I like to keep my R packages up to date, and in developing my own package, I want to stick to @Hadley's advice:
Generally, it’s always better to specify the version and to be conservative about which version to require. Unless you know otherwise, always require a version greater than or equal to the version you’re currently using.
So, I'll need some of those here in DESCRIPTION
:
Imports:
knitr (>= 1.13),
rmarkdown (>= 1.0)
Is there an existing tool to programmatically update the versions of dependencies based on the packages I currently have installed?
I know this is a small thing and I can do this manually, but I just know this is the kind of thing that is easily forgotten.
Using the locally installed version of a package as a minimum dependency also seems to make sense because I tested / build with those dependencies.
Is there a reason why I shouldn't do this?
This shld do what you ask (well, you still need to cut/paste the output into DESCRIPTION
:-)
#' Add curent version string to package dependencies
#'
#' Will \code{cat} out a cut/paste-able set of fields for a
#' \code{DESCRIPTION} file with minimum required versions for
#' each package based upon currently available package vesions
#' in CRAN.
#'
#' @param pkg package description, can be path or package name
#' @param fields fields to get & report dependencies for
#' @note R and the R version is NOT added to \code{Depends}
#' @examples
#' add_pkg_versions("qmethod")
#' add_pkg_versions("MASS")
#' \dontrun { # assumes you're in a pkg devel dir
#' add_pkg_versions()
#' }
add_pkg_versions <- function(pkg=".",
fields=c("Depends", "Imports", "LinkingTo", "Suggests")) {
require(purrr)
walk(c("dplyr", "tools", "stringi", "devtools"), require, character.only=TRUE)
stopifnot(is_scalar_character(pkg), pkg != "")
fields <- match.arg(fields, c("Depends", "Imports", "LinkingTo", "Suggests"),
several.ok=TRUE)
avail <- as_data_frame(available.packages())
if (pkg == ".") {
pkg_deps <- unclass(as_data_frame(read.dcf(file.path(package_file(), "DESCRIPTION"))))
pkg <- pkg_deps$Package
map(fields, ~stri_split_lines(pkg_deps[[.]])) %>%
map(function(x) {
if (length(x) > 0) {
unlist(x) %>%
stri_replace_all_regex(" \\(.*$|,", "") %>%
discard(`%in%`, c("", "R"))
} else { x }
}) -> pkg_deps
names(pkg_deps) <- fields
} else {
pkg_deps <- map(fields, ~flatten_chr((package_dependencies(pkg, which=.))))
names(pkg_deps) <- fields
}
pkg_deps <- discard(pkg_deps, function(x) {length(x)==0})
map(pkg_deps, function(x) {
non_base <- filter(avail, Package %in% x)
base <- setdiff(x, non_base$Package)
non_base %>%
mutate(pv=sprintf("%s (>= %s)", Package, Version)) %>%
select(pv) %>%
flatten_chr() -> pkg_plus_version
sort(c(pkg_plus_version, base))
}) -> pkg_deps
cat("Package: ", pkg, "\n", sep="")
walk(names(pkg_deps), function(x) {
cat(x, ":\n", sep="")
sprintf(" %s", pkg_deps[[x]]) %>%
paste0(collapse=",\n") %>%
cat()
cat("\n")
})
}
One of your packages:
add_pkg_versions("qmethod")
Package: qmethod
Imports:
digest (>= 0.6.10),
GPArotation (>= 2014.11-1),
knitr (>= 1.13),
methods,
psych (>= 1.6.6),
tools,
xtable (>= 1.8-2)
Just to show edge cases are handled:
add_pkg_versions("MASS")
Package: MASS
Depends:
graphics,
grDevices,
stats,
utils
Imports:
methods
Suggests:
lattice (>= 0.20-33),
nlme (>= 3.1-128),
nnet (>= 7.3-12),
survival (>= 2.39-5)