R NA-skipping modes ignore both NA and NaN: > mean(c(0, 1, NA), na.rm='true') [1] 0.5 > mean(c(0, 1, NaN), na.rm='true') [1] 0.5
Although without skipping enabled, the result does reflect the original NA/NaN: > mean(c(0, 1, NaN), na.rm='false') [1] NaN > mean(c(0, 1, NA), na.rm='false') [1] NA Given the factor of 10 slowdown in checking for NA mentioned by Jordi, I feel that distinguishing between the two for NA-skipping is not warranted, except perhaps as an additional selectable option. We could remove the shadowing concern of using the NaN toolbox by adding a mode like na.rm='true' to the main functions that would switch to the NaN toolbox functionality, while preserving by default the treatment of NaN as an invalid value. There could be a global flag that could be set by users to change the default to NA-skipping so that existing code that uses NaN-toolbox functions would not need to add a flag to use the new mode at each invocation of functions like mean. ------------------------------------------------------------------------------ Virtualization & Cloud Management Using Capacity Planning Cloud computing makes use of virtualization - but cloud computing also focuses on allowing computing to be delivered as a service. http://www.accelacomm.com/jaw/sfnl/114/51521223/ _______________________________________________ Octave-dev mailing list Octave-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/octave-dev
