Opened 7 years ago
Last modified 6 years ago
#16347 closed enhancement
Wilson's constructions of OA with 2 truncated groups — at Version 7
Reported by:  ncohen  Owned by:  

Priority:  major  Milestone:  sage6.3 
Component:  combinatorial designs  Keywords:  
Cc:  vdelecroix, knsam, dimpase  Merged in:  
Authors:  Nathann Cohen  Reviewers:  
Report Upstream:  N/A  Work issues:  
Branch:  u/ncohen/16347 (Commits)  Commit:  a4554f5c90ea6a4fee0a170b2fb8a862b21df724 
Dependencies:  #16295, #16391  Stopgaps: 
Description (last modified by )
Heeeeeeeere it is ! The new construction !
As a result, the MOLS table really took a lifetime to fill. So I turned TD and OA into cached functions and they give their answer MUCH faster. The bad side is that it also cached the OA/TD, and this is not necessarily a good idea. The idea would be to cache ONLY the boolean/integer answers, but I do not know how right now.
Anyway caching the OA/TD is not as bad as it seems, for it will only build those who are useful in some construction the users asks. When an OA is not useful, a existence check is made before building the actual design, so this is cool !
I asked on #15657 if the guys there knew a way to tune cached_function to only cache some inputs.
Note that MOLS is not a cached_function, because they output matrix objects .... Does not matter much anyway, the MOLS constructor does not contain much.
Nathann
TODO : Update #16423 when this patch is reviewed
Change History (7)
comment:1 Changed 7 years ago by
 Branch set to u/ncohen/16347
 Status changed from new to needs_review
comment:2 Changed 7 years ago by
 Commit set to 9fefb55ae381228affcfa085d6814fef6e8dd1bf
comment:3 Changed 7 years ago by
 Commit changed from 9fefb55ae381228affcfa085d6814fef6e8dd1bf to 00ee241a234506c6dfade78b97ce445c0d3b87b1
comment:4 Changed 7 years ago by
 Component changed from combinatorics to combinatorial designs
comment:5 Changed 6 years ago by
 Dependencies changed from #16295 to #16295, #16391
With this updated patch, we only have one version of wilson_construction
which can handle 0,1 or 2 truncated columns. It can actually handle much more situations that will appear in new constructions I am writing these days. I think that it is a much cleaner code :)
The two function which look for decomposition of integers are left as they are, they are still useful : they find sets of parameters with which the new wilson_construction
function is to be fed.
Nathann
comment:6 Changed 6 years ago by
 Commit changed from 00ee241a234506c6dfade78b97ce445c0d3b87b1 to a4554f5c90ea6a4fee0a170b2fb8a862b21df724
Branch pushed to git repo; I updated commit sha1. New commits:
90a72bd  trac #16370: OA(k,n) strongly regular graphs

395458f  trac #16391: Helper functions for OA constructions

0ed318e  trac #16391: Merged with 6.3.beta2

f86d187  trac #16391: small improvement

d954b93  trac #16391: Independent Set problem with a LP

c1bb903  trac #16347: Merged with #16391

a4554f5  trac #16347: Genelarized Wilson construction

comment:7 Changed 6 years ago by
 Description modified (diff)
Branch pushed to git repo; I updated commit sha1. Last 10 new commits:
trac #16241: useless if condition in MOLS constructor
trac #16241: Merged with updated #16235
trac #16236: Five mutually orthogonal latin squares of order 12
trac #16236: Merged with updated #16241
trac #16236: Broken doctests
trac #16236: Merged with updated #16241
trac #16236: Merged with #16272
trac #16295 : Faster is_orthogonal_array
trac #16295: Doctests and review
trac #16347: Wilson's constructions of OA with 2 truncated groups