Project

General

Profile

Bug #870

GBasis of product of ideals is wrong (Vadim Tropashko) --> I.myReset()

Added by John Abbott about 8 years ago. Updated almost 7 years ago.

Status:
Closed
Priority:
Urgent
Assignee:
Anna Maria Bigatti
Category:
enhancing/improving
Target version:
CoCoA-5.2.0 spring 2017
Start date:
26 Apr 2016
Due date:
% Done:

100%

Estimated time:
2.01 h
Spent time:
1.60 h

Description

Vadim Tropashko reported an incorrect result. Robbiano has simplified it. Here is the simplified version.

use QQ[x,y];
I1 := ideal(x,y);
I2 := ideal(x-1,y-1);
J := I1*I2;
J; -->  ideal(x^2 -x, x*y -x, x*y -y, y^2 -y)  prints out OK
GBasis(J); --> plainly wrong!!
GBasis(ideal(gens(J))); --> correct!

Related issues

Related to CoCoALib - Design #871: Redesign idealsNew2016-04-26

Related to CoCoALib - Design #924: FlagManager for bool/bool3 flagsNew2016-09-19

Related to CoCoALib - Feature #899: IsMaximal, IsPrimary for IDEAL (in cocoalib)Closed2016-06-27

History

#1 Updated by John Abbott about 8 years ago

Since this was reported by a CoCoA-5 user i have put it in the CoCoA-5 category even though I'm almost certain that the problem lies in CoCoALib.

My guess is that the function which creates a product of ideals incorrectly sets the GBasis field (with the wrong value).

I don't have much desire to delve in Max's code; Anna do you?

We should get this sorted out ASAP -- it's an embarrassing bug! Also it should be added to the CoCoA tests (preferably CoCoALib).

#2 Updated by John Abbott about 8 years ago

Try looking around SparsePolyRing.C:2029

#3 Updated by John Abbott about 8 years ago

  • Status changed from New to In Progress
  • % Done changed from 0 to 10

Here is the equivalent CoCoALib code -- it exhibits the problem.

  void program()
  {
    GlobalManager CoCoAFoundations;

    const ring Qxy = NewPolyRing(RingQQ(), symbols("x,y"));
    const RingElem x = indet(Qxy,0);
    const RingElem y = indet(Qxy,1);
    vector<RingElem> v;
    v.push_back(x); v.push_back(y);
    ideal I1 = ideal(v);
    v.clear();
    v.push_back(x-1); v.push_back(y-1);
    ideal I2 = ideal(v);
    I1 = I1*I2;
    cout << "GB: " << GBasis(I1) << endl;
    ideal J = ideal(gens(I1));
    cout << "GB2: " << GBasis(J) << endl;
  }

#4 Updated by John Abbott about 8 years ago

I continue to think that the current design of ideals is completely wrong (but I cannot find the redmine task about this to mark as a related issue). I am convinced that it is poor design to make the function myMul modify one of the args; I expect it would be cleaner if it took two args and produced a new ideal.

The problem is almost certainly around line 2025. The fields IhaveMonomialGens3Value and other similar ones should be reset always (unless both ideals are monomial). For some reason they are reset only if char is non-zero -- why??

I am also mystified as to why IamPrime3Flag is not reset (commented out).

#5 Updated by John Abbott about 8 years ago

#6 Updated by John Abbott almost 8 years ago

I suggest that Anna try to fix the functions already in SparsePolyRing.C.

Even though I'm hoping for a proper redesign of ideals (in the not too distant future?), I think it makes sense to fix the functions we already have, since I'm hoping they can be largely copied as they are to make the new impls... which means that time spent debugging these functions is not (entirely) wasted.

Should there be a procedure to "reset all 3-way flags" for ideals? Requiring each function to mention explicitly every internal flag is not "scalable" (as we have just seen in issue #870). Also a "reset all flags" procedure would mimic well the creation of new ideal for the result (which is the direction I think we should probably move in).

#7 Updated by Anna Maria Bigatti almost 8 years ago

John Abbott wrote:

I suggest that Anna try to fix the functions already in SparsePolyRing.C.

done! (cvs-ing)

Should there be a procedure to "reset all 3-way flags" for ideals? Requiring each function to mention explicitly every internal flag is not "scalable" (as we have just seen in issue #870). Also a "reset all flags" procedure would mimic well the creation of new ideal for the result (which is the direction I think we should probably move in).

Indeed, and even more: I implemented myReset() which resets all member fields but the generators (and removed myClearGBasis).
Any info which can be maintained/deduced has to be done explicitely.

We had the similar (by implementation) bug:

/**/ I := ideal(x+y, x-y);  MinGens(I);
[x, -y]

/**/ MinGens(I+ideal(z));
[x, -y]

#8 Updated by John Abbott almost 8 years ago

  • Assignee set to Anna Maria Bigatti
  • % Done changed from 10 to 50

I have added a new CoCoALib test (test-bug6.C) which tests for this specific problem.

It should later be expanded or replaced by a more thorough test on ideals.

Anna will add a similar test to the CoCoA-5 suite.

#9 Updated by Anna Maria Bigatti almost 8 years ago

  • Subject changed from GBasis of product of ideals is wrong (Vadim Tropashko) to GBasis of product of ideals is wrong (Vadim Tropashko) --> I.myReset()

#10 Updated by Anna Maria Bigatti about 7 years ago

  • Status changed from In Progress to Feedback

#11 Updated by Anna Maria Bigatti about 7 years ago

  • Related to Design #924: FlagManager for bool/bool3 flags added

#12 Updated by Anna Maria Bigatti about 7 years ago

  • Related to Feature #899: IsMaximal, IsPrimary for IDEAL (in cocoalib) added

#13 Updated by Anna Maria Bigatti almost 7 years ago

  • Status changed from Feedback to Closed
  • % Done changed from 50 to 100
  • Estimated time set to 2.01 h

Also available in: Atom PDF