Hi Bastien On Tue, Apr 9, 2013 at 7:06 PM, Bastien wrote: >> It is worth a small compatibility change: For a range with only empty >> fields it is now possible and necessary to choose different behaviors >> of vmean by adding the format specifiers E and/or N. > > I'll add this in the release notes. The wording of my paragraph quoted above is unfortunate, sorry. It is more that I made fail what is the same as a division by 0 instead of silently using 0 as the result for 0/0. Now one is forced to think about the mean value of an empty vector/list (it is not defined) and has to check for empty vector/list or consider EN, all explained with examples for Calc vmean in the manual now. I don't think such a detail is worth mentioning in the release notes. But in the release notes I would generally, not particularly because of the above, write: If empty fields are of interest it is recommended to reread the section "3.5.2 Formula syntax for Calc" of the manual because the description for the mode strings has been clarified and new examples have been added towards the end. > Maybe you could add a footnote in the manual for this? Do the attached patches clarify what you had in mind? And to repeat myself: I see the directory "testing" as part of the doc, in this case `test-org-table/empty-field'. ;-) Michael