[R] Light Libraries
Uwe Ligges
ligges at statistik.tu-dortmund.de
Tue Feb 26 15:35:30 CET 2013
Perhaps you want to tell people where the advantage is comparing it with
the very well established R package system.
Best,
Uwe Ligges
On 26.02.2013 02:43, ivo welch wrote:
> Dear R-Help group:
>
> I have been tinkering with how I want my personal standard library
> functions to look like. They are not designed to be professional and
> heavyweight, but lightweight. There are probably dozens of little bugs,
> because I don't know or have not properly taken care of a variety of
> internal R code issues. still, I like how this ended up, and there is no
> learning curve, so I thought I would share it.
>
> I have put all my functions into a directory ~/src/iaw/R/ . In my standard
> .Rprofile, I thus first added a list of my libraries (well, I have just
> one) and invoke it:
>
> options(strict="very")
> Libdirs <- c("~/src/iaw/R/")
> for (libdir in Libdirs) {
> source( Sys.glob(paste0(libdir, "Rprofile"))) )
> }
>
>
> (I prefer mnemonics to numbers. it is 2013. why is it text(..., pos="1"),
> when it should be text(...,pos="east")? "East", "E" should be
> abbreviations for it. Then we could also use NNW...but I am getting
> distracted...)
>
> in each of my "light" libraries, I now have an Rprofile file that looks
> like this
>
>
> library(compiler)
>
> cached <- paste0(libdir, "/library.Rdata")
>
> if (file.exists(cached) & all( file.info(cached)$mtime >
> file.info(Sys.glob(paste0(libdir,"/*.R")))$mtime
> ) ) {
> load(cached)
> cat("Loaded", cached, "\n")
> } else {
>
> Rprofile <- Sys.glob(paste0(libdir, "*.R"))
> for (n in Rprofile) {
> source(n)
> }
>
> ## libraries that I need to have in order to be able to compile
> library(utils)
> library(parallel)
> library(stats)
> library(graphics)
> library(grDevices)
>
> for (n in ls()) {
> (is.function(.GlobalEnv[[n]]))%or% next
> ##if ((n %in% c("n", "Rprofile", "cached"))) next
> cat("['", n, "']\n", sep="")
> .GlobalEnv[[n]] <- cmpfun(.GlobalEnv[[n]])
> }
>
> save.image(file=cached)
> cat("Saved", cached, "\n")
> }
>
>
> the basic organizational idea now is to stick each R function into its own
> .R file. the Rprofile code makes sure that whenever I change a function in
> the ~src/iaw/*.R directory, the library is rebuilt (all functions are
> recompiled and then saved into an .Rdata file). this is very fast in my
> case; if it were not, I could add some intelligence. all of this
> could/should be stuck into a universal function
> ("library.light(directoryname)"), but because I only have one library for
> now, it can just live in the Rprofile.
>
> with this organization, it is now also easy to keep vignettes, latex text,
> other code, etc., in the same directory. they will just be ignored because
> they do not end with .R.
>
>
> the actual functions follow a format that is different from existing
> documentation systems, incl Hadley's oxygen, but designed to plug in
> (eventually) into the standard R manual and help system.
>
> PREAMBLE <- c(
> doc = '
>
> @TITLE lagseries
>
> @AUTHOR ivo.welch at gmail.com
>
> @DATE Feb 25, 2013
>
> @DESCRIPTION
>
> "lagseries" takes a vector and shift its contents numlags items to the
> left, filling in appropriate missing values to retain the length of the
> vector. If panelid is named, then lagged value from another panelid will
> not be assigned to be the lag. (Usually, the panelid will be the firm id,
> and the panel must be sorted by firmid. Naturally, it makes little sense to
> use this unless the observations are also sorted by the time of the
> observation. This is, after all, a lagseries function.)
>
> @USAGE lagseries( seriesin, numlags =1, panelid=NULL)
>
> @ARGUMENTS
>
> seriesin: a numeric vector
>
> numlags: an integer, can be negative
>
> panelid: an optional panel id
>
> @DETAILS
>
> None
>
> @SEEALSO
>
> leadseries, chgseries, pchgseries, compoundseries
>
> @EXAMPLES
>
> x <- rnorm(10)
> xlag <- lagseries(x,2)
> lm( x ~ xlag )
>
> d <- data.frame( x <- c( rnorm(20), runif(30), rcauchy(40) ),
> who= c( rep("firm1",20), rep("firm2", 30),
> rep("firm3",40)),
> year= c( 1961:1980, 1971:2000, 1971:2010 ) )
>
> lagd <- data.frame( x=lagseries(d$x, panelid=who), who=d$who,
> year=lagseries(d$x, panelid=who) )
>
> ',
> test = '
>
> all( lagseries( 1:6, 2, c(1,1,2,2,2,2) ) == c(NA,NA,NA,NA,3,4),
> na.rm=TRUE )
>
> ')
>
> ################################################################################################################################
> lagseries <- function (seriesin, numlags = 1, panelid = NULL) {
>
> if (!is.null(getOption("strict"))) {
> (is.null(seriesin)) %and% "Looks like you are trying to calc a
> lagseries from a NULL or non-existing series"
> (is.vector(seriesin, mode="any")) %or% "Your series is not a vector,
> but a {{class(seriesin)}}."
> (length(seriesin) > 1) %or% "Need more observations than
> {{length(seriesin)}}"
>
> (is.vector(numlags, mode="numeric", length=1)) %or% "numlags must be a
> simple integer, not {{numlags}}."
>
> (is.null(panelid) | (is.vector(seriesin, mode="any",
> length=length(seriesin)))) %or%
> "panel id must be NULL or a vector of same length as seriesin. right
> now it is {{class(panelid)}}"
> }
>
> (numlags == 0) %and% return(seriesin)
>
> if (numlags < 0) return(leadseries(seriesin, -numlags, panelid))
>
> rv <- c(rep(NA, numlags), seriesin[1:(length(seriesin) - numlags)])
>
> if (is.null(panelid)) return(rv)
>
> (all(panelid >= lagseries(panelid), na.rm = TRUE)) %or% "The panel is not
> sorted upwards by panel id"
> ifelse(panelid != lagseries(panelid, numlags), NA, rv)
> }
>
>
> I think this code looks nicer than Hadley Wickham's way of marking up the
> docs, but this is obviously a matter of taste. Unlike me, Hadley knows
> what he is doing. Still, I wouldn't mind if Hadley adopted a second
> optional format like this in oxygen3. Writing an R function that parses
> this preamble for every .R file should be easy. Writing an ESS parser with
> some more intelligence to understand that any R file that begins with
> preamble <- c(doc='') is documentation is probably doable as well, but
> emacs hacking is way beyond me. Writing code to step through the tests is
> also easy. Moreover, the preamble could hold other useful info (such as a
> minimum version number) if need be.
>
> note that, in my lagseries function, I am trying to be anal about the input
> checking, but I am careless about output checking.
>
> I wish getOption("strict") would also enable internal R checking, too, but
> c'est la vis.
>
>
> in case someone wants to try this out, here is the sketch of my backup
> routines
>
> ## if someone from r-help wants to try it out, here are some background
> routines:
> original.is.vector <- is.vector
> is.vector <- function( x, mode ="any", length =(-1) ) {
> (original.is.vector(x, mode=mode)) %or% return(FALSE)
> ((length<0) | (length(x)==length)) %or% return(FALSE)
> TRUE
> }
>
>
> ## abort.estring should:
> ## [a] add the name of the preceding invoking function at the start of
> the error message, preferably with source line number
> ## so, the user would see an error like
> ## * lagseries:52:: Need more observations than 1*
> ## [b] evaluate every {{ }} construct and insert output into the string
> ## [c] abort
>
> "%or%" <- function (e1, e2) {
> if (!e1) { if (is.character(e2)) abort.estring(e2) else eval(e2) }
> }
>
>
> I do not know whether it is possible to build an abort.estring function
> that does what I want, but R seems flexible enough to do almost anything.
> I have a sketch of [b], thanks to Neal Fultz, but not of [a].
>
> I hope this organizational design helps some others.
>
> regards,
>
> /iaw
> ----
> Ivo Welch (ivo.welch at gmail.com)
> http://www.ivo-welch.info/
>
> <http://www.ivo-welch.info/>
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
More information about the R-help
mailing list