre-coded model files (#1345).

• mrgx::assign() is now available in the mrgx plugin; this is a convenience
  function to help you get R objects from your model code back to R when the
  simulation finishes (#1353).

• The plot() method for mrgsims objects gains a fixy argument to fix the
  y-axis limits across panels in the plot (#1349).

• Users can now find plot() method documentation through ?plot (#1349).

• env_get() refactored to allow access of R objects inside the model
  environment similar to base::get(); env_get_obj() is an alias to env_get(),
  analogous to env_get_env() (#1355).

Bugs Fixed

• Fixed a bug where the model row counter was not correct when a simulation
  was performed with only an idata set (#1351).

mrgsolve 1.8.0

• mrgsolve now requires R >= 4.1 (#1333).

• FINAL_ROW and FINAL_IROW macros are now available to indicate when the
  simulation is processing the final output record for the whole problem or for
  the current individual, respectively; both macros are reserved words
  (#1328, #1327).

Bugs Fixed

• Fixed S3 registration of the all.equal method (#1337).

• custom_rtol() and custom_atol() now return without error when called with
  no tolerance inputs (#1321, #1336).

• Row counters (self.rown, self.nrow, self.irown, self.inrow) now only
  update when the system advances to an output record; previously, non-output
  records between output records could cause the counters to misrepresent the
  current position in the output (#1323).

Internal

• Model parameter and compartment aliases in generated C++ code now use scoped
  reference variables rather than preprocessor #define directives (#1332).

  • During this refactor, it was noted that model parameters were mutable
    in $TABLE (and $EVENT), against the intention for the parameter
    vector. This vector continues to be mutable in these blocks, but plans will
    be made to fix this behavior in the future.

• Initialze Self to null pointer in evt::regimen objects (#1329).

• Refactor mat2df() (C++) for efficiency, avoiding Rcpp sugar (#1316).

mrgsolve 1.7.2

• An individual record counter has been added. Use self.inrow to get the number of
  output rows for the current individual; use self.irown to get the zero-indexed row
  number for the current individual; these are similar to self.nrow and self.rown,
  which count records for the entire output data set (#1313).

Bugs Fixed

• The code argument to mcode() now accepts a character vector with length
  greater than 1; passing multiple lines previously caused a compile error (#1310).

mrgsolve 1.7.1

• The 1.7.0 release added safety checks to help users avoid an unintentional
  mismatch between model code and data set. These checks now signal a warning
  instead of an error to give downstream tools that use mrgsolve more
  flexibility in how they adjust to the checks (#1307).

mrgsolve 1.7.0

• Safety checks added to avoid unintentional mismatch between model code and
  data set when infusion duration (Dn) or rate (Rn) are modeled in
  $PK (#1304).
  • If there is a positive infusion duration in the model (e.g., D1
    in $PK) and data set rate is not -2 on a dosing record, an error will be
    issued. If there is a positive infusion rate in the model (e.g., R1
    in $PK) and data set rate is not -1 on a dosing record, a similar error
    will be issued (check direction: model to data set).
  • These checks are in addition to existing behavior that issued an error
    when rate was -1 or -2 in the data and R1 or D1 were not set to a
    positive value in the model (the opposite check direction: data set to
    model).
  • Original behavior can be restored by turning off these checks either by a
    block option (@!check_modeled_infusions) on $PK or through the
    CHECK_MODELED_INFUSIONS macro available in $PREAMBLE only.

mrgsolve 1.6.1

• Internal refactor of self.mevent() and self.mtime() to not send event
  objects back to mrgsolve when the requested object time is in the past; this
  preserves previous functionality while respecting change in 1.6.0 that
  generates an error for events (or event sequences) that happen or begin in
  the past (#1297).

mrgsolve 1.6.0

• evt::tgrid() was added to the evtools plugin, allowing users to insert
  internal-only (non-output) records into the simulation for
  monitoring the system at a regular grid of times (#1291).

• evt::evid() was added to the evtools plugin allowing users
  to reset the event ID attribute of event objects created inside a model (#1291).

• EVID behavior was changed so that event IDs greater than or equal to 100
  do not trigger a reset of the ODE solver, functioning like observation records
  with EVID=0; for EVID < 100, a discontinuity will continue to happen through
  ODE solver reset (#1291).

• Event objects created in the model (typically through the evtools plugin) will
  now trigger an error when the event or event sequence starts before the
  current simulation time; these events were previously ignored silently
  (#1291).

• Absolute (atol) and relative (rtol) tolerances can now be customized for each
  compartment in an ODE-based model (#1264)

  • Usecustom_rtol(), custom_atol(), and custom_tol() to set tolerances
    on a compartment-by-compartment basis.
  • Use reset_rtol(), reset_atol(), or reset_tol() to reset custom
    tolerances.
  • Use use_custom_tol() and use_scalar_tol() to toggle the model to use
    either the customized (vector) values or the scalar values.
  • Use get_tol() or get_tol_list() to extract both custom and scalar
    tolerance values.
  • The model print method has been updated to reflect whether the model is
    currently set to use the scalar or customized (vector) values.

Bugs Fixed

• Fixed a bug where incorrectly spelled block names were inadvertently aliased
  resulting in a misleading error message (#1294).

• Fixed a bug where no OMEGA or SIGMA matrices were imported into the
  model via multiple $NMEXT or $NMXML blocks when the matrix from one
  block was ignored (#1276).

mrgsolve 1.5.3

• D_CMT, ALAG_CMT, and simeta() are now available to use in $TABLE
  ($ERROR), $EVENT, and $PRED code blocks (#1290).

• Added END_OF_INFUSION syntax, which is true on the internal record that
  stops an infusion (#1288).

• mwrite_yaml() and mwrite_cpp() now render block names in upper case,
  even for blocks coded with lower case in the original model file (#1282).

• When creating an evt::regimen object (via evtools plugin), the default
  start time is now the time when the object was created; before this change,
  the default time was always 0 (#1266).

• Added methods for evt::regimen objects (via evtools plugin) under
  the name time_next(); this either gets or sets the time of the
  next dose in the regimen (#1266).

Bugs Fixed

• Fixed an issue where time values in the simulated output data frame
  were off by a very small amount, due to issues with floating point
  mathematical operations (#1287).

• Fixed a problem writing a model to file with mwrite_yaml() and
  mwrite_cpp() when lower case block naming was used (#1282).

• Fixed a bug where invalid item passed through the $SET block caused
  an error when reading / building a model (#1271).

• Fixed a bug where mrgsolve was not matching NONMEM behavior when an
  EVID=4 record was included in the data set with SS=1; mrgsolve will
  now match NONMEM behavior in this case, with the output being identical
  to EVID=1 with SS=1 (#1262).

mrgsolve 1.5.2

• Add $EVENT block for writing code related to dosing or other events that
  are implemented through model code rather than the data set (#1230).

• Add evt::reset() and evt::reset(self) functions under the evtools
  plugin; these reset the compartments in a model; overloaded functions are
  also provided to reset and dose with bolus or infusion (#1222).

• Completed dosing functionality in evtools plugin; use evt::addl() to
  schedule additional doses through an evt::ev object; use evt::ii() to
  set the dosing interval; use evt::ss() to advance the pharmacokinetic
  system to steady state just prior to dosing; evt::cmt() sets the compartment
  number; evt::amt() sets the dose amount; evt::rate() sets the infusion
  rate; see the user guide for the specific signatures that are available
  (#1227).

• Add evtools model to modlib(), illustrating how to implement dosing
  regimens from inside the model a few different ways (#1230).

• Added more comprehensive checking for duplicate blocks in a model file;
  duplicate blocks are always handled when allowed; an error message is always
  issued when duplicates are not allowed (#1238).

• Code to audit $ODE (or $DES) code, looking for an equation for every
  model compartment was refactored to use a common approach for both traditional
  models and models written with the nm-vars plugin; regardless of approach,
  the user will be warned if mrgsolve does not detect code relevant to every
  model compartment; the audit system can be bypassed by including the
  @!audit block option to $ODE (#1235).

• The autodec plugin was lightly refactored to avoid false positive detection
  of variables declared as double; plans are in place to narrow the
  scope of what is detected for declaration in future releases (#1234).

• R help files (.Rd) reviewed and revised for consistency and formatting
  (#1246).

mrgsolve 1.5.1

• Fixed yaml_to_cpp() example code to prevent saving the file to the working
  directory (#1220).