Thanks! Wouldn't it be better no to try to load it at all?

The point of using getNamespaceVersion was to detect the rare case when the loaded version is different from the on-disk version. If the package is not loaded, we can just keep the loaded version empty.

Anyway, I can merge this now, and change it later.

Maybe, I can do that instead, one sec, will update the PR.

Codecov Report

Merging #4 into master will not change coverage.
The diff coverage is 100%.

@@          Coverage Diff          @@
##           master     #4   +/-   ##
=====================================
  Coverage     100%   100%           
=====================================
  Files           8      8           
  Lines         173    174    +1     
=====================================
+ Hits          173    174    +1

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8de1163...3806e7c. Read the comment docs.

Ok updated to only get namespace versions for loaded namespaces.

Are the NAs printed? How does the printed output look?

I'll update the mocked test, don't worry about that.

Hmm NA actually causes errors from base::package_version() because it is not a valid package version, as done "". "0.0.0" works, but the output is kind of gross.

─ Packages ───────────────────────────────────────────────
 package           * version    date       source

 acepack             0.0.0 (!)  2016-10-29 CRAN (R 3.4.0)

 AnnotationDbi       0.0.0 (!)  2017-04-25 Bioconductor

 ape                 0.0.0 (!)  2017-02-14 CRAN (R 3.4.0)

 archive             0.0.0 (!)  2017-07-28 Github (jimhester/archive@a203312)

 assertive.base      0.0.0 (!)  2016-12-30 CRAN (R 3.4.0)

 assertthat          0.2.0      2017-04-11 CRAN (R 3.4.0)

Yeah, I am afraid we need to change the print method.

Somewhere around here:

If it is not loaded, just use the on-disk version.

Ok turns out if you use package_version(NA, strict = FALSE) we get an NA package version, which we can then handle in the print method, which looks nicer and should still catch the case where a loaded version is different than the installed version as before (3806e7c).

Great! Thanks!