FAQ

versions of R prior to 2.7.0, or without support for cairo.  From R
2.7.0 'png()' by default uses the Quartz device on macOS, and that too
works in batch mode.

   Earlier versions of the 'png()' device used the X11 driver, which is
a problem in batch mode or for remote operation.  If you have
Ghostscript you can use 'bitmap()', which produces a PostScript or PDF
file then converts it to any bitmap format supported by Ghostscript.  On
some installations this produces ugly output, on others it is perfectly
satisfactory.  Many systems now come with Xvfb from X.Org
(https://www.x.org/) (possibly as an optional install), which is an X11
server that does not require a screen; and there is the *GDD*
(https://CRAN.R-project.org/package=GDD) package from CRAN, which
produces PNG, JPEG and GIF bitmaps without X11.

7.20 How can I get command line editing to work?
================================================

The Unix-like command-line interface to R can only provide the inbuilt
command line editor which allows recall, editing and re-submission of
prior commands provided that the GNU readline library is available at
the time R is configured for compilation.  Note that the 'development'
version of readline including the appropriate headers is needed: users
of Linux binary distributions will need to install packages such as
'libreadline-dev' (Debian) or 'readline-devel' (Red Hat).

7.21 How can I turn a string into a variable?
=============================================

If you have

     varname <- c("a", "b", "d")

you can do

     get(varname[1]) + 2

for

     a + 2

or

     assign(varname[1], 2 + 2)

for

     a <- 2 + 2

or

     eval(substitute(lm(y ~ x + variable),
                     list(variable = as.name(varname[1]))))

for

     lm(y ~ x + a)

   At least in the first two cases it is often easier to just use a
list, and then you can easily index it by name

     vars <- list(a = 1:10, b = rnorm(100), d = LETTERS)
     vars[["a"]]

without any of this messing about.

7.22 Why do lattice/trellis graphics not work?
==============================================

The most likely reason is that you forgot to tell R to display the
graph.  Lattice functions such as 'xyplot()' create a graph object, but
do not display it (the same is true of *ggplot2*
(https://CRAN.R-project.org/package=ggplot2) graphics, and Trellis
graphics in S-PLUS).  The 'print()' method for the graph object produces
the actual display.  When you use these functions interactively at the
command line, the result is automatically printed, but in 'source()' or
inside your own functions you will need an explicit 'print()' statement.

7.23 How can I sort the rows of a data frame?
=============================================

To sort the rows within a data frame, with respect to the values in one
or more of the columns, simply use 'order()' (e.g., 'DF[order(DF$a,
DF[["b"]]), ]' to sort the data frame 'DF' on columns named 'a' and
'b').

7.24 Why does the help.start() search engine not work?
======================================================

The browser-based search engine in 'help.start()' utilizes a Java
applet.  In order for this to function properly, a compatible version of
Java must installed on your system and linked to your browser, and both
Java _and_ JavaScript need to be enabled in your browser.

   There have been a number of compatibility issues with versions of
Java and of browsers.  For further details please consult section
"Enabling search in HTML help" in 'R Installation and Administration'.
This manual is included in the R distribution, *note What documentation
exists for R?::, and its HTML version is linked from the HTML search
page.

7.25 Why did my .Rprofile stop working when I updated R?
========================================================

Did you read the 'NEWS' file?  For functions that are not in the *base*
package you need to specify the correct package namespace, since the
code will be run _before_ the packages are loaded.  E.g.,

     ps.options(horizontal = FALSE)
     help.start()

needs to be

     grDevices::ps.options(horizontal = FALSE)
     utils::help.start()

('graphics::ps.options(horizontal = FALSE)' in R 1.9.x).

7.26 Where have all the methods gone?
=====================================

Many functions, particularly S3 methods, are now hidden in namespaces.
This has the advantage that they cannot be called inadvertently with
arguments of the wrong class, but it makes them harder to view.

   To see the code for an S3 method (e.g., '[.terms') use

     getS3method("[", "terms")

To see the code for an unexported function 'foo()' in the namespace of
package '"bar"' use 'bar:::foo'.  Don't use these constructions to call
unexported functions in your own code--they are probably unexported for
a reason and may change without warning.

7.27 How can I create rotated axis labels?
==========================================

To rotate axis labels (using base graphics), you need to use 'text()',
rather than 'mtext()', as the latter does not support 'par("srt")'.

     ## Increase bottom margin to make room for rotated labels
     par(mar = c(7, 4, 4, 2) + 0.1)
     ## Create plot with no x axis and no x axis label
     plot(1 : 8, xaxt = "n",  xlab = "")
     ## Set up x axis with tick marks alone
     axis(1, labels = FALSE)
     ## Create some text labels
     labels <- paste("Label", 1:8, sep = " ")
     ## Plot x axis labels at default tick marks
     text(1:8, par("usr")[3] - 0.25, srt = 45, adj = 1,
          labels = labels, xpd = TRUE)
     ## Plot x axis label at line 6 (of 7)
     mtext(1, text = "X Axis Label", line = 6)

When plotting the x axis labels, we use 'srt = 45' for text rotation
angle, 'adj = 1' to place the right end of text at the tick marks, and
'xpd = TRUE' to allow for text outside the plot region.  You can adjust
the value of the '0.25' offset as required to move the axis labels up or
down relative to the x axis.  See '?par' for more information.

   Also see Figure 1 and associated code in Paul Murrell (2003),
"Integrating grid Graphics Output with Base Graphics Output", _R News_,
*3/2*, 7-12.

7.28 Why is read.table() so inefficient?
========================================

By default, 'read.table()' needs to read in everything as character
data, and then try to figure out which variables to convert to numerics
or factors.  For a large data set, this takes considerable amounts of
time and memory.  Performance can substantially be improved by using the
'colClasses' argument to specify the classes to be assumed for the
columns of the table.

7.29 What is the difference between package and library?
========================================================

A "package" is a standardized collection of material extending R, e.g.
providing code, data, or documentation.  A "library" is a place
(directory) where R knows to find packages it can use (i.e., which were
"installed").  R is told to use a package (to "load" it and add it to
the search path) via calls to the function 'library'.  I.e., 'library()'
is employed to load a package from libraries containing packages.

   *Note R Add-On Packages::, for more details.  See also Uwe Ligges
(2003), "R Help Desk: Package Management", _R News_, *3/3*, 37-39.

7.30 I installed a package but the functions are not there
==========================================================

To actually _use_ the package, it needs to be _loaded_ using
'library()'.

   See *note R Add-On Packages:: and *note What is the difference
between package and library?:: for more information.

7.31 Why doesn't R think these numbers are equal?
=================================================

The only numbers that can be represented exactly in R's numeric type are
integers and fractions whose denominator is a power of 2.  All other
numbers are internally rounded to (typically) 53 binary digits accuracy.
As a result, two floating point numbers will not reliably be equal
unless they have been computed by the same algorithm, and not always
even then.  For example

     R> a <- sqrt(2)
     R> a * a == 2
     [1] FALSE
     R> a * a - 2
     [1] 4.440892e-16
     R> print(a * a, digits = 18)
     [1] 2.00000000000000044

   The function 'all.equal()' compares two objects using a numeric
tolerance of '.Machine$double.eps ^ 0.5'.  If you want much greater
accuracy than this you will need to consider error propagation
carefully.

   A discussion with many easily followed examples is in Appendix G
"Computational Precision and Floating Point Arithmetic", pages 753-771
of _Statistical Analysis and Data Display: An Intermediate Course with
Examples in R_, Richard M. Heiberger and Burt Holland (Springer 2015,
second edition).  This appendix is a free download from
<https://link.springer.com/content/pdf/bbm%3A978-1-4939-2122-5%2F1.pdf>.

   For more information, see e.g. David Goldberg (1991), "What Every
Computer Scientist Should Know About Floating-Point Arithmetic", _ACM
Computing Surveys_, *23/1*, 5-48, also available via
<https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html>.

   Here is another example, this time using addition:

     R> .3 + .6 == .9
     [1] FALSE
     R> .3 + .6 - .9
     [1] -1.110223e-16
     R> print(matrix(c(.3,  .6,  .9, .3 + .6)), digits = 18)
                          [,1]
     [1,] 0.299999999999999989
     [2,] 0.599999999999999978
     [3,] 0.900000000000000022
     [4,] 0.899999999999999911

7.32 How can I capture or ignore errors in a long simulation?
=============================================================

Use 'try()', which returns an object of class '"try-error"' instead of
an error, or preferably 'tryCatch()', where the return value can be
configured more flexibly.  For example

     beta[i,] <- tryCatch(coef(lm(formula, data)),
                          error = function(e) rep(NaN, 4))

would return the coefficients if the 'lm()' call succeeded and would
return 'c(NaN, NaN, NaN, NaN)' if it failed (presumably there are
supposed to be 4 coefficients in this example).

7.33 Why are powers of negative numbers wrong?
==============================================

You are probably seeing something like

     R> -2^2
     [1] -4

and misunderstanding the precedence rules for expressions in R. Write

     R> (-2)^2
     [1] 4

to get the square of -2.

   The precedence rules are documented in '?Syntax', and to see how R
interprets an expression you can look at the parse tree

     R> as.list(quote(-2^2))
     [[1]]
     `-`

     [[2]]
     2^2

7.34 How can I save the result of each iteration in a loop into a separate file?
================================================================================

One way is to use 'paste()' (or 'sprintf()') to concatenate a stem
filename and the iteration number while 'file.path()' constructs the
path.  For example, to save results into files 'result1.rda', ...,
'result100.rda' in the subdirectory 'Results' of the current working
directory, one can use

     for(i in 1:100) {
       ## Calculations constructing "some_object" ...
       fp <- file.path("Results", paste("result", i, ".rda", sep = ""))
       save(list = "some_object", file = fp)
     }

7.35 Why are p-values not displayed when using lmer()?
======================================================

Doug Bates has kindly provided an extensive response in a post to the
r-help list, which can be reviewed at
<https://stat.ethz.ch/pipermail/r-help/2006-May/094765.html>.

7.36 Why are there unwanted borders, lines or grid-like artifacts when viewing a plot saved to a PS or PDF file?
================================================================================================================

This can occur when using functions such as 'polygon()',
'filled.contour()', 'image()' or other functions which may call these
internally.  In the case of 'polygon()', you may observe unwanted
borders between the polygons even when setting the 'border' argument to
'NA' or '"transparent"'.

   The source of the problem is the PS/PDF viewer when the plot is
anti-aliased.  The details for the solution will be different depending
upon the viewer used, the operating system and may change over time.
For some common viewers, consider the following:

Acrobat Reader (cross platform)
     There are options in Preferences to enable/disable text smoothing,
     image smoothing and line art smoothing.  Disable line art
     smoothing.
Preview (macOS)
     There is an option in Preferences to enable/disable anti-aliasing
     of text and line art.  Disable this option.
GSview (cross platform)
     There are settings for Text Alpha and Graphics Alpha.  Change
     Graphics Alpha from 4 bits to 1 bit to disable graphic
     anti-aliasing.
gv (Unix-like X)
     There is an option to enable/disable anti-aliasing.  Disable this
     option.
Evince (Linux/GNOME)
     There is not an option to disable anti-aliasing in this viewer.
Okular (Linux/KDE)
     There is not an option in the GUI to enable/disable anti-aliasing.
     From a console command line, use:
          $ kwriteconfig --file okularpartrc --group 'Dlg Performance' \
                         --key GraphicsAntialias Disabled
     Then restart Okular.  Change the final word to 'Enabled' to restore
     the original setting.

7.37 Why does backslash behave strangely inside strings?
========================================================

This question most often comes up in relation to file names (see *note
How do file names work in Windows?::) but it also happens that people
complain that they cannot seem to put a single '\' character into a text
string unless it happens to be followed by certain other characters.

   To understand this, you have to distinguish between character strings
and _representations_ of character strings.  Mostly, the representation
in R is just the string with a single or double quote at either end, but
there are strings that cannot be represented that way, e.g., strings
that themselves contain the quote character.  So

     > str <- "This \"text\" is quoted"
     > str
     [1] "This \"text\" is quoted"
     > cat(str, "\n")
     This "text" is quoted

The _escape sequences_ '\"' and '\n' represent a double quote and the
newline character respectively.  Printing text strings, using 'print()'
or by typing the name at the prompt will use the escape sequences too,
but the 'cat()' function will display the string as-is.  Notice that
'"\n"' is a one-character string, not two; the backslash is not actually
in the string, it is just generated in the printed representation.

     > nchar("\n")
     [1] 1
     > substring("\n", 1, 1)
     [1] "\n"

   So how do you put a backslash in a string?  For this, you have to
escape the escape character.  I.e., you have to double the backslash.
as in

     > cat("\\n", "\n")
     \n

   Some functions, particularly those involving regular expression
matching, themselves use metacharacters, which may need to be escaped by
the backslash mechanism.  In those cases you may need a _quadruple_
backslash to represent a single literal one.

   In versions of R up to 2.4.1 an unknown escape sequence like '\p' was
quietly interpreted as just 'p'.  Current versions of R emit a warning.

7.38 How can I put error bars or confidence bands on my plot?
=============================================================

Some functions will display a particular kind of plot with error bars,
such as the 'bar.err()' function in the *agricolae*
(https://CRAN.R-project.org/package=agricolae) package, the 'plotCI()'
function in the *gplots* (https://CRAN.R-project.org/package=gplots)
package, the 'plotCI()' and 'brkdn.plot()' functions in the *plotrix*
(https://CRAN.R-project.org/package=plotrix) package and the
'error.bars()', 'error.crosses()' and 'error.bars.by()' functions in the
*psych* (https://CRAN.R-project.org/package=psych) package.  Within
these types of functions, some will accept the measures of dispersion
(e.g., 'plotCI'), some will calculate the dispersion measures from the
raw values ('bar.err', 'brkdn.plot'), and some will do both
('error.bars').  Still other functions will just display error bars,
like the dispersion function in the *plotrix*
(https://CRAN.R-project.org/package=plotrix) package.  Most of the above
functions use the 'arrows()' function in the base *graphics* package to
draw the error bars.

   The above functions all use the base graphics system.  The grid and
lattice graphics systems also have specific functions for displaying
error bars, e.g., the 'grid.arrow()' function in the *grid* package, and
the 'geom_errorbar()', 'geom_errorbarh()', 'geom_pointrange()',
'geom_linerange()', 'geom_crossbar()' and 'geom_ribbon()' functions in
the *ggplot2* (https://CRAN.R-project.org/package=ggplot2) package.  In
the lattice system, error bars can be displayed with 'Dotplot()' or
'xYplot()' in the *Hmisc* (https://CRAN.R-project.org/package=Hmisc)
package and 'segplot()' in the *latticeExtra*
(https://CRAN.R-project.org/package=latticeExtra) package.

7.39 How do I create a plot with two y-axes?
============================================

Creating a graph with two y-axes, i.e., with two sorts of data that are
scaled to the same vertical size and showing separate vertical axes on
the left and right sides of the plot that reflect the original scales of
the data, is possible in R but is not recommended.  The basic approach
for constructing such graphs is to use 'par(new=TRUE)' (see '?par');
functions 'twoord.plot()' (in the *plotrix*
(https://CRAN.R-project.org/package=plotrix) package) and
'doubleYScale()' (in the *latticeExtra*
(https://CRAN.R-project.org/package=latticeExtra) package) automate the
process somewhat.

7.40 How do I access the source code for a function?
====================================================

In most cases, typing the name of the function will print its source
code.  However, code is sometimes hidden in a namespace, or compiled.
For a complete overview on how to access source code, see Uwe Ligges
(2006), "Help Desk: Accessing the sources", _R News_, *6/4*, 43-45
(<https://CRAN.R-project.org/doc/Rnews/Rnews_2006-4.pdf>).

7.41 Why does summary() report strange results for the R^2 estimate when I fit a linear model with no intercept?
================================================================================================================

As described in '?summary.lm', when the intercept is zero (e.g., from 'y
~ x - 1' or 'y ~ x + 0'), 'summary.lm()' uses the formula R^2 = 1 -
Sum(R[i]^2) / Sum((y[i])^2) which is different from the usual R^2 = 1 -
Sum(R[i]^2) / Sum((y[i] - mean(y))^2).  There are several reasons for
this:
   * Otherwise the R^2 could be negative (because the model with zero
     intercept can fit _worse_ than the constant-mean model it is
     implicitly compared to).
   * If you set the slope to zero in the model with a line through the
     origin you get fitted values y*=0
   * The model with constant, non-zero mean is not nested in the model
     with a line through the origin.

   All these come down to saying that if you know _a priori_ that E[Y]=0
when x=0 then the 'null' model that you should compare to the fitted
line, the model where x doesn't explain any of the variance, is the
model where E[Y]=0 everywhere.  (If you don't know a priori that E[Y]=0
when x=0, then you probably shouldn't be fitting a line through the
origin.)

7.42 Why is R apparently not releasing memory?
==============================================

This question is often asked in different flavors along the lines of "I
have removed objects in R and run 'gc()' and yet 'ps'/'top' still shows
the R process using a lot of memory", often on Linux machines.

   This is an artifact of the way the operating system (OS) allocates
memory.  In general it is common that the OS is not capable of releasing
all unused memory.  In extreme cases it is possible that even if R frees
almost all its memory, the OS can not release any of it due to its
design and thus tools such as 'ps' or 'top' will report substantial
amount of resident RAM used by the R process even though R has released
all that memory.  In general such tools do _not_ report the actual
memory usage of the process but rather what the OS is reserving for that
process.

   The short answer is that this is a limitation of the memory allocator
in the operating system and there is nothing R can do about it.  That
space is simply kept by the OS in the hope that R will ask for it later.
The following paragraph gives more in-depth answer with technical
details on how this happens.

   Most systems use two separate ways to allocate memory.  For
allocation of large chunks they will use 'mmap' to map memory into the
process address space.  Such chunks can be released immediately when
they are completely free, because they can reside anywhere in the
virtual memory.  However, this is a relatively expensive operation and
many OSes have a limit on the number of such allocated chunks, so this
is only used for allocating large memory regions.  For smaller
allocations the system can expand the data segment of the process
(historically using the 'brk' system call), but this whole area is
always contiguous.  The OS can only move the end of this space, it
cannot create any "holes".  Since this operation is fairly cheap, it is
used for allocations of small pieces of memory.  However, the
side-effect is that even if there is just one byte that is in use at the
end of the data segment, the OS cannot release any memory at all,
because it cannot change the address of that byte.  This is actually
more common than it may seem, because allocating a lot of intermediate
objects, then allocating a result object and removing all intermediate
objects is a very common practice.  Since the result is allocated at the
end it will prevent the OS from releasing any memory used by the
intermediate objects.  In practice, this is not necessarily a problem,
because modern operating systems can page out unused portions of the
virtual memory so it does not necessarily reduce the amount of real
memory available for other applications.  Typically, small objects such
as strings or pairlists will be affected by this behavior, whereas large
objects such as long vectors will be allocated using 'mmap' and thus not
affected.  On Linux (and possibly other Unix-like systems) it is
possible to use the 'mallinfo' system call (also see the mallinfo
(https://rforge.net/mallinfo) package) to query the allocator about the
layout of the allocations, including the actually used memory as well as
unused memory that cannot be released.

7.43 How can I enable secure https downloads in R?
==================================================

When R transfers files over HTTP (e.g., using the 'install.packages()'
or 'download.file()' function), a download method is chosen based on the
'download.file.method' option.  There are several methods available and
the default behavior if no option is explicitly specified is to use R's
internal HTTP implementation.  In most circumstances this internal
method will not support HTTPS URLs so you will need to override the
default: this is done automatically for such URLs as from R 3.2.2.

   R versions 3.2.0 and greater include two download methods
('"libcurl"' and '"wininet"') that both support HTTPS connections: we
recommend that you use these methods.  The requisite code to add to
'.Rprofile' or 'Rprofile.site' is:

     options(download.file.method = "wininet", url.method = "wininet")
     (Windows)
     options(download.file.method = "libcurl", url.method = "libcurl")
     (Linux and macOS)

(Method '"wininet"' is the default on Windows as from R 3.2.2.)

   Note that the '"libcurl"' method may or may not have been compiled
in.  In the case that it was not, i.e. 'capabilities("libcurl") ==
FALSE', we recommend method '"wget"' on Linux and '"curl"' on macOS. It
is possible that system versions of '"libcurl"', 'wget' or 'curl' may
have been compiled without HTTPS support, but this is unlikely.  As from
R 3.3.0 '"libcurl"' with HTTPS support is required except on Windows.

7.44 How can I get CRAN package binaries for outdated versions of R?
====================================================================

Since March 2016, Windows and macOS binaries of CRAN packages for old
versions of R (released more than 5 years ago) are made available from a
central CRAN archive server instead of the CRAN mirrors.  To get these,
one should set the CRAN "mirror" element of the 'repos' option
accordingly, by something like
     local({r <- getOption("repos")
            r["CRAN"] <- "http://CRAN-archive.R-project.org"
            options(repos = r)
           })
(see '?options' for more information).

8 R Programming
***************

8.1 How should I write summary methods?
=======================================

Suppose you want to provide a summary method for class '"foo"'.  Then
'summary.foo()' should not print anything, but return an object of class
'"summary.foo"', _and_ you should write a method 'print.summary.foo()'
which nicely prints the summary information and invisibly returns its
object.  This approach is preferred over having 'summary.foo()' print
summary information and return something useful, as sometimes you need
to grab something computed by 'summary()' inside a function or similar.
In such cases you don't want anything printed.

8.2 How can I debug dynamically loaded code?
============================================

Roughly speaking, you need to start R inside the debugger, load the
code, send an interrupt, and then set the required breakpoints.

   See section "Finding entry points in dynamically loaded code" in
'Writing R Extensions'.  This manual is included in the R distribution,
*note What documentation exists for R?::.

8.3 How can I inspect R objects when debugging?
===============================================

The most convenient way is to call 'R_PV' from the symbolic debugger.

   See section "Inspecting R objects when debugging" in 'Writing R
Extensions'.

8.4 How can I change compilation flags?
=======================================

Suppose you have C code file for dynloading into R, but you want to use
'R CMD SHLIB' with compilation flags other than the default ones (which
were determined when R was built).

   Starting with R 2.1.0, users can provide personal Makevars
configuration files in '$HOME/.R' to override the default flags.  See
section "Add-on packages" in 'R Installation and Administration'.

   For earlier versions of R, you could change the file
'R_HOME/etc/Makeconf' to reflect your preferences, or (at least for
systems using GNU Make) override them by the environment variable
'MAKEFLAGS'.  See section "Creating shared objects" in 'Writing R
Extensions'.

8.5 How can I debug S4 methods?
===============================

Use the 'trace()' function with argument 'signature=' to add calls to
the browser or any other code to the method that will be dispatched for
the corresponding signature.  See '?trace' for details.

9 R Bugs
********

9.1 What is a bug?
==================

If R executes an illegal instruction, or dies with an operating system
error message that indicates a problem in the program (as opposed to
something like "disk full"), then it is certainly a bug.  If you call
'.C()', '.Fortran()', '.External()' or '.Call()' (or '.Internal()')
yourself (or in a function you wrote), you can always crash R by using
wrong argument types (modes).  This is not a bug.

   Taking forever to complete a command can be a bug, but you must make
certain that it was really R's fault.  Some commands simply take a long
time.  If the input was such that you _know_ it should have been
processed quickly, report a bug.  If you don't know whether the command
should take a long time, find out by looking in the manual or by asking
for assistance.

   If a command you are familiar with causes an R error message in a
case where its usual definition ought to be reasonable, it is probably a
bug.  If a command does the wrong thing, that is a bug.  But be sure you
know for certain what it ought to have done.  If you aren't familiar
with the command, or don't know for certain how the command is supposed
to work, then it might actually be working right.  For example, people
sometimes think there is a bug in R's mathematics because they don't
understand how finite-precision arithmetic works.  Rather than jumping
to conclusions, show the problem to someone who knows for certain.
Unexpected results of comparison of decimal numbers, for example '0.28 *
100 != 28' or '0.1 + 0.2 != 0.3', are not a bug.  *Note Why doesn't R
think these numbers are equal?::, for more details.

   Finally, a command's intended definition may not be best for
statistical analysis.  This is a very important sort of problem, but it
is also a matter of judgment.  Also, it is easy to come to such a
conclusion out of ignorance of some of the existing features.  It is
probably best not to complain about such a problem until you have
checked the documentation in the usual ways, feel confident that you
understand it, and know for certain that what you want is not available.
If you are not sure what the command is supposed to do after a careful
reading of the manual this indicates a bug in the manual.  The manual's
job is to make everything clear.  It is just as important to report
documentation bugs as program bugs.  However, we know that the
introductory documentation is seriously inadequate, so you don't need to
report this.

   If the online argument list of a function disagrees with the manual,
one of them must be wrong, so report the bug.

9.2 How to report a bug
=======================

When you decide that there is a bug, it is important to report it and to
report it in a way which is useful.  What is most useful is an exact
description of what commands you type, starting with the shell command
to run R, until the problem happens.  Always include the version of R,
machine, and operating system that you are using; type 'version' in R to
print this.

   The most important principle in reporting a bug is to report _facts_,
not hypotheses or categorizations.  It is always easier to report the
facts, but people seem to prefer to strain to posit explanations and
report them instead.  If the explanations are based on guesses about how
R is implemented, they will be useless; others will have to try to
figure out what the facts must have been to lead to such speculations.
Sometimes this is impossible.  But in any case, it is unnecessary work
for the ones trying to fix the problem.

   For example, suppose that on a data set which you know to be quite
large the command

     R> data.frame(x, y, z, monday, tuesday)

never returns.  Do not report that 'data.frame()' fails for large data
sets.  Perhaps it fails when a variable name is a day of the week.  If
this is so then when others got your report they would try out the
'data.frame()' command on a large data set, probably with no day of the
week variable name, and not see any problem.  There is no way in the
world that others could guess that they should try a day of the week
variable name.

   Or perhaps the command fails because the last command you used was a
method for '"["()' that had a bug causing R's internal data structures
to be corrupted and making the 'data.frame()' command fail from then on.
This is why others need to know what other commands you have typed (or
read from your startup file).

   It is very useful to try and find simple examples that produce
apparently the same bug, and somewhat useful to find simple examples
that might be expected to produce the bug but actually do not.  If you
want to debug the problem and find exactly what caused it, that is
wonderful.  You should still report the facts as well as any
explanations or solutions.  Please include an example that reproduces
(e.g., <https://en.wikipedia.org/wiki/Reproducibility>) the problem,
preferably the simplest one you have found.

   Invoking R with the '--vanilla' option may help in isolating a bug.
This ensures that the site profile and saved data files are not read.

   Before you actually submit a bug report, you should check whether the
bug has already been reported and/or fixed.  First, try the "Show open
bugs new-to-old" or the search facility on
<https://bugs.R-project.org/>.  Second, consult
<https://svn.R-project.org/R/trunk/doc/NEWS.Rd>, which records changes
that will appear in the _next_ release of R, including bug fixes that do
not appear on the Bug Tracker.  Third, if possible try the current
r-patched or r-devel version of R. If a bug has already been reported or
fixed, please do not submit further bug reports on it.  Finally, check
carefully whether the bug is with R, or a contributed package.  Bug
reports on contributed packages should be sent first to the package
maintainer, and only submitted to the R-bugs repository by package
maintainers, mentioning the package in the subject line.

   A bug report can be generated using the function 'bug.report()'.  For
reports on R this will open the Web page at
<https://bugs.R-project.org/>: for a contributed package it will open
the package's bug tracker Web page or help you compose an email to the
maintainer.

   There is a section of the bug repository for suggestions for
enhancements for R labelled 'wishlist'.  Suggestions can be submitted in
the same ways as bugs, but please ensure that the subject line makes
clear that this is for the wishlist and not a bug report, for example by
starting with 'Wishlist:'.

   Comments on and suggestions for the Windows port of R should be sent
to <R-windows@R-project.org>.

   Corrections to and comments on message translations should be sent to
the last translator (listed at the top of the appropriate '.po' file) or
to the translation team as listed at
<https://developer.R-project.org/TranslationTeams.html>.

10 Acknowledgments
******************

Of course, many many thanks to Robert and Ross for the R system, and to
the package writers and porters for adding to it.

   Special thanks go to Doug Bates, Peter Dalgaard, Paul Gilbert,
Stefano Iacus, Fritz Leisch, Jim Lindsey, Thomas Lumley, Martin
Maechler, Brian D. Ripley, Anthony Rossini, and Andreas Weingessel for
their comments which helped me improve this FAQ.

   More to come soon ...