On 2026-01-15 5:42 p.m., Dima Pasechnik wrote:
What does it mean, by the way: "some calculations giving different
results". Results are either right or wrong.
If a right result is differently packaged, so it be. After all, it's
unusual for different versions of
Sage code to give results which are differently
formatted, or have another representation, another set (but a correct
set) of generators for something, etc.
The new code works over weighted projective space instead of projective
space. So some calculations that were not broken may give different
results in the sense that the objects have a different representation.
Of course for the calculations that were broken (Jacobian arithmetic for
even degree hyperelliptic curves), I don't care about "breaking" them
because it's already broken.
On 2026-01-16 9:42 a.m., Giacomo Pope wrote:
The only addition I want to make to this thread is that arithmetic on
Jac(H) truly is "broken" when H has an even degree model (zero or two
points at infinity) in the sense that the group operation is just
incorrect. There are many issues reporting this I can try and find
some links...
https://github.com/sagemath/sage/issues/39956
https://github.com/sagemath/sage/issues/32024
https://github.com/sagemath/sage/issues/23154
https://github.com/sagemath/sage/issues/37109
However, because we go from one projective space to the new weighted
projective space then things for the odd degree model for the
HyperellipticCurve things which weren't broken may not behave slightly
differently.
I am personally in favour of directly replacing it but understand that
it's not my decision.
I am also in favour of directly replacing it. For the odd degree model,
is the only way this change could "break" existing code be if a user was
comparing against a specific value? Or in other words, if a user has
some Sage code that has no hardcoded values (so no assuming that a
specific calculation returns a specific point (X : Y : Z)), would the
same code /usually/ work with the new smooth model? So most code will
not break, and any code that does break is probably easy to fix?
On 2026-01-15 6:13 p.m., Nils Bruin wrote:
I think part of the problem you could run into is unanimity on whether
the old code is broken. I think, the code in its present form is
basically correct for the singular curve. The new code would have
different semantics by working with a nonsingular curve in an
appropriately weighted projective space. I'm pretty sure I would need
the new behaviour whenever I'd ask a computer algebra system, so I
wouldn't complain. However, if there is someone out there who actually
does want the sematics of the current model, they'd probably point at
the documentation and claim it's not "broken" or erroneous (I haven't
checked, but I assume the old code is properly documented for what it
does).
If you can get (very?) close to unanimity then a short-cut may be
feasible. I don't even know unanimity in which group you should be
striving for, though.
Giacomo explained how the old code is broken for even degree models. As
for unanimity in some group to bypass the normal deprecation process, I
wasn't sure which group either. That's why I asked here.
I'll tag a few people on the GitHub PR who I know have knowledge about
curve arithmetic. If we hear no objections from them or on this mailing
list in the next week or so, do people thing it's reasonable to just
replace the old hyperelliptic code with the new smooth model
implementation without a deprecation period?
As a side note, this is not the only part of Sage that could benefit
from a major rewrite and where following deprecation policies to the
letter might result in a more difficult transition than just rewriting
the whole thing. The backwards-incompatible changes here are fairly
minor, but it may be worth thinking about some kind of policy to allow
major backwards-incompatible rewrites in the future.
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion visit
https://groups.google.com/d/msgid/sage-devel/5ecef853-2d99-4ebc-ac3c-53d3cf121a44%40ucalgary.ca.
