I think that throwing out ratpoi entirely would be a mistake.
Appropriate documentation could explain its purpose and allow interested
persons to construct the rational (CRE) coefficient version.
Perhaps conditionalizing the code somehow like #+RATPOIS (......)
The point is that under some circumstances the RATPOIS setting would run
important problems 1000X faster...
This is not something that the test suite tests for.
RJF
Andreas Eder wrote:
>Richard Fateman wrote:
>
>
>
>>Yes, I wrote the poisson stuff.
>>
>>As I recall, there are two different encoding schemes used for the
>>coefficients, one more efficient (but restricted to CRE) than the other,
>>which uses the general simplify "SIMPLIFYA" program.
>>
>> It was intended that only ONE of the two poisson series packages be loaded.
>>
>>Someone (Perhaps Bill Schelter?) seemed to set up the system so that
>>they both were loaded.
>>
>>But then someone (else?) decided to comment out all the code in one of
>>them, pois2.
>>The version that is left, pois3, does not use CRE..
>>
>>Consequently, the code in ratpoi, which uses CRE, is overwritten and is
>>never used.
>>
>>
>
>Yes and apparently it runs the few tests that are in the testsuite
>ok.
>
>So we could eliminate ratpoi.lisp? And remove quite a few compiler
>warnings that way (I'm donw to 27 in clisp now, after the last
>changes in the fortran translated code).
>
>Robert would you remove ratpoi.lisp from cvs?
>
>Andreas
>
>