% -*- coding: utf-8 -*-
\newcommand{\Rlogo}{\if{html}{\figure{../../html/Rlogo.svg}{options: class="toplogo" alt="[R logo]"}}\if{latex}{\figure{Rlogo.pdf}{options: width=0.5in}}}

\name{NEWS}
\title{R News}
\encoding{UTF-8}

\section{\Rlogo CHANGES IN 4.0.0}{
  \subsection{SIGNIFICANT USER-VISIBLE CHANGES}{
    \itemize{
      \item Packages need to be (re-)installed under this version
      (4.0.0) of \R.

      \item \code{matrix} objects now also inherit from class
      \code{"array"}, so e.g., \code{class(diag(1))} is \code{c("matrix",
	"array")}.  This invalidates code incorrectly assuming that
      \code{class(matrix_obj))} has length one.

      S3 methods for class \code{"array"} are now dispatched for
      \code{matrix} objects.

      \item There is a new syntax for specifying \emph{raw} character
      constants similar to the one used in C++: \code{r"(...)"} with
      \code{...} any character sequence not containing the sequence
      \samp{)"}. %}) \samp{*}: Rd-parser bug for \code{*}"
      This makes it easier to write strings that contain backslashes or
      both single and double quotes.  For more details see \code{?Quotes}.

      \item \R now uses a \samp{stringsAsFactors = FALSE} default, and
      hence by default no longer converts strings to factors in calls
      to \code{data.frame()} and \code{read.table()}.

      A large number of packages relied on the previous behaviour and
      so have needed/will need updating.

      \item The \code{plot()} S3 generic function is now in package
      \pkg{base} rather than package \pkg{graphics}, as it is
      reasonable to have methods that do not use the \pkg{graphics}
      package.  The generic is currently re-exported from the
      \pkg{graphics} namespace to allow packages importing it from
      there to continue working, but this may change in future.

      Packages which define S4 generics for \code{plot()} should be
      re-installed and package code using such generics from other
      packages needs to ensure that they are imported rather than rely
      on their being looked for on the search path (as in a namespace,
      the base namespace has precedence over the search path).
    }
  }

  \subsection{REFERENCE COUNTING}{
    \itemize{
       \item Reference counting is now used instead of the \code{NAMED}
       mechanism for determining when objects can be safely mutated in
       base C code.  This reduces the need for copying in some cases and
       should allow further optimizations in the future.  It should
       help make the internal code easier to maintain.

       This change is expected to have almost no impact on packages
       using supported coding practices in their C/C++ code.
     }
   }

  \subsection{MIGRATION TO PCRE2}{
    \itemize{
       \item This version of \R is built against the PCRE2 library for
       Perl-like regular expressions, if available.  (On non-Windows
       platforms PCRE1 can optionally be used if PCRE2 is not available
       at build time.)  The version of PCRE in use can be obtained
       \emph{via} \code{extSoftVersion()}: PCRE1 (formerly known as
       \sQuote{PCRE}) has versions <= 8, PCRE2 versions >= 10.

       \item Making PCRE2 available when building \R from source is
       strongly recommended (preferably version 10.30 or later) as PCRE1
       is no longer developed: version 8.44 is
       \sQuote{likely to be the final release}.
       % https://lists.exim.org/lurker/message/20200212.174850.6edd8277.en.html

       \item PCRE2 reports errors for some regular expressions that were
       accepted by PCRE1.  A hyphen now has to be escaped in a character
       class to be interpreted as a literal (unless first or last in the
       class definition).  \samp{\\R}, \samp{\\B} and \samp{\\X} are no
       longer allowed in character classes (PCRE1 treated these as
       literals).

       \item Option \code{PCRE_study} is no longer used with PCRE2, and is
       reported as \code{FALSE} when that is in use.
     }
  }

  \subsection{NEW FEATURES}{
    \itemize{
      \item \code{assertError()} and \code{assertWarning()} (in package
      \pkg{tools}) can now check for \emph{specific} error or warning
      classes \emph{via} the new optional second argument \code{classes}
      (which is not back compatible with previous use of an unnamed second
      argument).

      \item \code{DF2formula()}, the utility for the data frame method of
      \code{formula()}, now works without parsing and explicit evaluation,
      starting from Suharto Anggono's suggestion in \PR{17555}.

      \item \code{approxfun()} and \code{approx()} gain a new argument
      \code{na.rm} defaulting to true.  If set to false, missing
      \code{y} values now propagate into the interpolated values.

      \item Long vectors are now supported as the \code{seq} argument of
      a \code{for()} loop.

      \item \code{str(x)} gets a new \code{deparse.lines} option with a
      default to speed it up when \code{x} is a large \code{call} object.

      \item The internal traceback object produced when an error is
      signalled (\code{.Traceback}), now contains the \code{call}s
      rather than the \emph{\code{deparse()}d} calls, deferring the
      deparsing to the user-level functions \code{.traceback()} and
      \code{traceback()}.  This fulfils the wish of \PR{17580}, reported
      including two patch proposals by Brodie Gaslam.

      \item \code{data.matrix()} now converts character columns to
      factors and from this to integers.

      \item \code{package.skeleton()} now explicitly lists all exports
      in the \file{NAMESPACE} file.

      \item New function \code{.S3method()} to register S3 methods in R
      scripts.

      \item \code{file.path()} has some support for file paths not in
      the session encoding, e.g.\sspace{}with UTF-8 inputs in a
      non-UTF-8 locale the output is marked as UTF-8.

      \item Most functions with file-path inputs will give an explicit
      error if a file-path input in a marked encoding cannot be
      translated (to the native encoding or in some cases on Windows to
      UTF-8), rather than translate to a different file path using
      escapes.  Some (such as \code{dir.exists()}, \code{file.exists()},
      \code{file.access()}, \code{file.info()}, \code{list.files()},
      \code{normalizePath()} and \code{path.expand()}) treat this like
      any other non-existent file, often with a warning.

      \item There is a new help document accessed by
      \code{help("file path encoding")} detailing how file paths with
      marked encodings are handled.

      \item New function \code{list2DF()} for creating data frames from
      lists of variables.

      \item \code{iconv()} has a new option \code{sub = "Unicode"} to
      translate UTF-8 input invalid in the \samp{to} encoding using
      \samp{<U+xxxx>} escapes.

      \item There is a new function \code{infoRDS()} providing information
      about the serialization format of a serialized object.

      \item S3 method lookup now by default skips the elements of the
      search path between the global and base environments.

      \item Added an argument \code{add_datalist(*, small.size = 0)} to
      allow the creation of a \file{data/datalist} file even when the
      total size of the data sets is small.

      \item The backquote function \code{bquote()} has a new argument
      \code{splice} to enable splicing a computed list of values into an
      expression, like \code{,@} in LISP's backquote.

      \item The formula interface to \code{t.test()} and
      \code{wilcox.test()} has been extended to handle one-sample and
      paired tests.

      \item The \code{palette()} function has a new default set of
      colours (which are less saturated and have better accessibility
      properties).  There are also some new built-in palettes, which
      are listed by the new \code{palette.pals()} function.  These
      include the old default palette under the name \code{"R3"}. Finally,
      the new \code{palette.colors()} function allows a subset of
      colours to be selected from any of the built-in palettes.

      \item \code{n2mfrow()} gains an option \code{asp = 1} to specify the
      aspect ratio, fulfilling the wish and extending the proposal of
      Michael Chirico in \PR{17648}.

      \item For \code{head(x, n)} and \code{tail()} the default and
      other S3 methods notably for \emph{vector} \code{n}, e.g. to get a
      \dQuote{corner} of a matrix, has been extended to \code{array}'s
      of higher dimension thanks to the patch proposal by Gabe Becker in
      \PR{17652}.  Consequently, optional argument \code{addrownums} is
      deprecated and replaced by the (more general) argument
      \code{keepnums}.  An invalid second argument \code{n} now leads
      to typically more easily readable error messages.

      \item New function \code{.class2()} provides the full character
      vector of class names used for S3 method dispatch.

      \item Printing \code{methods(..)} now uses a new \code{format()} method.

      \item \code{sort.list(x)} now works for non-atomic objects
      \code{x} and \code{method = "auto"} (the default) or
      \code{"radix"} in cases \code{order(x)} works.

      \item Where they are available, \code{writeBin()} allows
      long vectors.% (those of length \eqn{\ge 2^{31}}{>= 2^31}).

      \item New function \code{deparse1()} produces one string, wrapping
      \code{deparse()}, to be used typically in
      \code{deparse1(substitute(*))}, e.g., to fix \PR{17671}.

      \item \code{wilcox.test()} enhancements: In the (non-paired)
      two-sample case, \code{Inf} values are treated as very large for
      robustness consistency.  If exact computations are used, the
      result now has \code{"exact"} in the \code{method} element of its
      return value.  New arguments \code{tol.root} and
      \code{digits.rank} where the latter may be used for stability to
      treat very close numbers as ties.

      \item \code{readBin()} and \code{writeBin()} now report an error
      for an invalid \code{endian} value.  The affected code needs to be
      fixed with care as the old undocumented behavior was to swap
      endian-ness in such cases.

      \item \code{sequence()} is now an S3 generic with an internally
      implemented default method, and gains arguments to generate more
      complex sequences.  Based on code from the \pkg{S4Vectors}
      Bioconductor package and the advice of Hervé Pagès.

      \item \code{print()}'s default method and many other methods (by
      calling the default eventually and passing \code{...}) now make
      use of a new optional \code{width} argument, avoiding the need for
      the user to set and reset \code{options("width")}.

      \item \code{memDecompress()} supports the RFC 1952 format
      (e.g.\sspace{}in-memory copies of \command{gzip}-compressed files)
      as well as RFC 1950.

      \item \code{memCompress()} and \code{memDecompress()} support long
      raw vectors for types \code{"gzip"} and \code{"zx"}.

      \item \code{sweep()} and \code{slice.index()} can now use names
      of dimnames for their \code{MARGIN} argument (\code{apply} has
      had this for almost a decade).

      \item New function \code{proportions()} and
      \code{marginSums()}. These should replace the unfortunately named
      \code{prop.table()} and \code{margin.table()}. They are drop-in
      replacements, but also add named-margin functionality. The old
      function names are retained as aliases for back-compatibility.

      \item Functions \code{rbinom()}, \code{rgeom()}, \code{rhyper()},
      \code{rpois()}, \code{rnbinom(),} \code{rsignrank()} and
      \code{rwilcox()} which have returned integer since \R 3.0.0 and hence
      \code{NA} when the numbers would have been outside the integer range,
      now return double vectors (without NAs, typically) in these cases.

      \item \code{matplot(x,y)} (and hence \code{matlines()} and
      \code{matpoints()}) now call the corresponding methods of
      \code{plot()} and \code{lines()}, e.g, when \code{x} is a
      \code{"Date"} or \code{"POSIXct"} object; prompted by Spencer Graves'
      suggestion.

      \item \code{stopifnot()} now allows customizing error messages via
      argument names, thanks to a patch proposal by Neal Fultz in \PR{17688}.

      \item  \code{unlink()} gains a new argument \code{expand} to disable
      wildcard and tilde expansion.  Elements of \code{x} of value
      \code{"~"} are now ignored.

      \item \code{mle()} in the \code{stats4} package has had its
      interface extended so that arguments to the negative
      log-likelihood function can be one or more vectors, with similar
      conventions applying to bounds, start values, and parameter values
      to be kept fixed.  This required a minor extension to class
      \code{"mle"}, so saved objects from earlier versions may need to
      be recomputed.

      \item The default for \code{pdf()} is now
      \code{useDingbats = FALSE}.

      \item The default fill colour for \code{hist()} and
      \code{boxplot()} is now \code{col = "lightgray"}.

      \item The default order of the levels on the y-axis for
      \code{spineplot()} and \code{cdplot()} has been reversed.

      \item If the \env{R_ALWAYS_INSTALL_TESTS} environment variable is
      set to a true value, \command{R CMD INSTALL} behaves as if
      the \option{--install-tests} option is always specified. Thanks to
      Reinhold Koch for the suggestion.

      \item New function \code{R_user_dir()} in package \pkg{tools}
      suggests paths appropriate for storing \R-related user-specific
      data, configuration and cache files.

      \item \code{capabilities()} gains a new logical option \code{Xchk}
      to avoid warnings about X11-related capabilities.

      \item The internal implementation of \pkg{grid} units has changed,
      but the only visible effects at user-level should be
      \itemize{
	\item a slightly different print format for some units
	(especially unit arithmetic),
	\item faster performance (for unit operations) and
	\item two new functions \code{unitType()} and \code{unit.psum()}.
      }
      Based on code contributed by Thomas Lin Pedersen.

      \item When internal dispatch for \code{rep.int()}
      and \code{rep_len()} fails, there is an attempt to dispatch on the
      equivalent call to \code{rep()}.

      \item Object \code{.Machine} now contains new \code{longdouble.*}
      entries (when \R uses long doubles internally).

      \item \code{news()} has been enhanced to cover the news on \R 3.x
      and 2.x.

       \item For consistency, \code{N <- NULL; N[[1]] <- val} now turns
       \code{N} into a \code{list} also when \code{val)} has length one.
       This enables \code{dimnames(r1)[[1]] <- "R1"} for a 1-row matrix
       \code{r1}, fixing \PR{17719} reported by Serguei Sokol.

       \item \code{deparse(..)}, \code{dump(..)}, and \code{dput(x,
	 control = "all")} now include control option \code{"digits17"}
       which typically ensures 1:1 invertibility.  New option
       \code{control = "exact"} ensures numeric exact invertibility
       via \code{"hexDigits"}.

      \item When loading data sets via \code{read.table()},
       \code{data()} now uses \samp{LC_COLLATE=C} to ensure
       locale-independent results for possible string-to-factor
       conversions.

      \item A server socket connection, a new connection type
      representing a listening server socket, is created via
      \code{serverSocket()} and can accept multiple socket connections
      via \code{socketAccept()}.

      \item New function \code{socketTimeout()} changes the connection
      timeout of a socket connection.

      \item The time needed to start a homogeneous \samp{PSOCK} cluster on
      \samp{localhost} with many nodes has been significantly reduced
      (package \pkg{parallel}).

      \item New \code{globalCallingHandlers()} function to establish
      global condition handlers.  This allows registering default
      handlers for specific condition classes. Developed in
      collaboration with Lionel Henry.

      \item New function \code{tryInvokeRestart()} to invoke a specified
      restart if one is available and return without signaling an error
      if no such restart is found.  Contributed by Lionel Henry in
      \PR{17598}.

      \item \code{str(x)} now shows the length of \code{attributes} in some
      cases for a data frame \code{x}.

      \item \code{Rprof()} gains a new argument \code{filter.callframes}
      to request that intervening call frames due to lazy evaluation or
      explicit \code{eval()} calls be omitted from the recorded profile
      data.  Contributed by Lionel Henry in \PR{17595}.

      \item The handling of \code{$\{FOO-bar\}} and \code{$\{FOO:-bar\}}
      in \file{Renviron} files now follows POSIX shells (at least on a
      Unix-alike), so the first treats empty environment variables as
      set and the second does not.  Previously both ignored empty
      variables.  There are several uses of the first form in
      \file{etc/Renviron}.

      \item New \code{classes} argument for \code{suppressWarnings()}
      and \code{suppressMessages()} to selectively suppress only
      warnings or messages that inherit from particular classes.
      Based on patch from Lionel Henry submitted with \PR{17619}.

      \item New function \code{activeBindingFunction()} retrieves the
      function of an active binding.
    }
  }% end{ NEW FEATURES }

  \subsection{Windows}{
    \itemize{
      \item \command{Rterm} now works also when invoked from MSYS2
      terminals.  Line editing is possible when command \command{winpty} is
      installed.

      \item \code{normalizePath()} now resolves symbolic links and
      normalizes case of long names of path elements in case-insensitive
      folders (\PR{17165}).

      \item \code{md5sum()} supports UTF-8 file names with characters
      that cannot be translated to the native encoding (\PR{17633}).

      \item \command{Rterm} gains a new option \option{--workspace} to
      specify the workspace to be restored.  This allows equals to be
      part of the name when opening \emph{via} Windows file associations
      (reported by Christian Asseburg).

      \item \command{Rterm} now accepts \code{ALT+xxx} sequences also
      with NumLock on.  Tilde can be pasted with an Italian keyboard
      (\PR{17679}).

      \item \R falls back to copying when junction creation fails during
      package checking (patch from Duncan Murdoch).
   }
  }

  \subsection{DEPRECATED AND DEFUNCT}{
    \itemize{
      \item Make macro \samp{F77_VISIBILITY} has been removed
      and replaced by \samp{F_VISIBILITY}.

      \item Make macros \samp{F77}, \samp{FCPIFCPLAGS} and
      \samp{SHLIB_OPENMP_FCFLAGS} have been removed and replaced by
      \samp{FC}, \samp{FPICFLAGS} and \samp{SHLIB_OPENMP_FFLAGS}
      respectively.  (Most \command{make} programs will set \samp{F77}
      to the value of \samp{FC}, which is set for package compilation.
      But portable code should not rely on this.)

      \item The deprecated support for specifying C++98 for package
      installation has been removed.

      \item \command{R CMD config} no longer knows about the
      unused settings \samp{F77} and \samp{FCPIFCPLAGS}, nor
      \samp{CXX98} and similar.

      \item Either PCRE2 or PCRE1 >= 8.32 (Nov 2012) is required: the
      deprecated provision for 8.20--8.31 has been removed.
    }
  }

  \subsection{C-LEVEL FACILITIES}{
    \itemize{
      \item \code{installChar} is now remapped in \file{Rinternals.h} to
      \code{installTrChar}, of which it has been a wrapper since
      \R\sspace{}3.6.0.  Neither are part of the API, but packages using
      \code{installChar} can replace it if they depend on \samp{R >= 3.6.2}.

      \item Header \file{R_ext/Print.h} defines \samp{R_USE_C99_IN_CXX}
      and hence exposes \code{Rvprintf} and \code{REvprintf} if used
      with a C++11 (or later) compiler.

      \item There are new Fortran subroutines \code{dblepr1},
      \code{realpr1} and \code{intpr1} to print a scalar variable
      (\command{gfortran} 10 enforces the distinction between scalars
      and length-one arrays).  Also \code{labelpr} to print just a label.

      \item \code{R_withCallingErrorHandler} is now available for
      establishing a calling handler in C code for conditions inheriting
      from class \code{error}.
    }
  }

  \subsection{INSTALLATION on a UNIX-ALIKE}{
    \itemize{
      \item User-set \samp{DEFS} (e.g., in \file{config.site}) is now
      used for compiling packages (including base packages).

      \item There is a new variant option \option{--enable-lto=check}
      for checking consistency of BLAS/LAPACK/LINPACK calls --- see
      \sQuote{Writing R Extensions}.

      \item A C++ compiler default is set only if the C++11 standard is
      supported: it no longer falls back to C++98.

      \item PCRE2 is used if available.  To make use of PCRE1 if PCRE2
      is unavailable, configure with option \option{--with-pcre1}.

      \item The minimum required version of \code{libcurl} is now 7.28.0
      (Oct 2012).

      \item New make target \command{distcheck} checks
      \itemize{
	\item \R can be rebuilt from the tarball created by
	\command{make dist},
	\item the build from the tarball passes \command{make check-all},
	\item the build installs and uninstalls,
	\item the source files are properly cleaned by \command{make distclean}.
      }
    }
  }

  \subsection{UTILITIES}{
    \itemize{
      \item \command{R --help} now mentions the option \command{--no-echo}
      (renamed from \command{--slave}) and its previously undocumented
      short form \command{-s}.

      \item \command{R CMD check} now optionally checks
      \command{configure} and \command{cleanup} scripts for
      non-Bourne-shell code (\sQuote{bashisms}).

      \item \command{R CMD check --as-cran} now runs \code{\\donttest}
      examples (which are run by \code{example()}) instead of
      instructing the tester to do so.  This can be temporarily
      circumvented during development by setting environment variable
      \env{_R_CHECK_DONTTEST_EXAMPLES_} to a false value.
    }
  }

  \subsection{PACKAGE INSTALLATION}{
    \itemize{
      \item There is the beginnings of support for the recently approved
      C++20 standard, specified analogously to C++14 and C++17.  There is
      currently only limited support for this in compilers, with flags
      such as \option{-std=c++20} and \option{-std=c++2a}.  For the time
      being the \command{configure} test is of accepting one of these
      flags and compiling C++17 code.
    }
  }

  \subsection{BUG FIXES}{
    \itemize{
      \item \code{formula(x)} with \code{length(x) > 1} character vectors,
      is deprecated now.  Such use has been rare, and has \sQuote{worked}
      as expected in some cases only.  In other cases, wrong \code{x} have
      silently been truncated, not detecting previous errors.

      \item Long-standing issue where the X11 device could lose events
      shortly after startup has been addressed (\PR{16702}).

      \item The \code{data.frame} method for \code{rbind()} no longer
      drops \code{<NA>} levels from factor columns by default
      (\PR{17562}).

      \item \code{available.packages()} and hence \code{install.packages()}
      now pass their \code{...} argument to \code{download.file()},
      fulfilling the wish of \PR{17532}; subsequently,
      \code{available.packages()} gets new argument \code{quiet}, solving
      \PR{17573}.

      \item \code{stopifnot()} gets new argument \code{exprObject} to allow
      an \R object of class \code{expression} (or other \sQuote{language})
      to work more consistently, thanks to suggestions by Suharto Anggono.

      \item \code{conformMethod()} now works correctly in cases containing
      a \dQuote{\code{&&} logic} bug, reported by Henrik Bengtsson.  It now
      creates methods with \code{"missing"} entries in the signature.
      Consequently, \code{rematchDefinition()} is amended to use
      appropriate \code{.local()} calls with named arguments where needed.

      \item \code{format.default(*, scientific = FALSE)} now corresponds to
      a practically most extreme \code{options(scipen = n)} setting rather
      than arbitrary \code{n = 100}.

      \item \code{format(as.symbol("foo"))} now works (returning \code{"foo"}).

      \item \code{postscript(.., title = *)} now signals an error when
      the title string contains a character which would produce corrupt
      PostScript, thanks to \PR{17607} by Daisuko Ogawa.

      \item Certain \code{Ops} (notably comparison such as \code{==}) now
      also work for 0-length data frames, after reports by Hilmar Berger.

      \item \code{methods(class = class(glm(..)))} now warns more usefully
      and only once.

      \item \code{write.dcf()} no longer mangles field names (\PR{17589}).

      \item Primitive replacement functions no longer mutate a
      referenced first argument when used outside of a complex
      assignment context.

      \item  A better error message for \code{contour(*, levels = Inf)}.

      \item The return value of \code{contourLines()} is no longer
      \code{invisible()}.

      \item The Fortran code for calculating the \code{coefficients}
      component in \code{lm.influence()} was very inefficient. It has
      (for now) been replaced with much faster \R{} code (\PR{17624}).

      \item \code{cm.colors(n)} \emph{etc} no longer append the code for
      \code{alpha = 1}, \code{"FF"}, to all colors.  Hence all eight
      \code{*.colors()} functions and \code{rainbow()} behave
      consistently and have the same non-explicit default (\PR{17659}).

      \item \code{dnorm} had a problematic corner case with \code{sd ==
      -Inf} or negative \code{sd} which was not flagged as an error in
      all cases. Thanks to Stephen D. Weigand for reporting and Wang
      Jiefei for analyzing this; similar change has been made in
      \code{dlnorm()}.

      \item The optional \code{iter.smooth} argument of
      \code{plot.lm()}, (the \code{plot()} method for \code{lm} and
      \code{glm} fits) now defaults to \code{0} for all \code{glm} fits.
      Especially for binary observations with high or low fitted
      probabilities, this effectively deleted all observations of 1 or
      0.  Also, the type of residuals used in the \code{glm} case has
      been switched to \code{"pearson"} since deviance residuals do not
      in general have approximately zero mean.

      \item In \code{plot.lm}, Cook's distance was computed from unweighted
      residuals, leading to inconsistencies.  Replaced with usual weighted
      version. (\PR{16056})

      \item Time-series \code{ts(*, start, end, frequency)} with
      fractional \code{frequency} are supported more consistently;
      thanks to a report from Johann Kleinbub and analysis and patch by
      Duncan Murdoch in \PR{17669}.

      \item In case of errors \code{mcmapply()} now preserves attributes
      of returned \code{"try-error"} objects and avoids simplification,
      overriding \code{SIMPLIFY} to \code{FALSE}. (\PR{17653})

      \item \code{as.difftime()} gets new optional \code{tz = "UTC"}
      argument which should fix behaviour during
      daylight-savings-changeover days, fixing \PR{16764}, thanks to
      proposals and analysis by Johannes Ranke and Kirill Müller.

      \item \code{round()} does a better job of rounding
      \emph{\dQuote{to nearest}} by \emph{measuring} and
      \emph{\dQuote{to even}}; thanks to a careful algorithm originally
      prompted by the report from Adam Wheeler and then others, in
      \PR{17668}.  \cr \code{round(x, dig)} for \emph{negative} digits
      is much more rational now, notably for large \eqn{|dig|}.

      \item Inheritance information on S4 classes is maintained more
      consistently, particularly in the case of class unions (in part
      due to \PR{17596} and a report from Ezra Tucker).

      \item \code{is()} behaves more robustly when its argument
      \code{class2} is a \code{classRepresentation} object.

      \item The warning message when attempting to export an nonexistent
      class is now more readable; thanks to Thierry Onkelinx for
      recognizing the problem.

      \item \code{choose()} misbehaved in corner cases where it switched
      \code{n - k} for \code{k} and \code{n} was only \emph{nearly} integer
      (report from Erik Scott Wright).

      \item \code{mle()} in the \code{stats4} package had problems
      combining use of box constraints and fixed starting values (in
      particular, confidence intervals were affected).

      \item Operator \code{?} now has lower precedence than \code{=} to
      work as documented, so \code{=} behaves like \code{<-} in help
      expressions (\PR{16710}).

      \item \code{smoothEnds(x)} now returns \code{integer} type in
      \emph{both} cases when \code{x} is \code{integer}, thanks to a
      report and proposal by Bill Dunlap \PR{17693}.

      \item The \pkg{methods} package does a better job of tracking
      inheritance relationships across packages.

      \item \code{norm(diag(c(1, NA)), "2")} now works.

      \item \code{subset()} had problems with 0-col dataframes (reported
      by Bill Dunlap, \PR{17721}).

      \item Several cases of integer overflow detected by the
      \sQuote{undefined behaviour sanitizer} of \command{clang} 10 have
      been circumvented.  One in \code{rhyper()} may change the
      generated value for large input values.

      \item \code{dotchart()} now places the y-axis label (\code{ylab})
      much better, not overplotting labels, thanks to a report and
      suggestion by Alexey Shipunov.

      %% detected by clang-ASAN in examples/tests of adehabitatHR.
      \item A rare C-level array overflow in \code{chull()} has been
      worked around.

      \item Some invalid specifications of the day-of-the-year
      (\emph{via} \code{\%j}, e.g.\sspace{}day 366 in 2017) or week
      plus day-of-the-week are now detected by \code{strptime()}.
      They now return \code{NA} but give a warning as they may have
      given random results or corrupted memory in earlier versions of \R.

      \item \code{socketConnection(server = FALSE)} now respects the
      connection timeout also on Linux.

      \item \code{socketConnection(server = FALSE)} no longer leaks a
      connection that is available right away without waiting (e.g. on
      \samp{localhost}).

      \item Socket connections are now robust against spurious readability
      and spurious availability of an incoming connection.

      \item \code{blocking = FALSE} is now respected also on the server side
      of a socket connection, allowing non-blocking read operations.

      \item \code{anova.glm()} and \code{anova.glmlist()} computed
      incorrect score (Rao) tests in no-intercept cases. (André
      Gillibert, \PR{17734})

      \item \code{summaryRprof()} now should work correctly for the
      \code{Rprof(*, memory.profiling=TRUE)} case with small chunk size (and
      \code{"tseries"} or similar) thanks to a patch proposal by Benjamin
      Tyner, in \PR{15886}.

      \item \code{xgettext()} ignores strings passed to
      \code{ngettext()}, since the latter is handled by
      \code{xngettext()}. Thanks to Daniele Medri for the report and all
      the recent work he has done on the Italian translations.

      \item \code{data(package = P)} for \code{P} in \pkg{base} and
      \pkg{stats} no longer reports the data sets from package
      \pkg{datasets} (which it did for back compatibility for 16 years),
      fixing \PR{17730}.
    }
  }

  \subsection{BUG FIXES (Windows)}{
    \itemize{
      \item \code{Sys.glob()} now supports all characters
      from the Unicode Basic Multilingual Plane, no longer corrupting
      some (less commonly used) characters (\PR{17638}).

      \item \command{Rterm} now correctly displays
      multi-byte-coded characters representable in the current native
      encoding (at least on Windows 10 they were sometimes omitted,
      \PR{17632}).

      \item \code{scan()} issues with UTF-8 data when running in a DBCS
      locale have been resolved (\PR{16520}, \PR{16584}).

      \item \command{RTerm} now accepts enhanced/arrow keys also with
      ConPTY.

      \item \R can can now be started \emph{via} the launcher icon in a
      user documents directory whose path is not representable in the
      system encoding.

      \item \code{socketConnection(server = FALSE)} now returns instantly
      also on Windows when connection failure is signalled.

      \item Problems with UTF-16 surrogate pairs have been fixed in
      several functions, including \command{tolower()} and
      \command{toupper()} (\PR{17645}).
    }
  }
}

\section{CHANGES in previous versions}{
  \itemize{
    \item Older news can be found in text format in files
    \ifelse{html}{\href{../NEWS.0}{NEWS.0}, \href{../NEWS.1}{NEWS.1},
      \href{../NEWS.2}{NEWS.2} and
      and \href{../NEWS.3}{NEWS.3}}{\file{NEWS.0}, \file{NEWS.1},
      \file{NEWS.2} and \file{NEWS.3}}
    in the \file{doc} directory.  News in HTML format for
    \R  versions 3.x and from 2.10.0 to 2.15.3 is available at
    \ifelse{html}{\url{NEWS.3.html}}{\file{doc/html/NEWS.3.html}}
    and
    \ifelse{html}{\url{NEWS.2.html}}{\file{doc/html/NEWS.2.html}}.
  }
}