R devel - Sep 2007 - RUnit testing with R CMD check and more (was: R CMD check recursive copy of tests/)

Hello!

I have looked at Martin's recent work on "merging" unit tests with
R CMD check,
fused it with my and Tony's contributions and developed it a bit further.
Take
a look into the attached demo package on how it works.

In essence:

PKG/inst/RUnitTests
          |- runRUnitTests.R (script to run RUnit tests)
          \- ...             (any RUnit test file)

PKG/R/runRUnitTests.R        (wrapper function to call
PKG/inst/RUnitTests/runRUnitTests.R)
PKG/man/runRUnitTests.Rd     (help with examples! section)

The following can be achieved:

* useRs can run RUnit tests from within R after the package has been installed
  on their machine - this info is also given in PKG-package.Rd file to encourage
  useRs to test the package

* developeRs can run RUnit tests with:

  - R CMD check since examples in PKG/man/runRUnitTests.Rd are tested to work
    (test failures or R errors stop the R CMD check!)

  - launcing the script PKG/inst/RUnitTests/runRUnitTests.R with Rscript or
    R CMD check

I would like to hear any critiques, comments, suggestions, ... on above. Martin
mentioned (privately) that filenames have to much upper case letter, but I
wanted
to emphasize that the main work is being done by RUnit package - not developed
by
me!

Finnally, I would like to argue that from useRs point of view test and unit test
are the same thing and I believe that R would benefit if *all* tests could be
available in similar manner as above. This of course calls for the modification
of
packaging, testing, etc., but could provide more feedback (sure some of it will
be
bogous) on testing on various configurations.

With regards, Gregor

Gorjanc,

One important scenario I use during development is to run selected
unit tests, typically found in a single file. Usually this is
preceeded by INSTALL rather than check, as I optimistically think that
my changes will be an improvement and hence installing the package is
a starting point for the next iteration.

(another) Martin

Gorjanc Gregor <Gregor.Gorjanc at bfro.uni-lj.si> writes:
> Hello!
>
> I have looked at Martin's recent work on "merging" unit tests
with R
> CMD check, fused it with my and Tony's contributions and developed
> it a bit further. Take a look into the attached demo package on how
> it works.
>
> In essence:
>
> PKG/inst/RUnitTests
>           |- runRUnitTests.R (script to run RUnit tests)
>           \- ...             (any RUnit test file)
>
> PKG/R/runRUnitTests.R        (wrapper function to call
PKG/inst/RUnitTests/runRUnitTests.R)
> PKG/man/runRUnitTests.Rd     (help with examples! section)
>
> The following can be achieved:
>
> * useRs can run RUnit tests from within R after the package has been
> installed on their machine - this info is also given in
> PKG-package.Rd file to encourage useRs to test the package
>
> * developeRs can run RUnit tests with:
>
>   - R CMD check since examples in PKG/man/runRUnitTests.Rd are
>   tested to work (test failures or R errors stop the R CMD check!)
>
>   - launcing the script PKG/inst/RUnitTests/runRUnitTests.R with
>   Rscript or R CMD check
>
> I would like to hear any critiques, comments, suggestions, ... on
> above. Martin mentioned (privately) that filenames have to much
> upper case letter, but I wanted to emphasize that the main work is
> being done by RUnit package - not developed by me!
>
> Finnally, I would like to argue that from useRs point of view test
> and unit test are the same thing and I believe that R would benefit
> if *all* tests could be available in similar manner as above. This
> of course calls for the modification of packaging, testing, etc.,
> but could provide more feedback (sure some of it will be bogous) on
> testing on various configurations.
>
> With regards, Gregor
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
-- 
Martin Morgan
Bioconductor / Computational Biology
http://bioconductor.org

Hi!
> (another) Martin wrote:
> One important scenario I use during development is to run selected
> unit tests, typically found in a single file. Usually this is
> preceeded by INSTALL rather than check, as I optimistically think that
> my changes will be an improvement and hence installing the package is
> a starting point for the next iteration.
Yes, I also do that very often, especially if I work only on one function
and I only want to test if it works. With this in mind there is a Makefile
in PKG/inst/RUnitTests (this is based largely on work probably also by
you and other "BioC folk"), which provides handy

make # run unit tests
make all # install and run unit tests

Though, this runs all unit tests.

Gregor