R 3.5.0 (codename “Joy in Playing”) was released yesterday. You can get the latest binaries version from here. (or the .tar.gz source code from here).

This is a major release with many new features and bug fixes, the full list is provided below.

Upgrading R on Windows and Mac

If you are using Windows you can easily upgrade to the latest version of R using the installr package. Simply run the following code in Rgui:

install.packages("installr") # install 
setInternet2(TRUE) # only for R versions older than 3.3.0
installr::updateR() # updating R.
# If you wish it to go faster, run: installr::updateR(T)

Running “updateR()” will detect if there is a new R version available, and if so it will download+install it (etc.). There is also a step by step tutorial (with screenshots) on how to upgrade R on Windows, using the installr package. If you only see the option to upgrade to an older version of R, then change your mirror or try again in a few hours (it usually take around 24 hours for all CRAN mirrors to get the latest version of R).

If you are using Mac you can easily upgrade to the latest version of R using Andrea Cirillo’s updateR package. The package is not on CRAN, so you’ll need to run the following code in Rgui:

install.packages("devtools")
devtools::install_github("AndreaCirilloAC/updateR")
updateR(admin_password = "PASSWORD") # Where "PASSWORD" stands for your system password

Later this year Andrea and I intend to merge the updateR package into installr so that the updateR function will work seamlessly in both Windows and Mac. Stay tuned

CHANGES IN R 3.5.0

SIGNIFICANT USER-VISIBLE CHANGES

• All packages are by default byte-compiled on installation. This makes the installed packages larger (usually marginally so) and may affect the format of messages and tracebacks (which often exclude .Call and similar).

NEW FEATURES

• factor() now uses order() to sort its levels, rather than sort.list(). This allows factor() to support custom vector-like objects if methods for the appropriate generics are defined. It has the side effect of making factor() succeed on empty or length-one non-atomic vector(-like) types (e.g., "list"), where it failed before.
• diag() gets an optional names argument: this may require updates to packages defining S4 methods for it.
• chooseCRANmirror() and chooseBioCmirror() no longer have a useHTTPS argument, not needed now all R builds support https:// downloads.
• New summary() method for warnings() with a (somewhat experimental) print() method.
• (methods package.) .self is now automatically registered as a global variable when registering a reference class method.
• tempdir(check = TRUE) recreates the tempdir() directory if it is no longer valid (e.g. because some other process has cleaned up the ‘/tmp’ directory).
• New askYesNo() function and "askYesNo" option to ask the user binary response questions in a customizable but consistent way. (Suggestion of PR#17242.)
• New low level utilities ...elt(n) and ...length() for working with ... parts inside a function.
• isTRUE() is more tolerant and now true in

     x <- rlnorm(99)
     isTRUE(median(x) == quantile(x)["50%"])

  New function isFALSE() defined analogously to isTRUE().

• The default symbol table size has been increased from 4119 to 49157; this may improve the performance of symbol resolution when many packages are loaded. (Suggested by Jim Hester.)
• line() gets a new option iter = 1.
• Reading from connections in text mode is buffered, significantly improving the performance of readLines(), as well as scan() and read.table(), at least when specifying colClasses.
• order() is smarter about picking a default sort method when its arguments are objects.
• available.packages() has two new arguments which control if the values from the per-session repository cache are used (default true, as before) and if so how old cached values can be to be used (default one hour).These arguments can be passed from install.packages(), update.packages() and functions calling that: to enable this available.packages(), packageStatus() anddownload.file() gain a ... argument.
• packageStatus()‘s upgrade() method no longer ignores its ... argument but passes it to install.packages().
• installed.packages() gains a ... argument to allow arguments (including noCache) to be passed from new.packages(), old.packages(), update.packages() and packageStatus().
• factor(x, levels, labels) now allows duplicated labels (not duplicated levels!). Hence you can map different values of x to the same level directly.
• Attempting to use names<-() on an S4 derivative of a basic type no longer emits a warning.
• The list method of within() gains an option keepAttrs = FALSE for some speed-up.
• system() and system2() now allow the specification of a maximum elapsed time (‘timeout’).
• debug() supports debugging of methods on any object of S4 class "genericFunction", including group generics.
• Attempting to increase the length of a variable containing NULL using length()<- still has no effect on the target variable, but now triggers a warning.
• type.convert() becomes a generic function, with additional methods that operate recursively over list and data.frame objects. Courtesy of Arni Magnusson (PR#17269).
• lower.tri(x) and upper.tri(x) only needing dim(x) now work via new functions .row() and .col(), so no longer call as.matrix() by default in order to work efficiently for all kind of matrix-like objects.
• print() methods for "xgettext" and "xngettext" now use encodeString() which keeps, e.g. "\n", visible. (Wish of PR#17298.)
• package.skeleton() gains an optional encoding argument.
• approx(), spline(), splinefun() and approxfun() also work for long vectors.
• deparse() and dump() are more useful for S4 objects, dput() now using the same internal C code instead of its previous imperfect workaround R code. S4 objects now typically deparse perfectly, i.e., can be recreated identically from deparsed code.dput(), deparse() and dump() now print the names() information only once, using the more readable (tag = value) syntax, notably for list()s, i.e., including data frames.

  These functions gain a new control option "niceNames" (see .deparseOpts()), which when set (as by default) also uses the (tag = value) syntax for atomic vectors. On the other hand, without deparse options "showAttributes" and "niceNames", names are no longer shown also for lists. as.character(list( c (one = 1))) now includes the name, as as.character(list(list(one = 1))) has always done.

  m:n now also deparses nicely when m > n.

  The "quoteExpressions" option, also part of "all", no longer quote()s formulas as that may not re-parse identically. (PR#17378)

• If the option setWidthOnResize is set and TRUE, R run in a terminal using a recent readline library will set the width option when the terminal is resized. Suggested by Ralf Goertz.
• If multiple on.exit() expressions are set using add = TRUE then all expressions will now be run even if one signals an error.
• mclapply() gets an option affinity.list which allows more efficient execution with heterogeneous processors, thanks to Helena Kotthaus.
• The character methods for as.Date() and as.POSIXlt() are more flexible via new arguments tryFormats and optional: see their help pages.
• on.exit() gains an optional argument after with default TRUE. Using after = FALSE with add = TRUE adds an exit expression before any existing ones. This way the expressions are run in a first-in last-out fashion. (From Lionel Henry.)
• On Windows, file.rename() internally retries the operation in case of error to attempt to recover from possible anti-virus interference.
• Command line completion on :: now also includes lazy-loaded data.
• If the TZ environment variable is set when date-time functions are first used, it is recorded as the session default and so will be used rather than the default deduced from the OS if TZ is subsequently unset.
• There is now a [ method for class "DLLInfoList".
• glm() and glm.fit get the same singular.ok = TRUE argument that lm() has had forever. As a consequence, in glm(*, method = <your_own>), user specified methods need to accept a singular.ok argument as well.
• aspell() gains a filter for Markdown (‘.md’ and ‘.Rmd’) files.
• intToUtf8(multiple = FALSE) gains an argument to allow surrogate pairs to be interpreted.
• The maximum number of DLLs that can be loaded into R e.g. via dyn.load() has been increased up to 614 when the OS limit on the number of open files allows.
• Sys.timezone() on a Unix-alike caches the value at first use in a session: inter alia this means that setting TZ later in the session affects only the current time zone and not the system one.Sys.timezone() is now used to find the system timezone to pass to the code used when R is configured with –with-internal-tzcode.
• When tar() is used with an external command which is detected to be GNU tar or libarchive tar (aka bsdtar), a different command-line is generated to circumvent line-length limits in the shell.
• system(*, intern = FALSE), system2() (when not capturing output), file.edit() and file.show() now issue a warning when the external command cannot be executed.
• The “default” ("lm" etc) methods of vcov() have gained new optional argument complete = TRUE which makes the vcov() methods more consistent with the coef()methods in the case of singular designs. The former (back-compatible) behavior is given by vcov(*, complete = FALSE).
• coef() methods (for lm etc) also gain a complete = TRUE optional argument for consistency with vcov().
  For "aov", both coef() and vcov() methods remain back-compatibly consistent, using the other default, complete = FALSE.
• attach(*, pos = 1) is now an error instead of a warning.
• New function getDefaultCluster() in package parallel to get the default cluster set via setDefaultCluster().
• str(x) for atomic objects x now treats both cases of is.vector(x) similarly, and hence much less often prints "atomic". This is a slight non-back-compatible change producing typically both more informative and shorter output.
• write.dcf() gets optional argument useBytes.
• New, partly experimental packageDate() which tries to get a valid "Date" object from a package ‘DESCRIPTION’ file, thanks to suggestions in PR#17324.
• tools::resaveRdaFiles() gains a version argument, for use when packages should remain compatible with earlier versions of R.
• ar.yw(x) and hence by default ar(x) now work when x has NAs, mostly thanks to a patch by Pavel Krivitsky in PR#17366. The ar.yw.default()‘s AIC computations have become more efficient by using determinant().
• New warnErrList() utility (from package nlme, improved).
• By default the (arbitrary) signs of the loadings from princomp() are chosen so the first element is non-negative.
• If –default-packages is not used, then Rscript now checks the environment variable R_SCRIPT_DEFAULT_PACKAGES. If this is set, then it takes precedence over R_DEFAULT_PACKAGES. If default packages are not specified on the command line or by one of these environment variables, then Rscript now uses the same default packages as R. For now, the previous behavior of not including methods can be restored by setting the environment variable R_SCRIPT_LEGACY to yes.
• When a package is found more than once, the warning from find.package(*, verbose=TRUE) lists all library locations.
• POSIXt objects can now also be rounded or truncated to month or year.
• stopifnot() can be used alternatively via new argument exprs which is nicer and useful when testing several expressions in one call.
• The environment variable R_MAX_VSIZE can now be used to specify the maximal vector heap size. On macOS, unless specified by this environment variable, the maximal vector heap size is set to the maximum of 16GB and the available physical memory. This is to avoid having the R process killed when macOS over-commits memory.
• sum(x) and sum(x1,x2,..,x<N>) with many or long logical or integer vectors no longer overflows (and returns NA with a warning), but returns double numbers in such cases.
• Single components of "POSIXlt" objects can now be extracted and replaced via [ indexing with 2 indices.
• S3 method lookup now searches the namespace registry after the top level environment of the calling environment.
• Arithmetic sequences created by 1:n, seq_along, and the like now use compact internal representations via the ALTREP framework. Coercing integer and numeric vectors to character also now uses the ALTREP framework to defer the actual conversion until first use.
• Finalizers are now run with interrupts suspended.
• merge() gains new option no.dups and by default suffixes the second of two duplicated column names, thanks to a proposal by Scott Ritchie (and Gabe Becker).
• scale.default(x, center, scale) now also allows center or scale to be “numeric-alike”, i.e., such that as.numeric(.) coerces them correctly. This also eliminates a wrong error message in such cases.
• par*apply and par*applyLB gain an optional argument chunk.size which allows to specify the granularity of scheduling.
• Some as.data.frame() methods, notably the matrix one, are now more careful in not accepting duplicated or NA row names, and by default produce unique non-NA row names. This is based on new function .rowNamesDF(x, make.names = *) <- rNms where the logical argument make.names allows to specify how invalid row names rNms are handled. .rowNamesDF() is a “workaround” compatible default.
• R has new serialization format (version 3) which supports custom serialization of ALTREP framework objects. These objects can still be serialized in format 2, but less efficiently. Serialization format 3 also records the current native encoding of unflagged strings and converts them when de-serialized in R running under different native encoding. Format 3 comes with new serialization magic numbers (RDA3, RDB3, RDX3). Format 3 can be selected by version = 3 in save(), serialize() and saveRDS(), but format 2 remains the default for all serialization and saving of the workspace. Serialized data in format 3 cannot be read by versions of R prior to version 3.5.0.
• The "Date" and “date-time” classes "POSIXlt" and "POSIXct" now have a working `length<-` method, as wished in PR#17387.
• optim(*, control = list(warn.1d.NelderMead = FALSE)) allows to turn off the warning when applying the default "Nelder-Mead" method to 1-dimensional problems.
• matplot(.., panel.first = .) etc now work, as log becomes explicit argument and ... is passed to plot() unevaluated, as suggested by Sebastian Meyer in PR#17386.
• Interrupts can be suspended while evaluating an expression using suspendInterrupts. Subexpression can be evaluated with interrupts enabled using allowInterrupts. These functions can be used to make sure cleanup handlers cannot be interrupted.
• R 3.5.0 includes a framework that allows packages to provide alternate representations of basic R objects (ALTREP). The framework is still experimental and may undergo changes in future R releases as more experience is gained. For now, documentation is provided in https://svn.r-project.org/R/branches/ALTREP/ALTREP.html.

UTILITIES

• install.packages() for source packages now has the possibility to set a ‘timeout’ (elapsed-time limit). For serial installs this uses the timeout argument of system2(): for parallel installs it requires the timeout utility command from GNU coreutils.
• It is now possible to set ‘timeouts’ (elapsed-time limits) for most parts of R CMD check via environment variables documented in the ‘R Internals’ manual.
• The ‘BioC extra’ repository which was dropped from Bioconductor 3.6 and later has been removed from setRepositories(). This changes the mapping for 6–8 used by setRepositories(ind=).
• R CMD check now also applies the settings of environment variables _R_CHECK_SUGGESTS_ONLY_ and _R_CHECK_DEPENDS_ONLY_ to the re-building of vignettes.
• R CMD check with environment variable _R_CHECK_DEPENDS_ONLY_ set to a true value makes test-suite-management packages available and (for the time being) works around a common omission of rmarkdown from the VignetteBuilder field.

INSTALLATION on a UNIX-ALIKE

• Support for a system Java on macOS has been removed — install a fairly recent Oracle Java (see ‘R Installation and Administration’ §C.3.2).
• configure works harder to set additional flags in SAFE_FFLAGS only where necessary, and to use flags which have little or no effect on performance.In rare circumstances it may be necessary to override the setting of SAFE_FFLAGS.
• C99 functions expm1, hypot, log1p and nearbyint are now required.
• configure sets a -std flag for the C++ compiler for all supported C++ standards (e.g., -std=gnu++11 for the C++11 compiler). Previously this was not done in a few cases where the default standard passed the tests made (e.g. clang 6.0.0 for C++11).

C-LEVEL FACILITIES

• ‘Writing R Extensions’ documents macros MAYBE_REFERENCED, MAYBE_SHARED and MARK_NOT_MUTABLE that should be used by package C code instead NAMED or SET_NAMED.
• The object header layout has been changed to support merging the ALTREP branch. This requires re-installing packages that use compiled code.
• ‘Writing R Extensions’ now documents the R_tryCatch, R_tryCatchError, and R_UnwindProtect functions.
• NAMEDMAX has been raised to 3 to allow protection of intermediate results from (usually ill-advised) assignments in arguments to BUILTIN functions. Package C code usingSET_NAMED may need to be revised.

DEPRECATED AND DEFUNCT

• Sys.timezone(location = FALSE) is defunct, and is ignored (with a warning).
• methods:::bind_activation() is defunct now; it typically has been unneeded for years.The undocumented ‘hidden’ objects .__H__.cbind and .__H__.rbind in package base are deprecated (in favour of cbind and rbind).
• The declaration of pythag() in ‘Rmath.h’ has been removed — the entry point has not been provided since R 2.14.0.

BUG FIXES

• printCoefmat() now also works without column names.
• The S4 methods on Ops() for the "structure" class no longer cause infinite recursion when the structure is not an S4 object.
• nlm(f, ..) for the case where f() has a "hessian" attribute now computes LL’ = H + µI correctly. (PR#17249).
• An S4 method that “rematches” to its generic and overrides the default value of a generic formal argument to NULL no longer drops the argument from its formals.
• Rscript can now accept more than one argument given on the #! line of a script. Previously, one could only pass a single argument on the #! line in Linux.
• Connections are now written correctly with encoding "UTF-16LE". (PR#16737).
• Evaluation of ..0 now signals an error. When ..1 is used and ... is empty, the error message is more appropriate.
• (Windows mainly.) Unicode code points which require surrogate pairs in UTF-16 are now handled. All systems should properly handle surrogate pairs, even those systems that do not need to make use of them. (PR#16098)
• stopifnot(e, e2, ...) now evaluates the expressions sequentially and in case of an error or warning shows the relevant expression instead of the full stopifnot(..) call.
• path.expand() on Windows now accepts paths specified as UTF-8-encoded character strings even if not representable in the current locale. (PR#17120)
• line(x, y) now correctly computes the medians of the left and right group’s x-values and in all cases reproduces straight lines.
• Extending S4 classes with slots corresponding to special attributes like dim and dimnames now works.
• Fix for legend() when fill has multiple values the first of which is NA (all colours used to default to par(fg)). (PR#17288)
• installed.packages() did not remove the cached value for a library tree that had been emptied (but would not use the old value, just waste time checking it).
• The documentation for installed.packages(noCache = TRUE) incorrectly claimed it would refresh the cache.
• aggregate(<data.frame>) no longer uses spurious names in some cases. (PR#17283)
• object.size() now also works for long vectors.
• packageDescription() tries harder to solve re-encoding issues, notably seen in some Windows locales. This fixes the citation() issue in PR#17291.
• poly(<matrix>, 3) now works, thanks to prompting by Marc Schwartz.
• readLines() no longer segfaults on very large files with embedded '\0' (aka ‘nul’) characters. (PR#17311)
• ns() (package splines) now also works for a single observation. interpSpline() gives a more friendly error message when the number of points is less than four.
• dist(x, method = "canberra") now uses the correct definition; the result may only differ when x contains values of differing signs, e.g. not for 0-1 data.
• methods:::cbind() and methods:::rbind() avoid deep recursion, thanks to Suharto Anggono via PR#17300.
• Arithmetic with zero-column data frames now works more consistently; issue raised by Bill Dunlap.Arithmetic with data frames gives a data frame for ^ (which previously gave a numeric matrix).
• pretty(x, n) for large n or large diff(range(x)) now works better (though it was never meant for large n); internally it uses the same rounding fuzz (1e-10) as seq.default() — as it did up to 2010-02-03 when both were 1e-7.
• Internal C-level R_check_class_and_super() and hence R_check_class_etc() now also consider non-direct super classes and hence return a match in more cases. This e.g., fixes behaviour of derived classes in package Matrix.
• Reverted unintended change in behavior of return calls in on.exit expressions introduced by stack unwinding changes in R 3.3.0.
• Attributes on symbols are now detected and prevented; attempt to add an attribute to a symbol results in an error.
• fisher.test(*, workspace = <n>) now may also increase the internal stack size which allows larger problem to be solved, fixing PR#1662.
• The methods package no longer directly copies slots (attributes) into a prototype that is of an “abnormal” (reference) type, like a symbol.
• The methods package no longer attempts to call length<-() on NULL (during the bootstrap process).
• The methods package correctly shows methods when there are multiple methods with the same signature for the same generic (still not fully supported, but at least the user can see them).
• sys.on.exit() is now always evaluated in the right frame. (From Lionel Henry.)
• seq.POSIXt(*, by = "<n> DSTdays") now should work correctly in all cases and is faster. (PR#17342)
• .C() when returning a logical vector now always maps values other than FALSE and NA to TRUE (as documented).
• Subassignment with zero length vectors now coerces as documented (PR#17344).
  Further, x <- numeric(); x[1] <- character() now signals an error ‘replacement has length zero’ (or a translation of that) instead of doing nothing.
• (Package parallel.) mclapply(), pvec() and mcparallel() (when mccollect() is used to collect results) no longer leave zombie processes behind.
• R CMD INSTALL <pkg> now produces the intended error message when, e.g., the LazyData field is invalid.
• as.matrix(dd) now works when the data frame dd contains a column which is a data frame or matrix, including a 0-column matrix/d.f. .
• mclapply(X, mc.cores) now follows its documentation and calls lapply() in case mc.cores = 1 also in the case mc.preschedule is false. (PR#17373)
• aggregate(<data.frame>, drop=FALSE) no longer calls the function on parts but sets corresponding results to NA. (Thanks to Suharto Anggono’s patches in PR#17280).
• The duplicated() method for data frames is now based on the list method (instead of string coercion). Consequently unique() is better distinguishing data frame rows, fixing PR#17369 and PR#17381. The methods for matrices and arrays are changed accordingly.
• Calling names() on an S4 object derived from "environment" behaves (by default) like calling names() on an ordinary environment.
• read.table() with a non-default separator now supports quotes following a non-whitespace character, matching the behavior of scan().
• parLapplyLB and parSapplyLB have been fixed to do load balancing (dynamic scheduling). This also means that results of computations depending on random number generators will now really be non-reproducible, as documented.
• Indexing a list using dollar and empty string (l$"") returns NULL.
• Using \usage{ data(<name>, package="<pkg>") } no longer produces R CMD check warnings.
• match.arg() more carefully chooses the environment for constructing default choices, fixing PR#17401 as proposed by Duncan Murdoch.
• Deparsing of consecutive ! calls is now consistent with deparsing unary - and + calls and creates code that can be reparsed exactly; thanks to a patch by Lionel Henry inPR#17397. (As a side effect, this uses fewer parentheses in some other deparsing involving ! calls.)