Some serialization of schemes #4110
Conversation
HechtiDerLachs
force-pushed
the
localization_serialization
branch
from
338cfa7
to
7ac8e7d
Compare
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #4110 +/- ##
==========================================
+ Coverage 84.67% 84.71% +0.03%
==========================================
Files 626 627 +1
Lines 83980 84176 +196
==========================================
+ Hits 71108 71306 +198
+ Misses 12872 12870 -2
|
| @check begin | ||
| d = prod(length(f) > 0 ? f : one(OO(X))) | ||
| U = spec(R) | ||
| U == hypersurface_complement(X, d) | ||
| end | ||
| return new{base_ring_type(X), ring_type(U), typeof(X)}(X, U, lifted_numerator.(f)) |
There was a problem hiding this comment.
this looks fishy to me. U is not defined if check=false
There was a problem hiding this comment.
You're right. I'll look into this.
HechtiDerLachs
force-pushed
the
localization_serialization
branch
from
1a9a316
to
5e4398d
Compare
test/Serialization/IPC.jl
Outdated
| @testset "parallel smoothness test for schemes" begin | ||
| channels = Oscar.params_channels(Union{AffineScheme, Ring}) | ||
|
|
||
| IP2 = projective_space(QQ, [:x, :y, :z]) | ||
| S = homogeneous_coordinate_ring(IP2) | ||
| (x, y, z) = gens(S) | ||
|
|
||
| I = ideal(S, y^2*z + x^3 + x^2*z) | ||
|
|
||
| X, _ = sub(IP2, I) | ||
| X_cov = covered_scheme(X) | ||
| U = affine_charts(X_cov) | ||
| for a in U | ||
| Oscar.put_params(channels, ambient_coordinate_ring(a)) | ||
| Oscar.put_params(channels, OO(a)) | ||
| end | ||
| results = pmap(is_smooth, U) | ||
| @test results == [1, 1, 0] | ||
| end |
There was a problem hiding this comment.
ping @simonbrandhorst , @afkafkafk13 .
|
Can we merge #4121 first and then get a better view on what's going on here? It seems that this is built on top of the other. |
HechtiDerLachs
force-pushed
the
localization_serialization
branch
from
bc12d64
to
6ccadd7
Compare
|
@antonydellavecchia : I did another implementation which broadcasts all rings to all workers using |
| @testset "_is_smooth" begin | ||
| R, (x, y, z) = QQ[:x, :y, :z] | ||
|
|
||
| I1 = ideal(R, x*y) | ||
| IA3 = spec(R) | ||
| X1, _ = sub(IA3, I1) | ||
| @test !Oscar._is_smooth(X1) | ||
| @test !Oscar._is_smooth(X1; jacobian_cut_off=0) |
There was a problem hiding this comment.
@afkafkafk13 : I think I have now implemented a truly generic smoothness test here which can also handle non-reduced, non-equidimensional, and non-connected cases. I'm still a bit skeptic, since all the classical implementations seem to solve this via some primary/equidimensional decomposition which I seem to be able to fully avoid. If you have the time: Could you have a look and tell me what you think? Maybe you have some further exotic example in mind for which my implementation fails. If so, it would be good to know about it.
|
ping @danteluber : Check out this PR and try the following. using Distributed
process_ids = addprocs(4)
@everywhere using OscarThat starts the workers. Then you can do Replace |
Intermediate state of the art of our joint effort with @antonydellavecchia this week.