On 18/05/2017 10:34 AM, Martin Watts wrote:
I have created a package (which I host on a local server) that I have set
up to depend on exact versions of other packages, i.e my DESCRIPTION file
looks like this:
Package: myPackage
Depends:
R (>= 3.2.5)
Imports:
dplyr (== 0.5.0),
lazyeval (== 0.2.0),
lubridate (== 1.6.0),
magrittr (== 1.5),
reshape2 (== 1.4.2)
My expectation is that when I run install.packages('myPackage') that each
of these specific dependency versions will be installed, but this is not
the case. It looks like if a dependent package is not present at all then
it is installed to the correct version, but when it is present at a LATER
version it is not rolled-back to the desired version.
That could leave your library in an unusable state. Suppose one package
requires reshape2 1.4.2, and a later install requires reshape2 1.4.1.
Then the second install shouldn't roll back the version of reshape2 or
it would break the first package.
It's also wrong if it allows the second package to install with 1.4.2 in
place; that sounds like a bug.
Is there a way that we can force all requested versions of these packages
to be installed? We would like this behaviour as we'd like a way to
guarantee that all R environments are exactly the same when running this
code.
Uninstall everything before installing the new one.
There are also a couple of efforts for reproducing exact environments.
Can't think of the names right now.
Note that I've been testing using devtools::install, but I assume that this
mirrors the behaviour of install.packages.
You should test the code you're complaining about, not some other code.
All we know from what you've reported is that devtools::install has a
bug. We don't know that install.packages has the same bug.
Duncan Murdoch
Also posted on SO: http://stackoverflow.com/questions/44028355/r-force-
download-of-specific-package-versions-in-description-file
Thanks,
Martin
[[alternative HTML version deleted]]
______________________________________________
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
______________________________________________
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
