I pride myself on having an extensive and comprehensive test suite for the survival package. However, I don't have a good procedure for testing the plotting routines. (Forget "good", I don't even have a third-rate procedure for this.) This was brought home with the latest release; within 24 hours I got a note about a problem. They y-axis range for a plot was calculated incorrectly, giving no warning or error, just a graph with 2/3 of the data outside the plot window. It was a side effect of a change for another reason, and of course I had checked out that consequence thoroughly. Any good ideas for this issue? Terry Therneau
Hi, Winston Chang designed the vtest package for a similar purpose, https://github.com/wch/ggplot2/wiki/Visual-test-system https://github.com/wch/vtest I don't know if / how well it works outside the ggplot2 framework, but it's pretty slick. HTH, baptiste On 22 January 2014 11:35, Therneau, Terry M., Ph.D. <therneau@mayo.edu>wrote:> I pride myself on having an extensive and comprehensive test suite for the > survival package. > > However, I don't have a good procedure for testing the plotting routines. > (Forget "good", I don't even have a third-rate procedure for this.) This > was brought home with the latest release; within 24 hours I got a note > about a problem. They y-axis range for a plot was calculated incorrectly, > giving no warning or error, just a graph with 2/3 of the data outside the > plot window. It was a side effect of a change for another reason, and of > course I had checked out that consequence thoroughly. > > Any good ideas for this issue? > > Terry Therneau > > ______________________________________________ > R-devel@r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-devel >[[alternative HTML version deleted]]
On 22/01/2014 14:35, Therneau, Terry M., Ph.D. wrote:> I pride myself on having an extensive and comprehensive test suite for > the survival package. > > However, I don't have a good procedure for testing the plotting > routines. (Forget "good", I don't even have a third-rate procedure for > this.) This was brought home with the latest release; within 24 hours I > got a note about a problem. They y-axis range for a plot was calculated > incorrectly, giving no warning or error, just a graph with 2/3 of the > data outside the plot window. It was a side effect of a change for > another reason, and of course I had checked out that consequence > thoroughly. > > Any good ideas for this issue? > > Terry Therneau >The R tests are comparisons of uncompressed PDF files (which are almost text files). See the 'tests' directory. Paul Murrell gave a talk about another method at an early DSC: https://www.stat.auckland.ac.nz/~paul/Talks/qa.pdf and now has a graphicsQC package. -- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595