Talk:Numerical tower
From Wikipedia, the free encyclopedia
Where do IEEE 754 floats fit in this tower? Do -0.0, +Inf and NaN fit in with "Real"? Or (in my opinion) is this tower an artificial construct with little basis in practical computing? —The preceding unsigned comment was added by 213.114.96.51 (talk • contribs) .
- According to R5RS, IEEE floats may be used to represent numbers inexactly. Using floats for real numbers seems to be the usual practice. By that standard, the presence of ±Inf, NaN and −0 would be implementation-dependent. For example, Chicken implements them, whereas, according to [1], Scheme 48 does not. More generally, the hierarchy just expresses the fact that, but for minor issues,
, which is something that non-sadistic language designers will usually not dispute by specifying something different. What you choose as an implementation of those things is up to you (although, if I may voice an opinion here, it is generally a good idea to implement both the C99-style types and arbitrary precision types). —xyzzyn 14:32, 7 June 2006 (UTC)
