Re: Re: Re: Bug in interpretation of mma Series command?
- To: mathgroup at smc.vnet.net
- Subject: [mg2371] Re: [mg2287] Re: [mg2268] Re: Bug in interpretation of mma Series command?
- From: siegman at ee.stanford.edu (A. E. Siegman)
- Date: Sun, 29 Oct 1995 22:48:46 -0500
- Organization: Stanford University
As the original complainant on this topic, I'd just like to reiterate two quick points in the discussion of why having f1[x] * Series[f2[x]] evaluate to Series[f1[x] * f2[x]] (rather than remain unevaluated, for example, or do something, anything, else) still seems to me a bad thing. 1) It's _grossly_ unusual or unfamiliar behavior: I'd challenge the Wolfram (or other) respondents on this issue to name any other common situation where f1 * Operator[f2] is _expected_ by a user to behave as Operator[f1 * f2] (or maybe Operator[f1] * Operator[f2]), with Operator being an operator that really does something to the f's. The  symbols are just widely understood to mark the bounds of what (any) Operator operates on. 2) It can lead to, indeed promotes, _large_ undetected errors: If I understand the situation (and I've missed some of the messages), the unwanted action is essentially a forced type conversion in combining f1 and Series[f2] . I think even the relatively unsophisticated user may understand that if one multiplies or adds an integer and a real there may be a forced conversion of the product to real, and a small loss of accuracy. But this loss in accuracy is normally both small and also of small import, though it may, of course, be very important in certain rare cases. But a forced (and unnoted) series expansion of f1 is _likely_ to lead to _large_ inaccuracies, since the above input is most likely to be used precisely in situations where Series[f2] is appropriate and acceptable, but Series[f1] is NOT (which is why f1 was left outside in the first place). --AES