[R] Packages: What expressions are allowed outside/before .First.lib?

[Henrik Bengtsson]

Sorry for this long message. The two first questions asks for general coding
standard when creating packages, the third one asks if the .R files are
concatenated in lexical order or not, the fourth question is a "how-to"
question. All questions are related. If there is a reference where I can
read about this please tell me, because then I could repost a more
restricted set of questions.

Q1. When loading pakages, if exists, the .First.lib function will be called
in the end. Is it correct that it is only in .First.lib one should have
"executable" code? Is it also true that "outside" .First.lib, only function
declarations and objects assignments such as '.conflicts.OK <- TRUE' should
be used?

Q2. Is it considered ok to use attr() outside .First.lib? I would like set
an attribute 'modifiers' to all my functions as they are declared. I have
tried it and it work. This is an example of how one of my generated package
files looks like:

  callMe0 <- function(x)
    cat("hey\n")
  attr(callMe0, "modifiers") <- c("private")

  callMe <- function(x)
    callMe0(x)
  attr(callMe, "modifiers") <- c("public")

I know that I could put the attr() assignments in the .First.lib function,
but since I/we write code in (many) separated .R files, which are
concatenated when the library is build, it would be doomed to miss one or
have one to many. So, is the above code considered to be alright? (The
'private' and 'public' attributes are used to "hide" some of the functions
from the end user.)


Q3. After having used the attr() approach for a while, we wanted to add some
control of the assignment of 'modifiers'. I wrote a function "modifiers<-"
that assert that the values are valid before setting the attribute. Now the
code would look like

  callMe0 <- function(x)
    cat("hey\n")
  modifiers(callMe0) <- c("private")

  callMe <- function(x)
    callMe0(x)
  modifiers(callMe) <- c("public")

I think this code is a little bit more readable, but bullet-proof in the
sence that unknown modifiers can not be set. But, by doing this I know I am
in deep water; to load this library I rely on the fact that "modifiers<-" is
already declared. It looks like the library source file is created by
concatenation the .R files in lexical order and one can force the
declaration of "modifiers<-" to appear first by putting it in a file called
"000.R" (zeros). However, I don't know if one can rely on this?!


Q4. What I really want to do is to declare "modifiers<-" in another package,
lets call it, 'myrootlib', which all other libraries we write relies on.
Trying

  require(myrootlib)  # Declares "modifiers<-"

  callMe0 <- function(x)
    cat("hey\n")
  modifiers(callMe0) <- c("private")

  callMe <- function(x)
    callMe0(x)
  modifiers(callMe) <- c("public")

won't work! The reason for this seems to be that the package 'myrootlib' is
not loaded (evaluated) until afterwards. This brings up the issues asked in
Q1 and Q2. Is there another way to do it or do we have to go back to the
approaches explained in Q3, Q2 or even Q1?

Thank you so much!

Henrik Bengtsson

Lund University/UC Berkeley
hb at maths.lth.se

PS. Any questions I ever ask about [R] and possible improvements for it, is
to improve my (and hopefully others) understanding and never to complain. I
appreciate all the work done by all the contributors and testers. Thanks.
DS.

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._