Bug in IncidenceStructure.dual_design

[Nathann Cohen]

As the title says.

With some other improvements.

The original code seems to have been written by a mathematician.

Nathann

[git]

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

​fed439e

trac #16032: Bug in IncidenceStructure.dual_design

[Dima Pasechnik]

What is the bug in question?

[Dima Pasechnik]

there is a bug in the optional test: the line 422 in incidence_structures.py should be

sage: PP = designs.ProjectiveGeometryDesign(2,1,GF(4,'conway'), algorithm="gap") # optional - gap_design

actually, it's weird that ProjectivePlaneDesign does not known about algorithm= at all. This is a bug to be fixed, IMHO, perhaps on this ticket...

---

the patch is:

--- a/src/sage/combinat/designs/incidence_structures.py
+++ b/src/sage/combinat/designs/incidence_structures.py
@@ -419,7 +419,7 @@ class IncidenceStructure(object):
             sage: PP = designs.ProjectivePlaneDesign(4)
             sage: PP.dual_design().is_block_design()
             (True, [2, 21, 5, 1])
-            sage: PP = designs.ProjectivePlaneDesign(4, algorithm="gap") # optional - gap_design
+            sage: PP = designs.ProjectiveGeometryDesign(2,1,GF(4,'conway'), algorithm="gap") # optional - gap_design
             sage: PP.dual_design().is_block_design()                     # optional - gap_design
             (True, [2, 21, 5, 1])

without it, the test

sage -t --long --optional=sage,gap_design src/sage/combinat/designs/incidence_structures.py 

fails.

[Nathann Cohen]

Yoooooooo !

there is a bug in the optional test: the line 422 in incidence_structures.py should be

sage: PP = designs.ProjectiveGeometryDesign(2,1,GF(4,'conway'), algorithm="gap") # optional - gap_design

Nononono, I meant this "algorithm=gap" to be an argument of the function I tests, i.e. dual_design.

And I rename this gap_design (which is not even the name of a spkg) to gap_packages.

actually, it's weird that ProjectivePlaneDesign does not known about algorithm= at all. This is a bug to be fixed, IMHO, perhaps on this ticket...

The standard of the bugs I fix is MUCH higher than the bugs that are not fixed in combinat, really. If THIS is a bug their code should not even be allowed to run :-P

without it, the test

sage -t --long --optional=sage,gap_design src/sage/combinat/designs/incidence_structures.py 

fails.

Well, as I said above the "algorithm=gap" was intended for the dual_block method. I don't see why we would call gap to compute a dual, but anyway :-P

Aaaaaaaaaaaaaand I added this algorithm argument in ProjectivePlane too.

Nathann

[git]

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

​f82b15d

trac #16032: reviewer's remarks

[Dima Pasechnik]

OK!

[Nathann Cohen]

You don't mind my removing this "if", do you ? :-PPPPP

Nathann

[git]

Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:

​7dcff95

trac 16032: This can be simplified... :-P

[Dima Pasechnik]

shouldn't these matrices be created sparse?

[Nathann Cohen]

Ahahaha. Yes of course. Do I change that ?

This code is awful, I am already happy to see those two "for" reverted in order to avoid a useless test. Mathematicians should not write code, all they care about is whether the code can be read like a mathematical proof, and efficiency can go to hell.

Nathann

[Dima Pasechnik]

Replying to ncohen:

Ahahaha. Yes of course. Do I change that ?

yes please...

This code is awful, I am already happy to see those two "for" reverted in order to avoid a useless test. Mathematicians should not write code, all they care about is whether the code can be read like a mathematical proof, and efficiency can go to hell.

they can be taught to write good code, IMHO :)

[Nathann Cohen]

Yooooooo !

yes please...

Done !

they can be taught to write good code, IMHO :)

Probably. But most of your efforts will be wasted, and in the meantime you have to re-do the work when they turn their back.

Nathann

[git]

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

​f6da4b7

trac #16032: These matrices should be sparse