R 4.0.0 Is Released, Hacker News

         Peter Dalgaard      pd @ me @ @end | ng | rom cb @@ dk        
(Fri Apr) : 41: (CEST)

The build system rolled up R-4.0.0.tar.gz (codename “Arbor Day”) this morning. The list below details the changes in this release. You can get the source code from 4 / R-4.0.0.tar.gz or wait for it to be mirrored at a CRAN site nearer to you. Binaries for various platforms will appear in due course. For the R Core Team, Peter Dalgaard These are the checksums (md5 and SHA – 391) for the freshly created files, in case you wish to check that they are uncorrupted: MD5 (AUTHORS)=b9c (f9f) (cab) (ad) (bebc) b4 MD5 (COPYING)=eb (b) (feef) (de) e (b5c) a MD5 (COPYING.LIB)=a6f (e) d9b6cdffcea4f (e) MD5 (FAQ)=4afa (cd) (aaa) f0ba e2e7bc5d6 MD5 (INSTALL)=(f) ca (f1ccf) ad7b MD5 (NEWS)=9aafc4b (fdf) (cf) (f) (ae) MD5 (NEWS.0)=bfcd7c (b) (d) c6f 70 e5a8 MD5 (NEWS.1)=eb c4d (ec9c) (b) cf0c2ebea
MD5 (NEWS.2)=(c) e2def (cebccddfb) ac6 MD5 (NEWS.3)=(e7f4a) (cc8ec) bf3f0ff 7801 ec8 MD5 (R-latest.tar.gz)=(c) (c) (cb) (f) (c) (a) b MD5 (README)=f (f) (c) (e) (a1b) decbbe6 MD5 (RESOURCES)=(fd3ffef) (d0a) MD5 (THANKS)=bb (f) (c) d (c) (fd) (f) (da MD5 (VERSION-INFO.dcf)=f6e4e a 507 cc (a) (d9f) (c4c) b8 MD5 (R-4 / R-4.0.0.tar.gz)=(c) (c) (cb) (f) (c) (a) b 2cde (a7b) (e5f) (b) (c) (c) (be0f) (fa) (b7ddefef) (c) e (c) AUTHORS e6d6a e 452 fe e (fcb) (f) dae (de) (ab) (b) cb4 COPYING e9ffa (dd) (f) (aa) (b) (c9ed) (f3d6bf8a) dc5d2dec8ccc0ef3 COPYING.LIB eddf (b) (c7b3b) cbc9b (c1beab) (b) e3dcd (bf) d2f6a8b2a c2a1 FAQ f (be6cbaecc4dce) (ac) e5bd (b) (ccdadaf) (cb9b) (e) (f) INSTALL b9c6a (f2b) (d) (ee) (c) (a8e5a) (b6f3e1edf) deb (a) (c8ed) ed NEWS 4e (b) (f) (b) (f) (fceab) (d) (c7d) (e) (a) (ba8e) (f) (f) NEWS.0 (b) (c) b1b2b (a6dcd) (a) (f) (fc) (b) (f) (d) NEWS.1 e de (c) f (ff) (fa) (b) (f) (a7fa5c) (e) e4ed7d5f6e NEWS.2 (d) (afa) (b5e) (d) (dc) edf (b) bc1ed3dce0 NEWS.3 (beb) (b) (eb0dcb5d) (ec) (bdfb4e) (e7a5a) R-latest.tar.gz 2fdd3e (f) (f) (d4b3a0c) (f2c) (a) d (f8cadf) c3ddc README ecc6e (fdb2cf7a9dbb1a6cb) (c) (f) (b8c) d1fd2f1ab9 RESOURCES 2a8dca (cd) ef9e (f) (ca) (c) (b) (b) (dac) a THANKS (ad) (cd5bbf2ad7a3de) b (b) d4f (a) (aa) (d8c1cf) c VERSION-INFO.dcf (beb) (b) (eb0dcb5d) (ec) (bdfb4e) (e7a5a) (R-4 / R-4.0.0.tar.gz This is the relevant part of the NEWS file CHANGES IN 4.0.0:   SIGNIFICANT USER-VISIBLE CHANGES:     Packages need to be (re-) installed under this version (4.0.0) of       R.     matrix objects now also inherit from class “array”, so e.g.,       class (diag (1)) is c (“matrix”, “array”). This invalidates code       incorrectly assuming that class (matrix_obj)) has length one.       S3 methods for class “array” are now dispatched for matrix       objects.     There is a new syntax for specifying _raw_ character constants       Similar to the one used in C : r “(…)” with … any character       sequence not containing the sequence) “. This makes it easier to       write strings that contain backslashes or both single and double       quotes. For more details see? Quotes.     R now uses a stringsAsFactors=FALSE default, and hence by       default no longer converts strings to factors in calls to       data.frame () and read.table ().       A large number of packages relied on the previous behavior and       so have needed / will need updating.     The plot () S3 generic function is now in package base rather than       package graphics, as it is reasonable to have methods that do not       use the graphics package. The generic is currently re-exported       from the graphics namespace to allow packages importing it from       there to continue working, but this may change in future.       Packages which define S4 generics for 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 on the search path (as in a namespace,       the base namespace has precedence over the search path).   REFERENCE COUNTING:     Reference counting is now used instead of the 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.   MIGRATION TO PCRE2:     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 _via_       extSoftVersion (): PCRE1 (formerly known as ‘PCRE’) has versions       =20.     Making PCRE2 available when building R from source is strongly       recommended (preferably version [[-n] or later) as PCRE1 is no       longer developed: version 8. is’ likely to be the final       release ‘.     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). R, B and X are no longer allowed in       character classes (PCRE1 treated these as literals).     Option PCRE_study is no longer used with PCRE2, and is reported       as FALSE when that is in use.   NEW FEATURES:     assertError () and assertWarning () (in package tools) can now       check for _specific_ error or warning classes _via_ the new       optional second argument classes (which is not back compatible       with previous use of an unnamed second argument).     DF2formula (), the utility for the data frame method of formula (),       now works without parsing and explicit evaluation, starting from       Suharto Anggono’s suggestion in PR # .     approxfun () and approx () gain a new argument na.rm defaulting to       true. If set to false, missing y values ​​now propagate into the       interpolated values.     Long vectors are now supported as the seq argument of a for ()       loop.     str (x) gets a new deparse.lines option with a default to speed it       up when x is a large call object.     The internal traceback object produced when an error is signalled       (.Traceback), now contains the calls rather than the _deparse () d_       calls, deferring the deparsing to the user-level functions       .traceback () and traceback (). This fulfils the wish of PR # 17671,       reported including two patch proposals by Brodie Gaslam.     data.matrix () now converts character columns to factors and from       this to integers.     package.skeleton () now explicitly lists all exports in the       NAMESPACE file.     New function .S3method () to register S3 methods in R scripts.     file.path () has some support for file paths not in the session       encoding, e.g. With UTF-8 inputs in a non-UTF-8 locale the output       is marked as UTF-8.     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 dir.exists (), file.exists (), file.access (), (), list.files (), normalizePath () and path.expand ())       Treat this like any other non-existent file, often with a       warning.     There is a new help document accessed by help (“file path       encoding “) detailing how file paths with marked encodings are       handled.     New function list2DF () for creating data frames from lists of       variables.     iconv () has a new option sub=”Unicode” to translate UTF-8 input       invalid in the to encoding using escapes.     There is a new function infoRDS () providing information about the       serialization format of a serialized object.     S3 method lookup now by default skips the elements of the search       path between the global and base environments.     Added an argument add_datalist (*, small.size=0) to allow the       creation of a data / datalist file even when the total size of the       data sets is small.     The backquote function bquote () has a new argument splice to       enable splicing a computed list of values ​​into an expression,       like, @ in LISP’s backquote.     The formula interface to t.test () and wilcox.test () has been       extended to handle one-sample and paired tests.     The palette () function has a new default set of colors (which       are less saturated and have better accessibility properties).       There are also some new built-in palettes, which are listed by       the new palette.pals () function. These include the old default       palette under the name “R3”. Finally, the new palette.colors ()       function allows a subset of colors to be selected from any of       the built-in palettes.     n2mfrow () gains an option asp=1 to specify the aspect ratio,       fulfilling the wish and extending the proposal of Michael Chirico       in PR # 17653.     For head (x, n) and tail () the default and other S3 methods       notably for _vector_ n, e.g. to get a “corner” of a matrix, has       been extended to array’s of higher dimension thanks to the patch       proposal by Gabe Becker in PR # 17668. Consequently, optional       argument addrownums is deprecated and replaced by the (more       general) argument keepnums. An invalid second argument n now       leads to typically more easily readable error messages.     New function .class2 () provides the full character vector of       class names used for S3 method dispatch.     Printing methods (..) now uses a new format () method.     sort.list (x) now works for non-atomic objects x and method=      “auto” (the default) or “radix” in cases order (x) works.     Where they are available, writeBin () allows long vectors.     New function deparse1 () produces one string, wrapping deparse (),       to be used typically in deparse1 (substitute [ date ]), e.g., to fix       PR # 17719.     wilcox.test () enhancements: In the (non-paired) two-sample case,       Inf values ​​are treated as very large for robustness consistency.       If exact computations are used, the result now has “exact” in the       method element of its return value. New arguments tol.root and       digits.rank where the latter may be used for stability to treat       very close numbers as ties.     readBin () and writeBin () now report an error for an invalid       endian value. The affected code needs to be fixed with care as       the old undocumented behavior was to swap endian-ness in such       cases.     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 S4Vectors Bioconductor package       and the advice of Herv’e Pag`es.     print () ‘s default method and many other methods (by calling the       default eventually and passing …) now make use of a new       optional width argument, avoiding the need for the user to set       and reset options (“width”).     memDecompress () supports the RFC format (eg in-memory       copies of gzip-compressed files) as well as RFC 2473.     memCompress () and memDecompress () support long raw vectors for       types “gzip” and “zx”.     sweep () and slice.index () can now use names of dimnames for their       MARGIN argument (apply has had this for almost a decade).     New function proportions () and marginSums (). These should replace       the unfortunately named prop.table () and margin.table (). They are       drop-in replacements, but also add named-margin functionality.       The old function names are retained as aliases for       back-compatibility.     Functions rbinom (), rgeom (), rhyper (), rpois (), rnbinom (),       rsignrank () and rwilcox () which have returned integer since R       3.0.0 and hence NA when the numbers would have been been the       integer range, now return double vectors (without NAs, typically)       in these cases.     matplot (x, y) (and hence matlines () and matpoints ()) now call the       corresponding methods of plot () and lines (), e.g, when x is a       “Date” or “POSIXct” object; prompted by Spencer Graves’       suggestion.     stopifnot () now allows customizing error messages via argument       names, thanks to a patch proposal by Neal Fultz in PR # 17730.     unlink () gains a new argument expand to disable wildcard and       tilde expansion. Elements of x of value “~” are now ignored.     mle () in the 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 “mle”, so saved objects from       earlier versions may need to be recomputed.     The default for pdf () is now useDingbats=FALSE.     The default fill color for hist () and boxplot () is now col=      “lightgray”.     The default order of the levels on the y-axis for spineplot () and       cdplot () has been reversed.     If the R_ALWAYS_INSTALL_TESTS environment variable is set to a       true value, R CMD INSTALL behaves as if the –install-tests       option is always specified. Thanks to Reinhold Koch for the       suggestion.     New function R_user_dir () in package tools suggests paths       appropriate for storing R-related user-specific data,       configuration and cache files.     capabilities () gains a new logical option Xchk to avoid w arnings       about X – related capabilities.     The internal implementation of grid units has changed, but the       only visible effects at user-level should be         a slightly different print format for some units (especially           unit arithmetic),         faster performance (for unit operations) and         two new functions unitType () and unit.psum ().       Based on code contributed by Thomas Lin Pedersen.     When internal dispatch for () and rep_len () fails, there       is an attempt to dispatch on the equivalent call to rep ().     Object .Machine now contains new longdouble. Entries (when R       uses long doubles internally).     news () has been enhanced to cover the news on R 3.x and 2.x.     For consistency, N is required: the       deprecated provision for 8. 41 – 8.
has been removed.     Defunct functions mem.limits (), .readRDS (),       .saveRDS (), .. find.package (), and .path.package () from package       base and allGenerics (), getAccess (), getAllMethods (),       getClassName (), getClassPackage (), getExtends (), getProperties (),       getPrototype (), getSubclasses (), getVirtual (), mlistMetaName (),       removeMethodsObject (), seemsS4Object (), traceOff (), and traceOn ()       from methods have been removed.   C-LEVEL FACILITIES:     installChar is now remapped in Rinternals.h to installTrChar, of       which it has been a wrapper since R 3.6.0. Neither are part of       the API, but packages using installChar can replace it if they       depend on R>=3.6.2.     Header R_ext / Print.h defines R_USE_C 281 _ IN_CXX and hence exposes       Rvprintf and REvprintf if used with a C (or later) compiler.     There are new Fortran subroutines dblepr1, realpr1 and intpr1 to       print a scalar variable (gfortran 23 enforces the distinction       between scalars and length-one arrays). Also labelpr to print       just a label.     R_withCallingErrorHandler is now available for establishing a       calling handler in C code for conditions inheriting from class       error.   INSTALLATION on a UNIX-ALIKE:     User-set DEFS (e.g., in is now used for compiling       packages (including base packages).     There is a new variant option –enable-lto=check for checking       consistency of BLAS / LAPACK / LINPACK calls – see ‘Writing R       Extensions’.     A C compiler default is set only if the C standard is       supported: it no longer falls back to C .     PCRE2 is used if available. To make use of PCRE1 if PCRE2 is       unavailable, configure with option –with-pcre1.     The minimum required version of libcurl is now 7. 0 (Oct
) .     New make target distcheck checks         R can be rebuilt from the tarball created by make dist,         the build from the tarball passes make check-all,         the build installs and uninstalls,         the source files are properly cleaned by make distclean.   UTILITIES:     R –help now mentions the option –no-echo (renamed from –slave)       and its previously undocumented short form -s.     R CMD check now optionally checks configure and cleanup scripts       for non-Bourne-shell code (‘bashisms’).     R CMD check –as-cran now runs donttest examples (which are run       by example ()) instead of instructing the tester to do so. This       can be temporarily circumvented during development by setting       environment variable _R_CHECK_DONTTEST_EXAMPLES_ to a false       value.   PACKAGE INSTALLATION:     There is the beginnings of support for the recently approved       C (standard, specified analogously to C ) and C 37. There       It is currently only limited support for this in compilers, with       flags such as -std=c and -std=c 2a. For the time being the       configure test is of accepting one of these flags and compiling       C 32 code.   BUG FIXES:     formula (x) with length (x)> 1 character vectors, is deprecated       now. Such use has been rare, and has ‘worked’ as expected in       some cases only. In other cases, wrong x have silently been       truncated, not detecting previous errors.     Long-standing issue where the X device could lose events       shortly after startup has been addressed (PR # 17532.     The data.frame method for rbind () no longer drops levels       from factor columns by default (PR # 17580.     available.packages () and hence install.packages () now pass their       … argument to download.file (), fulfilling the wish of PR # 17595;       subsequently, available.packages () gets new argument quiet,       solving PR # .     stopifnot () gets new argument exprObject to allow an R object of       class expression (or other ‘language’) to work more consistently,       thanks to suggestions by Suharto Anggono.     conformMethod () now works correctly in cases containing a “&&       logic “bug, reported by Henrik Bengtsson. It now creates methods       with “missing” entries in the signature. Consequently,       rematchDefinition () is amended to use appropriate .local () calls       with named arguments where needed.     format.default (*, scientific=FALSE) now corresponds to a       practically most extreme options (scipen=n) setting rather than       arbitrary n=.     format (as.symbol (“foo”)) now works (returning “foo”).     postscript (.., title=now signals an error when the title       string contains a character which would produce corrupt       PostScript, thanks to PR # by Daisuko Ogawa.     Certain Ops (notably comparison such as==) now also work for       0-length data frames, after reports by Hilmar Berger.     methods (class=class (glm (..))) now warns more usefully and only       once.     write.dcf () no longer mangles field names (PR # ).     Primitive replacement functions no longer mutate a referenced       first argument when used outside of a complex assignment context.     A better error message for contour (*, levels=Inf).     The return value of contourLines () is no longer invisible ().     The Fortran code for calculating the coefficients component in       lm.influence () was very inefficient. It has (for now) been       replaced with much faster R code (PR # 17638.     cm.colors (n) _etc_ no longer append the code for alpha=1, “FF”,       to all colors. Hence all eight .colors () functions and       rainbow () behave consistently and have the same non-explicit       default (PR # 569978484 ).     dnorm had a problematic corner case with sd==-Inf or negative       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 dlnorm ().     The optional iter.smooth argument of plot.lm (), (the plot ()       method for lm and glm fits) now defaults to 0 for all 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 glm case has been       switched to “pearson” since deviance residuals do not in general       have approximately zero mean.     In plot.lm, Cook’s distance was computed from unweighted       residuals, leading to inconsistencies. Replaced with usual       weighted version. (PR # 11484673906     Time-series ts (*, start, end, frequency) with fractional       frequency are supported more consistently; thanks to a report       from Johann Kleinbub and analysis and patch by Duncan Murdoch in       PR # 17693.     In case of errors mcmapply () now preserves attributes of returned       “try-error” objects and avoids simplification, overriding       SIMPLIFY to FALSE. (PR # 11484673906 )     as.difftime () gets new optional tz=”UTC” argument which should       fix behavior during daylight-savings-changeover days, fixing       PR # , thanks to proposals and analysis by Johannes Ranke and       Kirill M “uller.     round () does a better job of rounding _ “to nearest” _ by       _measuring_ and _ “to even” _; thanks to a careful algorithm       originally prompted by the report from Adam Wheeler and then       others, in PR # 17688.       round (x, dig) for _negative_ digits is much more rational now,       notably for large | dig |.     Inheritance information on S4 classes is maintained more       consistently, particularly in the case of class unions (in part       due to PR # 17624 and a report from Ezra Tucker).     is () behaves more robustly when its argument class2 is a       classRepresentation object.     The warning message when attempting to export an nonexistent       class is now more readable; thanks to Thierry Onkelinx for       recognizing the problem.     choose () misbehaved in corner cases where it switched n – k for k       and n was only _nearly_ integer (report from Erik Scott Wright).     mle () in the stats4 package had problems combining use of box       constraints and fixed starting values ​​(in particular, confidence       intervals were affected).     Operator? now has lower precedence than=to work as documented,       so=behaves like ).     smoothEnds (x) now returns integer type in _both_ cases when x is       integer, thanks to a report and proposal by Bill Dunlap PR # 17734.     The methods package does a better job of tracking inheritance       relationships across packages.     norm (diag (c (1, NA)), “2”) now works.     subset () had problems with 0-col dataframes (reported by Bill       Dunlap, PR # 17730) .     Several cases of integer overflow detected by the ‘undefined       behavior sanitizer ‘of clang have been circumvented. One in       rhyper () may change the generated value for large input values.     dotchart () now places the y-axis label (ylab) much better, not       overplotting labels, thanks to a report and suggestion by Alexey       Shipunov.     A rare C-level array overflow in chull () has been worked around.     Some invalid specifications of the day-of-the-year (_via_% j,       e.g. day (in) ) or week plus day-of-the-week are now       detected by strptime (). They now return NA but give a warning as       they may have given random results or corrupted memory in earlier       versions of R.     socketConnection (server=FALSE) now respects the connection       timeout also on Linux.     socketConnection (server=FALSE) no longer leaks a connection       that is available right away without waiting (e.g. on localhost).     Socket connections are now robust against spurious readability       and spurious availability of an incoming connection.     blocking=FALSE is now respected also on the server side of a       socket connection, allowing non-blocking read operations.     anova.glm () and anova.glmlist () computed incorrect score (Rao)       tests in no-intercept cases. (Andr’e Gillibert, PR # )     summaryRprof () now should work correctly for the Rprof (*,       memory.profiling=TRUE) case with small chunk size (and “tseries”       or similar) thanks to a patch proposal by Benjamin Tyner, in       PR # 16710.     xgettext () ignores strings passed to ngettext (), since the latter       is handled by xngettext (). Thanks to Daniele Medri for the report       and all the recent work he has done on the Italian translations.     data (package=”P”) for P in base and stats no longer reports the       data sets from package datasets (which it did for back       compatibility for (years), fixing PR #

.     x [[Inf]] (returning NULL) no longer leads to undefined behavior,       thanks to a report by Kirill M “uller in PR # 22839. Further,       x [[-Inf]] and x [[-n]] now give more helpful error messages.     Gamma () family sometimes had trouble storing link name PR # 16764   BUG FIXES (Windows):     Sys.glob () now supports all characters from the Unicode Basic       Multilingual Plane, no longer corrupting some (less commonly       used) characters (PR # 17638.     Rterm now correctly displays multi-byte-coded characters       representable in the current native encoding (at least on Windows       20 they were sometimes omitted, PR # 17648).     scan () issues with UTF-8 data when running in a DBCS locale have       been resolved (PR # 444506264355 , PR # 17165).     RTerm now accepts enhanced / arrow keys also with ConPTY.     R can can now be started _via_ the launcher icon in a user       documents directory whose path is not representable in the system       encoding.     socketConnection (server=FALSE) now returns instantly also on       Windows when connection failure is signalled.     Problems with UTF – surrogate pairs have been fixed in several       functions, including tolower () and toupper () (PR # . CHANGES in previous versions:   Older news can be found in text format in files NEWS.0, NEWS.1,     NEWS.2 and NEWS.3 in the doc directory. News in HTML format for R     versions 3.x and from 2. 0 to 2. . 3 is available at     doc / html / NEWS.3.html and doc / html / NEWS.2.html. – Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, Frederiksberg, Denmark Phone: ( 62) Office: A 4. 47 Email: pd.mes using (Priv: PDalgd using

More information about the R-announce mailing list
Read More

What do you think?

Leave a Reply

Your email address will not be published.

GIPHY App Key not set. Please check settings

New Tintin Game is the Nostalgic Action-Adventure We Deserve, Crypto Coins News

New Tintin Game is the Nostalgic Action-Adventure We Deserve, Crypto Coins News

How to make pizza like a Neapolitan master, Hacker News

How to make pizza like a Neapolitan master, Hacker News