Colleagues,
My interest is not in writing ad hoc functions (which I might use once to
analyze my data), but rather what I will call a system function that might be
part of a package. The lm function is a paradigm of what I call a system
function.
The lm function begins by processing the arguments passed to the function
(represented in the function as parameters, see code below.) Much of this
processing is only peripherally related to running a regression, but the code is
necessary to determine exactly what the user of the system function wants the
function to do. It would be helpful if there was a document that would describe
best practices when writing system functions, with clear explanations of what
each step in system function is designed to do and how the line accomplishes its
task. It would also be nice if the system function had documentation. I have
pushed my way through the lm function, and with the help of R help files, I have
come to understand how the function works, but this is not an efficient way to
learn best practices that should be used when writing a system function.
Perhaps there is a document that does what I would like to see done, but I do
not know of one.
John
lmlm
function (formula, data, subset, weights, na.action, method = "qr",
model = TRUE, x = FALSE, y = FALSE, qr = TRUE, singular.ok = TRUE,
contrasts = NULL, offset, ...)
{
ret.x <- x
ret.y <- y
cl <- match.call()
mf <- match.call(expand.dots = FALSE)
m <- match(c("formula", "data", "subset",
"weights", "na.action",
"offset"), names(mf), 0L)
mf <- mf[c(1L, m)]
mf$drop.unused.levels <- TRUE
mf[[1L]] <- quote(stats::model.frame)
mf <- eval(mf, parent.frame())
if (method == "model.frame")
return(mf)
else if (method != "qr")
warning(gettextf("method = '%s' is not supported. Using
'qr'",
method), domain = NA)
mt <- attr(mf, "terms")
y <- model.response(mf, "numeric")
w <- as.vector(model.weights(mf))
if (!is.null(w) && !is.numeric(w))
stop("'weights' must be a numeric vector")
offset <- model.offset(mf)
mlm <- is.matrix(y)
ny <- if (mlm)
nrow(y)
else length(y)
if (!is.null(offset)) {
if (!mlm)
offset <- as.vector(offset)
if (NROW(offset) != ny)
stop(gettextf("number of offsets is %d, should equal %d (number
of observations)",
NROW(offset), ny), domain = NA)
}
if (is.empty.model(mt)) {
x <- NULL
z <- list(coefficients = if (mlm) matrix(NA_real_, 0,
ncol(y)) else numeric(), residuals = y, fitted.values = 0 *
y, weights = w, rank = 0L, df.residual = if (!is.null(w)) sum(w !
0) else ny)
if (!is.null(offset)) {
z$fitted.values <- offset
z$residuals <- y - offset
}
}
else {
x <- model.matrix(mt, mf, contrasts)
z <- if (is.null(w))
lm.fit(x, y, offset = offset, singular.ok = singular.ok,
...)
else lm.wfit(x, y, w, offset = offset, singular.ok = singular.ok,
...)
}
class(z) <- c(if (mlm) "mlm", "lm")
z$na.action <- attr(mf, "na.action")
z$offset <- offset
z$contrasts <- attr(x, "contrasts")
z$xlevels <- .getXlevels(mt, mf)
z$call <- cl
z$terms <- mt
if (model)
z$model <- mf
if (ret.x)
z$x <- x
if (ret.y)
z$y <- y
if (!qr)
z$qr <- NULL
z
}
John David Sorkin M.D., Ph.D.
Professor of Medicine, University of Maryland School of Medicine;
Associate Director for Biostatistics and Informatics, Baltimore VA Medical
Center Geriatrics Research, Education, and Clinical Center;?
PI?Biostatistics and Informatics Core, University of Maryland School of Medicine
Claude D. Pepper Older Americans Independence Center;
Senior Statistician University of Maryland Center for Vascular Research;
Division of Gerontology and Paliative Care,
10 North Greene Street
GRECC (BT/18/GR)
Baltimore, MD 21201-1524
Cell phone 443-418-5382
________________________________________
From: Jorgen Harmse <JHarmse at roku.com>
Sent: Tuesday, January 7, 2025 1:47 PM
To: r-help at r-project.org; ikwsimmo at gmail.com; Bert Gunter; Sorkin, John;
jdnewmil at dcn.davis.ca.us
Subject: Re: Extracting specific arguments from "..."
Interesting discussion. A few things occurred to me.
Apologies to Iris Simmons: I mixed up his answer with Bert's question.
Bert raises questions about promises, and I think they are related to John
Sorkin's question. A big difference between R and most other languages is
that function arguments are computed lazily. match.call & substitute tell us
what expressions will be evaluated if function arguments are needed but not the
environments in which that will happen. The usual suspects are environment() and
parent.frame(), but parent.frame(k) & maybe even other environments are
possible. If you are really determined then I guess you can keep evaluating
match.call() in parent frames until you have accounted for all the inputs.
It's not clear to what extent John Sorkin is concerned about writing
functions as opposed to using functions. Lazy computation has advantages but
leads to some issues.
Exactly matching the function's default expression for an input is not
necessarily the same as omitting the input. The evaluation environment is
different.
If the caller uses an expression with side effects then there is no guarantee
that the side effects will happen. If there are side effects from two or more
inputs then the order is uncertain. (If an argument is not supplied and the
default has side effects then they might not happen either. However, I don't
know why the function writer would specify any side effect except stop(), and
then he or she has probably arranged for it to happen exactly when it should.)
If a default value depends on another input and that input is modified inside
the function then order of evaluation of inputs becomes important. Even if you
know exactly what you're doing when you write the function, you should make
it clear to future maintainers. An explicit call to force clarifies that the
input needs to be computed with the existing values of anything that is used in
the default, even if the code is refactored so that the value is not used
immediately. If you really want to modify another input before evaluating the
default then specify that in a comment.
Jeff Newmiller makes a good point. You can still change your mind about
inspecting a particular input without breaking old code that uses your function,
and you don?t necessarily need default values.
Old definition: f <- function(?) {<code that passes ? to other functions
and does some other things>}
New definition:
f <- function(?, a = <default expression, possibly stop()>)
{ <pass ?, a=a to another function>
<do something with the output>
}
OR
f <- function(?, a)
{ if (missing(a)) # OK, this becomes clunky if there are several such inputs
{ < pass ? to another function >}
else
{ <inspect or modify a> # Pitfall: Changing the order of evaluation may
break old code, but then the design was probably too devious in the first place.
<pass ?, a=a to another function>
}
<do something with the output>
}
Regards,
Jorgen Harmse.