From: Bruce S. <Bru...@nc...> - 2009-06-29 06:25:28
|
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> <meta content="text/html;charset=UTF-8" http-equiv="Content-Type"> <title></title> </head> <body bgcolor="#ffffff" text="#000000"> It's good that you provide two versions, but there's an informational problem for someone trying to figure out the ramifications of which to use.<br> <br> There aren't just two sides to this; there are more like four sides. Sides 3 and 4 are these:<br> <br> 3) There exists a large body of existing VPython programs developed over the last 9 years that various constituencies depend on working, and removing the numpy and math imports will break many (most?) of these. Many of these users are students and faculty who cannot be easily reached to explain a major change; they don't read computer lists.<br> <br> 4) There is a tricky point concerning mathematical functions that appear in both math and numpy. When Arthur Siegel did the work to convert Visual from dependence on Numeric to dependence on numpy, performance on many existing programs was horrible, because numpy scalar floats, unlike Numeric scalar floats, were no longer captured by the fast math functions but instead were processed by the array-handling numpy functions. There is arcane code in Visual to get around this problem. Few users would be able to figure this out for themselves.<br> <br> Bruce Sherwood<br> <br> Guy K. Kloss wrote: <blockquote cite="mid:200...@ma..." type="cite"> <pre wrap="">Hold your horses there ... On Mon, 29 Jun 2009 17:52:11 Bruce Sherwood wrote: </pre> <blockquote type="cite"> <pre wrap="">Warning: I do not accept the notion of removing the full importation of numpy and math without vastly more discussion. </pre> </blockquote> <pre wrap=""><!----> That's why that package is provided *additionally*, but the basic one is given first on that page. </pre> <blockquote type="cite"> <pre wrap="">There were and are good reasons for the existing scheme, and changing this will break existing programs right and left, programs that large numbers of people depend on. Most of these people are students and instructors who aren't working on Linux, so there's little immediate danger. And I accept that for some advanced users cleaning the name space can be useful. But it should be understood that this is for "consenting adults." My concern is that it could be very difficult for someone who is not expert to be able to figure out what has gone wrong. </pre> </blockquote> <pre wrap=""><!----> Yes, so can be the difficulty trying to figure out why the code is all of a sudden behaving erratically, although you did everything the way you usually do in Python. There are strong arguments for both sides. The one to make it easier, and the one to keep it in a Pythonic way. And after all, the ones using it are using Python, so that's the common denominator one should adhere to. And telling them to do a wild card import should be easy as anyway, if they want it to be simple but messy. Guy </pre> </blockquote> </body> </html> |