Automatic guess of the groups in a GDD

[brettpim]

This branch lets one define a GDD without specifying its groups. The automatic guess is performed in the is_group_divisible_design routine, as it also requires to list the pairs which appear in a block.

Nathann

CC: @nathanncohen @brettpim @sagetrac-danziger @videlec

Component: combinatorial designs

Keywords: group divisible design, block design

Author: Nathann Cohen

Branch/Commit: 7336c5c

Reviewer: Vincent Delecroix

Issue created by migration from https://trac.sagemath.org/ticket/17941

[nathanncohen]

Branch: public/17941

[nathanncohen]

Author: Nathann Cohen

[nathanncohen]

Commit: 6859373

[nathanncohen]

New commits:

6859373

trac #17941: Automatic guess of the groups in a GDD

[videlec]

Reviewer: Vincent Delecroix

[videlec]

comment:2

sage -t --long src/sage/combinat/designs/designs_pyx.pyx
**********************************************************************
File "src/sage/combinat/designs/designs_pyx.pyx", line 207, in sage.combinat.designs.designs_pyx.is_group_divisible_design
Failed example:
    is_group_divisible_design([range(40)],[["e",2]],40,G=[5],lambd=1,verbose=True)
Expected:
    a group has size 40 while G=[5]
    False
Got:
    e does not belong to [0,...,39]
    False
**********************************************************************

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

a328329	trac #17941: Automatic guess of the groups in a GDD
884a359	trac #17941: Broken doctest

[sagetrac-git]

Changed commit from 6859373 to 884a359

[nathanncohen]

comment:4

Sorry ! rebased and fixed.

Nathann

[videlec]

comment:5

Hello,

Shouldn't

if matrix[i*n+j] == 0:
    ...

be

if matrix[i*n+j] < l:
   ...

This suggest to add an example with lambd > 1.

Vincent

[nathanncohen]

comment:6

Hello,

Shouldn't

if matrix[i*n+j] == 0:
    ...

be

if matrix[i*n+j] < l:
   ...

Both work. Consider the whole matrix: for the design to be a GDD, it must have two types of entries only: 0 or l.

1. If the matrix is a real GDD, you can see that the two are equivalent
2. If the matrix is not a GDD, the 'error' message will differ. (wrong size of groups, or a pair not present the right number of times, or two points of a group contained in the same set, ...)

This suggest to add an example with lambd > 1.

Do you have an example in mind?

Nathann

[sagetrac-git]

Changed commit from 884a359 to 7336c5c

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

7336c5c

Trac #17941: one missing sage_free + doc

[videlec]

comment:8

Replying to @nathanncohen:

Hello,

Shouldn't

Both work. Consider the whole matrix: for the design to be a GDD, it must have two types of entries only: 0 or l.

right!

This suggest to add an example with lambd > 1.

Do you have an example in mind?

I added a dummy example.

I am done with the review. Tell me what you think of ​7336c5c.

Vincent

[nathanncohen]

comment:9

Oops. Thanks for the 'free' I forgot!

Nathann

[vbraun]

Changed branch from public/17941 to 7336c5c