Ticket #3793 (closed defect: fixed)

Opened 4 months ago

Last modified 4 months ago

[with patch, with positive review] Some elliptic curve doctests fail when the optional database is installed

Reported by: cremona Assigned to: was
Priority: minor Milestone: sage-3.1
Component: algebraic geometry Keywords:
Cc:

Description

A few of the doctests in ell_rational_field.py fail when the optional package database_cremona_ellcurve-20071019 is installed, mainly because for curves in the database the gens() as supplied by the database may differ from those computed on the fly. (In almost all cases the generators are not uniquely determined, being the generators of a finitely-generated abelian group. We have put some thought into how to make the generators canonical but have not yet succeeded.)

Attachments

10109.patch (3.0 kB) - added by cremona on 08/09/2008 05:53:12 AM.

Change History

08/09/2008 05:53:12 AM changed by cremona

  • attachment 10109.patch added.

08/09/2008 05:54:54 AM changed by cremona

  • summary changed from Some elliptic curve doctests fail when the optional database is installed to [with patch, needs review] Some elliptic curve doctests fail when the optional database is installed.

To test this you should really test the doctests in ell_rational_field.py both before and after installing the database.

08/09/2008 05:31:29 PM changed by mabshoff

  • milestone set to sage-3.1.

08/10/2008 01:57:03 PM changed by ncalexan

  • summary changed from [with patch, needs review] Some elliptic curve doctests fail when the optional database is installed to [with patch, with positive review] Some elliptic curve doctests fail when the optional database is installed.

I ran the tests on 3.0.6 before and after installing the database, *without* applying the patch, and both tests passed everything. So... is this really necessary?

But I still think this looks good and should be applied, since it addresses some ambiguity that could be annoying.

08/10/2008 02:16:00 PM changed by cremona

The point is that there was randomness in the old doctests: whenever they use E.gens() where E is an elliptic curve we cannot guarantee that the same gens are computed (on different systems, etc). As a special case of this ambiguity, the gens obtained from the database (which don't change! -- or at least ont very rarely, e.g. if they are found to be wrong) may not agree with computed gens.

I dealt with this by either inserting "# random", or by using explicit points instead of gens().

I hope that with this explanation you can give this (admittedly rather trivial) patch a positive review.

08/10/2008 02:20:32 PM changed by cremona

ok, so it already has a positive review! Thanks!

08/10/2008 09:59:07 PM changed by mabshoff

  • status changed from new to closed.
  • resolution set to fixed.

Merged in Sage 3.1.alpha1