Ticket #4593 (new defect)

Opened 2 months ago

Last modified 2 weeks ago

do not unconditionally use M2 for Gbasis computations over ZZ if it is installed

Reported by: mabshoff Assigned to: malb
Priority: blocker Milestone: sage-3.4
Component: commutative algebra Keywords:
Cc:

Description

This is a left over from #4589: The doctest below from sage/rings/polynomial/multi_polynomial_ideal.py changes depending on whether M2 is installed or not since the GBasis computation uses the optional M2 if it is installed. But the interface should offer an option what code is used since results should not vary depending on optional spkg 

@@ -164,7 +166,7 @@

         sage: I.change_ring(P.change_ring( IntegerModRing(2*7) )).groebner_basis()
         verbose 0 (...: multi_polynomial_ideal.py, groebner_basis) Warning: falling back to very slow toy implementation.
-        [x + y + z, y^2 + y + 8, y*z + y + 2, 2*y + 6, z^2 + 3, 2*z + 10]
+        [x + y + z^3 + z^2 + 11, y^2 + y + 5*z^3 + 2*z^2 + 3*z + 10, y*z + y + 9*z^3 + 5*z^2 + 9*z + 11, 2*y + 2*z^3 + 4*z^2 + 4*z + 8, z^2 + 3, 2*z + 10]

     Modulo any other prime the Groebner basis is trivial so there are
     no other solutions. For example:

Cheers,

Michael

Change History

11/30/2008 10:25:29 PM changed by mabshoff

I have also made the other doctest "optional - m2" since I have been seeing the following failure: 

File "/Users/mabshoff/sage-3.2.1.alpha2/devel/sage/sage/rings/polynomial/multi_polynomial_ideal.py", line 145:
    sage: I.groebner_basis()
Expected:
    verbose 0 ... Warning: falling back to very slow toy implementation.
    [x + y + z, y^2 + y + 23234, y*z + y + 26532, 2*y + 158864, z^2 + 17223, 2*z + 41856, 164878]
Got:
    verbose 0 (1792: multi_polynomial_ideal.py, groebner_basis) Warning: falling back to very slow toy implementation.
    [x + y + 3*z + 206734, y^2 + y + 2*z + 65090, y*z + y + 26532, 2*y + 158864, z^2 + 2*z + 223957, 2*z + 206734, 164878]

This is on varro, i.e. the MacOSX 10.4 box on SkyNet?.

Cheers,

Michael

12/23/2008 12:05:11 PM changed by was

  • priority changed from critical to blocker.

12/23/2008 12:34:33 PM changed by was

For the record, thinking about this, it does seem stupid to not just use M2 when it is available, just because we can't proper doctest this. A natural way to fix this would be to make the output random, get rid of the verbose warning altogether, and include an extra example that is marked 

   # optional -- macaulay2

I think *all* warnings about "using slow toy implementations" for any algorithm in sage should use some sort of unified framework, which is different (but related to) the verbose flag. E.g., each slow function would at the start call a function defined in misc: 

   slow("Groebner over ZZ -- install M2 to speed this up")

This would print a warning iff a certain flag is set, which would be off by default.

It would be good to add calls like that everywhere e.g., in the base classes for matrices and other places where we have "slow generic code". Then when people run code, they can set the flag and see a list of messages indicating exactly what they need to do to eliminate generic code from that a given code path, which would tip them off about how to speed up their code.

Obviously, this has to be done carefully. E.g., it would be bad to put slow(...) in the matrix getitem method, since that would slow the getitem method itself down.

-- William

12/23/2008 12:42:55 PM changed by mabshoff

  • milestone changed from sage-3.4 to sage-3.2.3.

I am pretty sure that the generic framework for slow generic code should be its own ticket, but I really do like the idea.

Moving to 3.2.3, too :)

Cheers,

Michael

12/23/2008 05:58:35 PM changed by was

I am pretty sure that the generic framework for slow generic code should be its own ticket, but I really do like the idea.

I strongly agree that it should be another ticket.

12/24/2008 06:35:05 AM changed by mabshoff

  • summary changed from do not uncinditionally use M2 for Gbasis computations over ZZ if it is installed to do not unconditionally use M2 for Gbasis computations over ZZ if it is installed.

12/26/2008 09:31:54 AM changed by mabshoff

  • milestone changed from sage-3.2.3 to sage-3.4.

For the record: The two failures listed above already are optional, so I asked Jaap who has been seeing those failure to open another ticket in case they are still around.

Since we will not be fixing the fundamental problem here in 3.2.2 I am reassigning this to 3.4.

Cheers,

Michael

12/26/2008 09:44:44 AM changed by mabshoff

The ticket to make another test optional for now until this is resolved is #4882

Cheers,

Michael