[Bioc-devel] Announcing newtest coverage shields

[Henrik Bengtsson]

On Wed, Jun 24, 2015 at 5:42 PM, Dan Tenenbaum <dtenenba at fredhutch.org> wrote:
> Hello Bioconductors,
>
> We're pleased to announce a new shield to join the ones we rolled out in May.
>
> The new shield measures test coverage of a package, as determined by Jim Hester's covr package.
>
> Coverage is a measure of the degree to which package code is tested by your unit tests (https://en.wikipedia.org/wiki/Code_coverage). If you don't know what unit tests are, read our guidelines at http://bioconductor.org/developers/how-to/unitTesting-guidelines/ .
>
> These shields are on all package landing pages for software packages in release and devel. An example shield can be seen at
>
> http://bioconductor.org/packages//Biobase/
>
> It links to a detailed coverage report at https://codecov.io/ .
>
> If package coverage cannot be determined (shield value is 'unknown'), the shield links to a section of
> http://bioconductor.org/developers/how-to/unitTesting-guidelines/#coverage explaining why
> this might be.
>
> Note that the coverage calculation happens on our linux build machines only and is not run as part of the nightly builds, but it is run several times a week. Only packages whose code has changed since the last calculation are run through covr.
>
> We hope this shield motivates package developers to add unit tests (if they don't have them already) and improve their package's unit test coverage. Refer to the covr
> documentation (http://cran.r-project.org/web/packages/covr/README.html) for more
> information on how to do this.
>
> Questions and comments are welcome as always on the bioc-devel list.

[snare drums] ... hi-hat!

Thank you very much for adding this.  For folks who yet haven't looked
into code coverage - it's extremely useful:

* You get line-by-line coverage estimates for your R code, e.g.
  https://codecov.io/github/Bioconductor-mirror/DNAcopy/R/segment.R

* Also for your native code, e.g.
  https://codecov.io/github/Bioconductor-mirror/DNAcopy/src/changepoints.f?ref=master

* The line-to-line reports makes it very easy to design new tests.

My experience from turning uncovered ("red") code lines into covered
("green") is that you are quite likely to discover a few more bugs
along the way.  I'd say it's one of the most efficient ways to find
unknown bugs that I ever used.  A useful rule of thumb is to always
make sure that the code coverage never decreases whenever a new
version is released.

/Henrik

>
> Thanks,
> Dan
>
> _______________________________________________
> Bioc-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/bioc-devel