I have a data frame and I want to learn how the summary generates it's information. Specifically, how does summary generate a count for the number of elements in each level of a factor. I can use summary, but I want to learn how to work with factors better. When I try ?summary, I just get the general info. Is this impossible because it is in bytecode?
Reading Source Code for R Functions – How to Guide
rr-faq
Related Solutions
The R source code of pnorm
is:
function (q, mean = 0, sd = 1, lower.tail = TRUE, log.p = FALSE)
.Call(C_pnorm, q, mean, sd, lower.tail, log.p)
So, technically speaking, typing "pnorm" does show you the source code. However, more usefully: The guts of pnorm
are coded in C, so the advice in the previous question view source code in R is only peripherally useful (most of it concentrates on functions hidden in namespaces etc.).
Uwe Ligges's article in R news, Accessing the Sources (p. 43), is a good general reference. From that document:
When looking at R source code, sometimes calls to one of the following functions show up:
.C()
,.Call()
,.Fortran()
,.External()
, or.Internal()
and.Primitive()
. These functions are calling entry points in compiled code such as shared objects, static libraries or dynamic link libraries. Therefore, it is necessary to look into the sources of the compiled code, if complete understanding of the code is required. ... The first step is to look up the entry point in file ‘$R HOME/src/main/names.c’, if the calling R function is either.Primitive()
or.Internal()
. This is done in the following example for the code implementing the ‘simple’ R functionsum()
.
(Emphasis added because the precise function you asked about (sum
) is covered in Ligges's article.)
Depending on how seriously you want to dig into the code, it may be worth downloading and
unpacking the source code as Ligges suggests (for example, then you can use command-line tools
such as grep
to search through the source code). For more casual inspection, you can view
the sources online via the R Subversion server or Winston Chang's github mirror or the R-svn github mirror (links here are specifically to src/nmath/pnorm.c
). (Guessing the right place to look, src/nmath/pnorm.c
, takes some familiarity with the structure of the R source code.)
mean
and sum
are both implemented in summary.c.
UseMethod("t")
is telling you that t()
is a (S3) generic function that has methods for different object classes.
The S3 method dispatch system
For S3 classes, you can use the methods
function to list the methods for a particular generic function or class.
> methods(t)
[1] t.data.frame t.default t.ts*
Non-visible functions are asterisked
> methods(class="ts")
[1] aggregate.ts as.data.frame.ts cbind.ts* cycle.ts*
[5] diffinv.ts* diff.ts kernapply.ts* lines.ts
[9] monthplot.ts* na.omit.ts* Ops.ts* plot.ts
[13] print.ts time.ts* [<-.ts* [.ts*
[17] t.ts* window<-.ts* window.ts*
Non-visible functions are asterisked
"Non-visible functions are asterisked" means the function is not exported from its package's namespace. You can still view its source code via the :::
function (i.e. stats:::t.ts
), or by using getAnywhere()
. getAnywhere()
is useful because you don't have to know which package the function came from.
> getAnywhere(t.ts)
A single object matching ‘t.ts’ was found
It was found in the following places
registered S3 method for t from namespace stats
namespace:stats
with value
function (x)
{
cl <- oldClass(x)
other <- !(cl %in% c("ts", "mts"))
class(x) <- if (any(other))
cl[other]
attr(x, "tsp") <- NULL
t(x)
}
<bytecode: 0x294e410>
<environment: namespace:stats>
The S4 method dispatch system
The S4 system is a newer method dispatch system and is an alternative to the S3 system. Here is an example of an S4 function:
> library(Matrix)
Loading required package: lattice
> chol2inv
standardGeneric for "chol2inv" defined from package "base"
function (x, ...)
standardGeneric("chol2inv")
<bytecode: 0x000000000eafd790>
<environment: 0x000000000eb06f10>
Methods may be defined for arguments: x
Use showMethods("chol2inv") for currently available ones.
The output already offers a lot of information. standardGeneric
is an indicator of an S4 function. The method to see defined S4 methods is offered helpfully:
> showMethods(chol2inv)
Function: chol2inv (package base)
x="ANY"
x="CHMfactor"
x="denseMatrix"
x="diagonalMatrix"
x="dtrMatrix"
x="sparseMatrix"
getMethod
can be used to see the source code of one of the methods:
> getMethod("chol2inv", "diagonalMatrix")
Method Definition:
function (x, ...)
{
chk.s(...)
tcrossprod(solve(x))
}
<bytecode: 0x000000000ea2cc70>
<environment: namespace:Matrix>
Signatures:
x
target "diagonalMatrix"
defined "diagonalMatrix"
There are also methods with more complex signatures for each method, for example
require(raster)
showMethods(extract)
Function: extract (package raster)
x="Raster", y="data.frame"
x="Raster", y="Extent"
x="Raster", y="matrix"
x="Raster", y="SpatialLines"
x="Raster", y="SpatialPoints"
x="Raster", y="SpatialPolygons"
x="Raster", y="vector"
To see the source code for one of these methods the entire signature must be supplied, e.g.
getMethod("extract" , signature = c( x = "Raster" , y = "SpatialPolygons") )
It will not suffice to supply the partial signature
getMethod("extract",signature="SpatialPolygons")
#Error in getMethod("extract", signature = "SpatialPolygons") :
# No method found for function "extract" and signature SpatialPolygons
Functions that call unexported functions
In the case of ts.union
, .cbindts
and .makeNamesTs
are unexported functions from the stats
namespace. You can view the source code of unexported functions by using the :::
operator or getAnywhere
.
> stats:::.makeNamesTs
function (...)
{
l <- as.list(substitute(list(...)))[-1L]
nm <- names(l)
fixup <- if (is.null(nm))
seq_along(l)
else nm == ""
dep <- sapply(l[fixup], function(x) deparse(x)[1L])
if (is.null(nm))
return(dep)
if (any(fixup))
nm[fixup] <- dep
nm
}
<bytecode: 0x38140d0>
<environment: namespace:stats>
Functions that call compiled code
Note that "compiled" does not refer to byte-compiled R code as created by the compiler package. The <bytecode: 0x294e410>
line in the above output indicates that the function is byte-compiled, and you can still view the source from the R command line.
Functions that call .C
, .Call
, .Fortran
, .External
, .Internal
, or .Primitive
are calling entry points in compiled code, so you will have to look at sources of the compiled code if you want to fully understand the function. This GitHub mirror of the R source code is a decent place to start. The function pryr::show_c_source
can be a useful tool as it will take you directly to a GitHub page for .Internal
and .Primitive
calls. Packages may use .C
, .Call
, .Fortran
, and .External
; but not .Internal
or .Primitive
, because these are used to call functions built into the R interpreter.
Calls to some of the above functions may use an object instead of a character string to reference the compiled function. In those cases, the object is of class "NativeSymbolInfo"
, "RegisteredNativeSymbol"
, or "NativeSymbol"
; and printing the object yields useful information. For example, optim
calls .External2(C_optimhess, res$par, fn1, gr1, con)
(note that's C_optimhess
, not "C_optimhess"
). optim
is in the stats package, so you can type stats:::C_optimhess
to see information about the compiled function being called.
Compiled code in a package
If you want to view compiled code in a package, you will need to download/unpack the package source. The installed binaries are not sufficient. A package's source code is available from the same CRAN (or CRAN compatible) repository that the package was originally installed from. The download.packages()
function can get the package source for you.
download.packages(pkgs = "Matrix",
destdir = ".",
type = "source")
This will download the source version of the Matrix package and save the corresponding .tar.gz
file in the current directory. Source code for compiled functions can be found in the src
directory of the uncompressed and untared file. The uncompressing and untaring step can be done outside of R
, or from within R
using the untar()
function. It is possible to combine the download and expansion step into a single call (note that only one package at a time can be downloaded and unpacked in this way):
untar(download.packages(pkgs = "Matrix",
destdir = ".",
type = "source")[,2])
Alternatively, if the package development is hosted publicly (e.g. via GitHub, R-Forge, or RForge.net), you can probably browse the source code online.
Compiled code in a base package
Certain packages are considered "base" packages. These packages ship with R and their version is locked to the version of R. Examples include base
, compiler
, stats
, and utils
. As such, they are not available as separate downloadable packages on CRAN as described above. Rather, they are part of the R source tree in individual package directories under /src/library/
. How to access the R source is described in the next section.
Compiled code built into the R interpreter
If you want to view the code built-in to the R interpreter, you will need to download/unpack the R sources; or you can view the sources online via the R Subversion repository or Winston Chang's github mirror.
Uwe Ligges's R news article (PDF) (p. 43) is a good general reference of how to view the source code for .Internal
and .Primitive
functions. The basic steps are to first look for the function name in src/main/names.c
and then search for the "C-entry" name in the files in src/main/*
.
Best Answer
What we see when you type
summary
isThis is telling us that summary is a generic function and has many methods attached to it. To see what those methods are actually called we can try
Here we see all the methods associated with the
summary
function. What this means is that there is different code for when you call summary on an lm object than there is when you call summary on a data.frame. This is good because we wouldn't expect the summary to be conducted the same way for those two objects.To see the code that is run when you call summary on a data.frame you can just type
as shown in the methods list. You'll be able to examine it and study it and do whatever you want with the printed code. You mentioned that you were interested in factors so you will probably want to examine the output of
summary.factor
. Now you might notice that some of the methods printed had an asterisk (*) next to them which implies that they're non-visible. This essentially means that you can't just type the name of the function to try to view the code.However, if you're determined to see what the code actually is you can use the
getAnywhere
function to view it.Hopefully this helps you explore the code in R much more easily in the future.
For even more details you can view Volume 6/4 of The R Journal (warning, pdf) and read Uwe Ligge's "R Help Desk" section which deals with viewing the source code of R functions.