Re: SF.net SVN: fricas:[600] branches/new-lambda

[Ralf Hemmecke]

Waldek,

can you say anything about the kind of regression?
http://groups.google.com/group/fricas-commit/browse_thread/thread/4e9b666d2324c97e

Below you find a diff to -r601 for which I have have seen no essential
differences in the .lsp files.

Should I commit?

Ralf

>svk diff
=== src/algebra/fparfrac.spad.pamphlet
==================================================================
--- src/algebra/fparfrac.spad.pamphlet (revision 3124)
+++ src/algebra/fparfrac.spad.pamphlet (local)
@@ -102,7 +102,7 @@
ans

UP2SUP p ==
- map(#1::UP::RF, p)$UnivariatePolynomialCategoryFunctions2(F, UP,
RF, SUP)
+ map((z1: F): RF +-> z1::UP::RF,
p)$UnivariatePolynomialCategoryFunctions2(F, UP, RF, SUP)

-- returns Trace_k^k(a) (h(a) / (x - a)^n) where d(a) = 0
REC2RF(d, h, n) ==
=== src/algebra/fs2expxp.spad.pamphlet
==================================================================
--- src/algebra/fs2expxp.spad.pamphlet (revision 3124)
+++ src/algebra/fs2expxp.spad.pamphlet (local)
@@ -139,9 +139,7 @@
-- and log
smpElem(numer f) / smpElem(denom f)

- -- smpElem p == map(k2Elem, (x : R) : FE +-> x::FE, p)$PCL
- -- PCL ==> PolynomialCategoryLifting(IndexedExponents
K,K,R,SMP,FE)
- smpElem p == map(k2Elem,#1::FE,p)$PCL
+ smpElem p == map(k2Elem, (x1: R) : FE +-> x1::FE, p)$PCL

k2Elem k ==
-- rewrites a kernel; all trig functions are
=== src/algebra/fs2ups.spad.pamphlet
==================================================================
--- src/algebra/fs2ups.spad.pamphlet (revision 3124)
+++ src/algebra/fs2ups.spad.pamphlet (local)
@@ -459,8 +459,7 @@
-- functions are expressed in terms of exp
smpElem(numer f) / smpElem(denom f)

- smpElem p == map(k2Elem,#1::FE,p)$PCL
- -- smpElem p == map(k2Elem, x +-> x::FE, p)$PCL
+ smpElem p == map(k2Elem, (x1: R): FE +-> x1::FE, p)$PCL

k2Elem k ==
-- rewrites a kernel; all trig functions are

[Waldek Hebisch]

>
> Waldek,
>
> can you say anything about the kind of regression?
> http://groups.google.com/group/fricas-commit/browse_thread/thread/4e9b666d2324c97e
>

Testsuite erorrs like:

testEquals("integrate(1/(1+x^4),x = %minusInfinity..%plusInfinity)", _
"sqrt(2)/2*%pi")

Daly Bug
>> Error detected within library code:
cannot retract nonconstant polynomial

Continuing to read the file...

or:

limit(x/exp(x),x = %minusInfinity) -- %minusInfinity


(5) "failed"
Type: Union("failed",...)

> Below you find a diff to -r601 for which I have have seen no essential
> differences in the .lsp files.
>
> Should I commit?
>

I think it deserves a bit more testing. If it works it would be
good to understand why one version works and the other does not.
Or at least know what triggers problems so we can avoid them.

--
Waldek Hebisch
heb...@math.uni.wroc.pl

[Ralf Hemmecke]

> I think it deserves a bit more testing. If it works it would be
> good to understand why one version works and the other does not.
> Or at least know what triggers problems so we can avoid them.

I don't know exactly from which of the tree problematic cases this comes
from but I realized that you had a replacement in fparfrac.spad which was

UP2SUP p ==
map(x +-> x::UP::RF,
p)$UnivariatePolynomialCategoryFunctions2(F, UP, RF, SUP)

before you reverted and now is

UP2SUP p ==

map((z1: F): RF +-> z1::UP::RF,
p)$UnivariatePolynomialCategoryFunctions2(F, UP, RF, SUP)

The big difference is that I use z1 and you used x. Look a bit more back
to find

x := monomial(1, 1)$UP

I am not sure, whether this caused the regression. It shouldn't, since
the x in x +-> x::UP::RF should hide the outer meaning of x. But you are
probably better in finding out what happened there.

I did not look into the other two, but just added the types. But maybe
it is a similar naming conflict.

Ralf

[Ralf Hemmecke]

I've just built new-lambda (including my patch
http://groups.google.com/group/fricas-devel/browse_thread/thread/1c5900d519b8c192)
without seeing any regression.

So, I'll commit soon...

Ralf