On Apr 7, 2006, at 9:43 AM, Jacques Carette wrote:
[Lots of stuff about IEEE 754] Is this an H' worthy item? It is worth taking a serious look in conjunction with completely redoing
Robert Dockins wrote: the Num class. Minor tweaking of the behaviour on NaNs (which requires a large amount of background work) does not seem to be worthwhile on its own (in my not-at-all-humble-opinion, having done this before). If getting the Num class "right" is in the cards for H', which probably presupposes that some kinds of class alias support is also 'in', then looking at a proper implementation of IEEE 754 is worth it. Note that this would also need that arithmetic operators be able to throw signals, which can be either caught and 'handled' or turned (automatically) into exceptions. While 'signals' here can be _completely_ independent from O/S signals, it also intersects with current discussions on exceptions. Right now, I am waiting to see where the dust settles on Num, class aliases and exceptions. If it looks like all these things will happen, then I will throw my hat in to *do* something about proper IEEE 754 support for Float, Double, etc. Jacques