R 4.4.0

CHANGES IN R 4.4.0

SIGNIFICANT USER-VISIBLE CHANGES:

• Startup banners, R --version, sessionInfo() and R CMD check no
  longer report (64-bit) as part of the platform as this is almost
  universal - the increasingly rare 32-bit platforms will still
  report (32-bit).

  On Windows, ditto for window titles.

• is.atomic(NULL) now returns FALSE, as NULL is not an atomic
  vector. Strict back-compatibility would replace is.atomic(foo)
  by (is.null(foo) || is.atomic(foo)) but should happen only
  sparingly.

NEW FEATURES:

• The confint() methods for "glm" and "nls" objects have been
  copied to the stats package. Previously, they were stubs which
  called versions in package MASS. The MASS namespace is no longer
  loaded if you invoke (say) confint(glmfit). Further, the "glm"
  method for profile() and the plot() and pairs() methods for class
  "profile" have been copied from MASS to stats. (profile.nls()
  and plot.profile.nls() were already in stats.)

• The confint() and profile() methods for "glm" objects have gained a
  possibility to do profiling based on the Rao Score statistic in
  addition to the default Likelihood Ratio. This is controlled by a
  new test = argument.

• The pairs() method for "profile" objects has been extended with a
  which = argument to allow plotting only a subset of the
  parameters.

• The "glm" method for anova() computes test statistics and
  p-values by default, using a chi-squared test or an F test
  depending on whether the dispersion is fixed or free. Test
  statistics can be suppressed by giving argument test a false
  logical value.

• In setRepositories() the repositories can be set using their
  names via name = instead of index ind =.

• methods() and .S3methods() gain a all.names option for the (rare)
  case where functions starting with a . should be included.

• Serializations can now be interrupted (e.g., by Ctrl-C on a
  Unix-alike) if they take too long, e.g., from save.image(),
  thanks to suggestions by Ivan Krylov and others on R-devel.

• New startup option --max-connections to set the maximum number of
  simultaneous connections for the session. Defaults to 128 as
  before: allowed values up to 4096 (but resource limits may in
  practice restrict to smaller values).

• R on Windows (since Windows 10 2004) now uses the new Segment
  Heap allocator. This may improve performance of some
  memory-intensive applications.

• When R packages are built, typically by R CMD build <pkg>, the
  new --user=<build_user> option overrides the (internally
  determined) user name, currently Sys.info()["user"] or LOGNAME.
  This is a (modified) fulfillment of Will Landau's suggestion in
  PR#17530.

• tools::testInstalledBasic() gets new optional arguments outDir
  and testSrcdir, e.g., allowing to use it in a <builddir> != <srcdir> setup, and in standard "binary" Windows installation
  if a source tests/ folder is present.

• range(<DT_with_Inf>, finite = TRUE) now work for objects of class
  "Date", "POSIXct", and "POSIXlt" with infinite entries,
  analogously to range.default(), as proposed by Davis Vaughan on
  R-devel. Other range()-methods can make use of new function
  .rangeNum().

• New .internalGenerics object complementing .S3PrimitiveGenerics,
  for documentation and low-level book-keeping.

• grid() now invisibly returns the x- and y- coordinates at which
  the grid-lines were drawn.

• norm(., type) now also works for complex matrices.

• kappa(., exact = TRUE, norm = *) now works for all norms and also
  for complex matrices. In symmetric / triangular cases, the new
  argument uplo = "U" | uplo = "L" allows the upper or lower triangular
  part to be specified.

• memDecompress(type = "unknown") recognizes compression in the
  default 'zlib' format as used by memCompress(type = "gzip").

• memCompress() and memDecompress() will use the libdeflate library
  (https://github.com/ebiggers/libdeflate) if installed. This
  uses the same type of compression for type = "gzip" but is 1.5-2x
  faster than the system libz library on some common platforms: the
  speed-up may depend on the library version.

• diff() for objects of class "Date", "POSIXct", and "POSIXlt"
  accepts a units argument passed via ....

• Dynamic help now does a much better job of rendering package
  DESCRIPTION metadata.

• Rprof() gains an event argument and support for elapsed (real)
  time profiling on Unix (PR#18076).

• filled.contour() gains a key.border argument.

• tools::update_pkg_po() gets arguments pot_make and mo_make for
  not re-making the corresponding files, and additionally a
  verbose argument.

• Hexadecimal string colour specifications are now accepted in
  short form, so, for example, we can use "#123", which is
  equivalent to "#112233".

  Thanks to MikeFC for the original idea and Ella Kaye, Malcolm
  Barrett, George Stagg, and Hanne Oberman for the patch.

• Plain-text help shows \var markup by angle brackets.

• The new experimental primitive function declare() is intended to
  eventually allow information about R code to be communicated to
  the interpreter, compiler, and code analysis tools. The syntax
  for declarations is still being developed.

• Functions psmirnov(), qsmirnov() and rsmirnov() in package stats
  have had argument two.sided renamed to alternative, to take into
  account that the permutation distributions of the one-sided
  statistics can be different in the case of ties. Consequence of
  PR#18582.

• sort() is now an implicit S4 generic in methods.

• Formatting and printing, format(z), print(z), of complex vectors
  z no longer zap relatively small real or imaginary parts to zero,
  fixing PR#16752. This is an API change, as it was documented
  previously to round real and imaginary parts together on purpose,
  producing nicer looking output. As mentioned, e.g. in the PR,
  this change is compatible with many other "R-like" programming
  environments.

  We have simplified the internal code and now basically format the
  real and imaginary parts independently of each other.

• New experimental functions Tailcall() and Exec() to support
  writing stack-space-efficient recursive functions.

• Where characters are attempted to be plotted by pdf(),
  postscript() and xfig() which are not in the selected 8-bit
  character set (most often Latin-1) and the R session is using a
  UTF-8 locale, the warning messages will show the UTF-8 character
  rather than its bytes and one dot will be substituted per
  character rather than per byte. (Platforms whose iconv() does
  transliteration silently plot the transliteration.)

  In a UTF-8 locale some transliterations are now done with a
  warning (e.g., dashes and Unicode minus to hyphen, ligatures are
  expanded, permille (‰) is replaced by o/oo), although the OS may
  have got there first. These are warnings as they will continue
  to be replaced by dots in earlier versions of R.

• The matrix multiplication functions crossprod() and tcrossprod()
  are now also primitive and S3 generic, as %*% had become in R
  4.3.0.

• source() and example() have a new optional argument catch.aborts
  which allows continued evaluation of the R code after an error.

• The non-Quartz tiff() devices allow additional types of
  compression if supported by the platform's libtiff library.

• The list of base and recommended package names is now provided by
  tools::standard_package_names().

• cairo_pdf() and cairo_ps() default to onefile = TRUE to closer
  match pdf() and postscript().

• New option catch.script.errors provides a documented way to catch
  errors and then continue in non-interactive use.

• L %||% R newly in base is an expressive idiom for the phrases
  if(!is.null(L)) L else R or if(is.null(L)) R else L.

• The return value from warnings() now always inherits from
  "warnings" as documented, now also in the case of no warnings
  where it previously returned NULL.

• as.complex("1i") now returns 0 + 1i instead of NA with a warning.

• z <- c(NA, 1i) now keeps the imaginary part Im(z[1]) == 0, no
  longer coercing to NA_complex_. Similarly, cumsum(z) correctly
  sums real and imaginary parts separately, i.e., without
  "crosstalk" in case of NAs.

• On Alpine Linux iconv() now maps "latin2", "latin-2", "latin9"
  and "latin-9" to encoding names the OS knows about
  (case-insensitively).

• iconv(sub = "Unicode") now always zero-pads to four (hex) digits,
  rather than to 4 or 8. (This seems to have become the convention
  once Unicode restricted the number of Unicode points to 2^21 - 1
  and so will never need more than 6 digits.)

• NCOL(NULL) now returns 0 instead of 1, for consistency with
  cbind().

• The information for the Euro glyph missing from the Adobe .afm
  files for the Courier, Helvetica and Times families has been
  copied from their URW equivalents - this will improve vertical
  centring in the pdf() and postscript() devices.

• The included BLAS sources have been updated to those shipped with
  LAPACK version 3.12.0. The changes are almost entirely cosmetic.

• The included LAPACK sources have been updated to version 3.12.0
  and some further double-complex routines added.

• There are new font families for the 2014--5 URW 2.0 fonts (see
  ?pdf) which are included in recent versions of Ghostscript.
  These have font widths for most Greek glyphs and a few others
  which were missing from the original versions (whose font
  families remain available for reproducibility, although
  Ghostscript-based viewers will render using the 2.0 versions).

• Improve the large-n efficiency of as.matrix(<dist>), thanks an R
  contributors effort, notably by Tim Taylor and Heather Turner,
  see PR#18660.

• The default and numeric methods of all.equal() get a check.class
  option.

• zapsmall() gets new optional arguments, function mFUN and min.d,
  for extra flexibility; fulfills a wish in PR#18199. Also, it is
  now an implicit S4 generic in package methods.

• The Rd filter for aspell() gains an ignore argument.

• New generic function sort_by(), primarily useful for the
  data.frame method which can be used to sort rows of a data frame
  by one or more columns.

• The icence headers for the RPC code in src/extra/xdr have been
  updated to use the GPL-compatible licence published by Oracle
  America in 2010.

• New function pkg2HTML() in tools to create single-page HTML
  reference manuals for R packages.

• The byte code evaluator now uses less C stack space for recursive
  calls to byte-compiled functions. It also makes more of an effort
  to avoid allocations for scalar return values.

• New completion option backtick (disabled by default) allows
  non-syntactic completions to be wrapped in backquotes. This is
  currently only useful for Jupyter notebooks via the IRkernel
  package, and may cause problems for other backends.

INSTALLATION:

• The parser has been updated to work with bison 3.8.2, which is
  now used for the pre-generated parsers in gram.c, file.c, and
  gramRd.c. A few parser error messages have changed, which may
  affect code that relies on exact messages.

INSTALLATION on a UNIX-ALIKE:

• System valgrind headers are now required to use configure option
  --with-valgrind-instrumentation with value 1 or 2.

• configure will warn if it encounters a 32-bit build, as that is
  nowadays almost untested.

• Environment variable R_SYSTEM_ABI is no longer used and so no
  longer recorded in etc/Renviron (it was not on Windows and was
  only ever used when preparing package tools).

• If the libdeflate library and headers are available, libdeflate
  rather than libz is used to (de)compress R objects in lazy-load
  databases. Typically tasks spend up to 5% of their time on such
  operations, although creating lazy-data databases is one of the
  exceptions.

  This can be suppressed if the library is available by the
  configure option --without-libdeflate-compression.

• configure option --enable-lto=check has not worked reliably since
  2019 and has been removed.

• The minimum autoconf requirement for a maintainer build has been
  increased to autoconf 2.71.

  It is intended to increase this to 2.72 for R 4.5.0: the
  distributed configure file was generated using 2.72.

• The minimum version requirement for an external LAPACK has been
  reduced to 3.9.0.

• No default C++ compiler is set if no C++17 compiler is detected:
  there is no longer an automatic fallback to C++14 or C++11.

  Compilers from the last five years should have sufficient
  support: for others macros CXX and CXXSTD can be set in file
  config.site to provide a fallback if needed.

  The Objective-C++ compiler now by default uses the standard
  selected by R for C++ (currently C++17) rather than the default
  standard for the C++ compiler (which on macOS is still C++98).

INSTALLATION on macOS:

• A new configure option --with-newAccelerate makes use of Apple's
  'new' BLAS / LAPACK interfaces in their Accelerate framework.
  Those interfaces are only available in macOS 13.3 or later, and
  building requires SDK 13.3 or later (from the Command Line Tools
  or Xcode 14.3 or later).

  By default the option uses new Accelerate for BLAS calls: to also
  use it for LAPACK use option --with-newAccelerate=lapack. The
  later interfaces provide LAPACK 3.9.1 rather than 3.2.1: 3.9.1 is
  from 2021-04 and does not include the improved algorithms
  introduced in LAPACK 3.10.0 (including for BLAS calls).

INSTALLATION on WINDOWS:

• The makefiles and installer scripts for Windows have been
  tailored to Rtools44, an update of the Rtools43 toolchain. It is
  based on GCC 13 and newer versions of MinGW-W64, binutils and
  libraries (targeting 64-bit Intel CPUs). R-devel can no longer
  be built using Rtools43 without changes.

• Rtools44 has experimental support for 64-bit ARM (aarch64) CPUs
  via the LLVM 17 toolchain using lld, clang/flang-new and
  libc++.

UTILITIES:

• R CMD check notes when S4-style exports are used without
  declaring a strong dependence on package methods.

• tools::checkRd() (used by R CMD check) detects more problems with
  \Sexpr-based dynamic content, including bad nesting of \Sexprs
  and invalid arguments.

• tools::checkRd() now reports Rd titles and section names ending
  in a period; this is ignored by R CMD check unless environment
  variable _R_CHECK_RD_CHECKRD_MINLEVEL_ is set to -5 or smaller.

• R CMD check now notes Rd files without an \alias, as long
  documented in 'Writing R Extensions' §1.3.1. The check for a
  missing \description has been moved from tools::checkRd() to
  tools::checkRdContents().

• R CMD check now visits inst/NEWS.Rd and OS-specific man
  subdirectories when checking Rd files.

• tools::checkDocFiles() and tools::checkRdContents() now also
  check internal Rd files by default, but "specially" (ignoring
  missing documentation of arguments).

• R CMD Rdiff gets option --useEx.

• R CMD check now warns on non-portable uses of Fortran KIND such
  as INTEGER(KIND=4) and REAL(KIND=8).

  To see the failing lines set environment variable
  _R_CHECK_FORTRAN_KIND_DETAILS_ to a true value.

• When checking Rd files, R CMD check now notes some of the "lost
  braces" that tools::checkRd() finds. Typical problems are Rd
  macros missing the initial backslash (e.g., code{...}), in-text
  set notation (e.g., {1, 2}, where the braces need escaping), and
  \itemize lists with \describe-style entries of the form
  \item{label}{description}.

• R CMD INSTALL (and hence check) will compile C++ code with
  -DR_NO_REMAP if environment variable _R_CXX_USE_NO_REMAP_ is set
  to a true value. It is planned that this will in future become
  the default for compiling C++.

• The new built-in Rd macro \dontdiff{} can be used to mark example
  code whose output should be ignored when comparing check output
  to reference output (tests/Examples/<pkg>-Ex.Rout.save). The
  \dontdiff tag, like \donttest, is not shown on the rendered
  help page, so provides a clean alternative to
  ## IGNORE_RDIFF_(BEGIN|END) comments.

• R CMD build when there is no NAMESPACE, now uses the recommended
  exportPattern("^[^.]"), instead of exporting everything.

• R CMD check now warns about non-ASCII characters in the NAMESPACE
  file (in addition to R files). Such packages are not portable and
  may fail to install on some platforms.

C-LEVEL FACILITIES:

• Headers R_ext/Applic.h and R-ext/Linpack.h used to include
  R_ext/BLAS.h although this was undocumented and unneeded by their
  documented entry points. They no longer do so.

• New function R_missing(), factored out from do_missing(), used to
  fix PR#18579.

• SEXP type S4SXP has been renamed to OBJSXP to support
  experimenting with alternative object systems. The S4SXP value
  can still be used in C code but is now deprecated. Based on
  contributions from the R Consortium's Object-Oriented Programming
  Working Group.

• New function pow1p(x,y) for accurate (1+x)^y.

• mkCharLenCE was incorrectly documented to take a size_t length
  but was implemented with int (and character strings in R are
  limited to 2^31 - 1 bytes).

DEPRECATED AND DEFUNCT:

• data() no longer handles zipped data from long-defunct (since R
  2.13.0) --use-zip-data installations.

• The legacy graphics devices pictex() and xfig() are now
  deprecated. They do not support recent graphics enhancements and
  their font handling is rudimentary. The intention is to retain
  them for historical interest as long as they remain somewhat
  functional.

• Support for encoding = "MacRoman" has been removed from the pdf()
  and postscript() devices - this was a legacy encoding supporting
  classic macOS up to 2001 and no longer has universal libiconv
  support.

• is.R() is deprecated as no other S dialect is known to be in use
  (and this could only identify historical dialects, not future
  ones).

  Further information on calls can be obtained by setting the
  environment variable _R_DEPRECATED_IS_R_ to error which turns the
  deprecation warning into an error and so by default gives a
  traceback. (This is done by R CMD check --as-cran.)

• UseMethod() no longer forwards local variables assigned in the
  generic function into method call environments before evaluating
  the method body. This makes method calls behave more like
  standard function calls and makes method code easier to analyze
  correctly.

• The twelve as.data.frame.<class>() methods which were deprecated
  only via _R_CHECK_AS_DATA_FRAME_EXPLICIT_METHOD_ and in R CMD check --as-cran are formally deprecated now in favour of calling
  as.data.frame() or as.data.frame.vector(). The deprecation
  "check" now works also when as.data.frame() is S4 generic thanks
  to Ivan Krylov.

• The default method for the directional comparison operators <, >,
  <=, and >= now signals an error when one of the operands is a
  language object, i.e. a symbol or a call.

• For terms.formula(), deprecate abb and neg.out arguments
  formally in addition to just documenting it.

BUG FIXES:

• The methods package is more robust to not being attached to the
  search path. More work needs to be done.

• pairwise.t.test() misbehaved when subgroups had 0 DF for
  variance, even with pool.sd = TRUE. (PR#18594 by Jack Berry).

• Probability distribution functions [dpq]<distrib>(x, *), but also
  bessel[IKJY](x, .) now consistently preserve attributes(x) when
  length(x) == 0, e.g., for a 2 x 0 matrix, thanks to Karolis
  Koncevičius' report PR#18509.

• Group "Summary" computations such as sum(1:3, 4, na.rm = 5, NA, 7, na.rm = LL) now give an error instead of either 17 or NN for
  LL true or false, as proposed by Ivan Krylov on the R-devel
  mailing list. (This also means it is now an error to specify
  na.rm more than once.)

• as.complex(x) now returns complex(real = x, imaginary = 0) for
  all numerical and logical x, notably also for NA or
  NA_integer_.

• Directories are now omitted by file.copy(, recursive = FALSE) and
  in file.append() (PR#17337).

• gsub() and sub() are now more robust to integer overflow when
  reporting errors caused by too large input strings (PR#18346).

• Top-level handlers are now more robust to attempts to remove a
  handler whilst handlers are running (PR#18508).

• The handling of Alt+F4 in dialogs created on Windows using
  GraphApp has been fixed (PR#13870).

• density() more consistently computes grid values for the
  FFT-based convolution, following Robert Schlicht's analysis and
  proposal in PR#18337, correcting density values typically by a
  factor of about 0.999. Argument old.coords = TRUE provides back
  compatibility.

• palette.colors() gains a name argument that defaults to FALSE
  controlling whether the vector of colours that is returned has
  names (where possible). PR#18529.

• tools::xgettext() no longer extracts the (non-translatable) class
  names from warningCondition and errorCondition calls.

• S3method(<gen>, <class>, <func>) in the NAMESPACE file now works
  (again) when <func> is visible from the namespace, e.g.,
  imported, or in base.

• getParseData(f) now also works for a function defined in the
  first of several <pkg>/R/*.R source files, thanks to Kirill
  Müller's report and Duncan Murdoch's patch in PR#16756.

• Rd \Sexpr macros with nested #ifdef conditionals were not
  processed.

• A non-blocking connection with non-default encoding (such as a
  socket) now correctly returns from readLines() after new data has
  arrived also when its EOF had been reached previously. Thanks to
  Peter Meilstrup's report on R-devel and Ivan Krylov's report and
  patch proposal in PR#18555.

• tools::checkRdContents() failed to detect empty argument
  descriptions when they spanned multiple lines, including those
  generated by prompt(). These cases are now noted by R CMD check.

• Plain-text help no longer outputs spurious colons in the
  arguments list (for multi-line \item labels in the Rd source).

• kappa() and rcond() work correctly in more cases; kappa(., norm = "2") now warns that it computes the 1-norm with (default) exact = FALSE; prompted by Mikael Jagan's quite comprehensive PR#18543.

• Rd skeletons generated by prompt() or promptData() now use a
  dummy title (so R CMD build works). tools::checkRdContents() has
  been updated to detect such template leftovers, including from
  promptPackage().

• When S4 method dispatch fails because no method was found, the
  error message now includes the signature argument names; thanks
  to Michael Chirico's proposal on the R-devel list.

• withAutoprint({ .. }) now preserves srcrefs previously lost,
  thanks to Andrew Simmons' report plus fix in PR#18572.

• transform.data.frame() no longer adjusts names; in particular,
  untransformed variables are kept as-is, including those with
  syntactically invalid names (PR#17890).

• The keep.source option for Rd \Sexpr blocks is no longer ignored.

• The formula methods for t.test() and wilcox.test() now catch when
  paired is passed, addressing PR#14359; use Pair(x1, x2) ~ 1 for a
  paired test.

• The level reported in the browser prompt was often too large. It
  now shows the number of browser contexts on the stack.

• For cbind() and rbind(), the optional deparse.level argument is
  now properly passed to methods, thanks to Mikael Jagan's PR#18579
  and comments there.

• Some error and warning messages for large ('long vector')
  matrix(v, nr, nc) and dim(m) <- d are now correct about sizes,
  using long long formatting, fixing PR#18612 (and more) reported
  by Mikael Jagan.

• readChar(useBytes = TRUE) now terminates strings even when the
  underlying connection uses extra spaces in the input buffer.
  This fixes problems with extra garbage seen with gzip
  connections, PR#18605.

• Named capture in PCRE regular expressions now works also with
  more than 127 named groups (PR#18588).

• Datetime functions are now robust against long jumps when dealing
  with internal time zone changes. This avoids confusing warnings
  about an invalid time zone, previously triggered by turning
  warnings into errors or handling them via tryCatch (PR#17966,
  PR#17780).

• Datetime functions now restore even an empty TZ environment
  variable after internal time zone changes (PR#17724). This makes
  results of datetime functions with this (typically unintentional)
  setting more predictable.

• drop.terms(*) now drops response as by default, keep.response = FALSE, fixing PR#18564 thanks to Mikael Jagan.

• dummy.coef(.) now also works for lm()-models with character
  categorical predictor variables rather than factor ones, fixing
  PR#18635 reported by Jinsong Zhao.

• formals(f) <- formals(f) now also works for a function w/o
  arguments and atomic constant body(f).

• Correct as.function(<invalid list>, .)'s error message.

• removeSource() is yet more thorough in finding and removing
  "srcref" and the other source references from parsed R language
  chunks, fixing PR#18638 thanks to Andrew Simmons.

• dgeom() is more accurate now, notably when its result is very
  small, fixing PR#18642 thanks to the proposal of Morten Welinder,
  also improving other instances where C level binom_raw(x, n, ..)
  has x == 0 or x== n.

• warning() with options(warn = 1) has improved output for
  multi-line messages.

• axis.Date() and axis.POSIXct() now respect the par("lab") setting
  for the number of pretty() intervals.

• Comparisons for language objects (which are based on deparsing)
  are now more careful about using accurate deparsed results
  (PR#18676).

• Plain-text help (Rd2txt) now correctly preserves blank lines
  following single-line \dontrun code.

• <POSIXlt>[*] no longer sets wrong "balanced" attribute, fixing
  PR#18681 thanks to Mikael Jagan.

• str(<classed-call>) now deparses the call as expected, fixing
  PR#18684, reported by Dave Slager.

• In Rd examples, code following the closing brace of a \dontrun,
  \dontshow or \donttest tag on the same line is no longer skipped
  when R CMD check runs the examples.

• as.data.frame(matrix(*, ncol=0)) now gets valid names() and
  colnames(); reported by Davis Vaughan on the R-devel list.

• Internal Mathlib function stirlerr(n) is now almost fully
  (52-bit) accurate for all n >= ~5.9 and more accurate also in the
  range 1 -- 5.9. This entails small ("after 12th decimal")
  changes in density functions, e.g., dgamma() in some regions of
  their support. The fix was partly prompted by Morten Welinder's
  PR#18640.

• Numbers like 9876543.2 are now considered non-integer by Mathlib
  internal R_nonint(), amending original fix of PR#15734.

• Rd comment lines no longer cause broken paragraphs in the
  rendered PDF and plain-text help. In code blocks, pure comment
  lines (starting with %) no longer produce an empty line.

• xtabs(Freq ~ .) now consistently defaults to na.action = na.pass,
  using na.rm = FALSE (added as an argument) when summing over Freq
  (PR#17770).

• tools::testInstalledPackage() is no longer silent about failures
  from running examples or tests and its return code no longer
  ignores failures from checking vignettes.