Add some basic methods for MinimalSemigroupGeneratingSet #407
Conversation
There was a problem hiding this comment.
This looks good. I agree that it is difficult to implement this in general, and that limiting what we try is a good idea.
Would it be much work to add methods for MinimalMonoidGeneratingSet and the ones for inverse semigroups, along the lines of what you've done already? It seems like it wouldn't be. If you'd rather not do it, then please add an issue stating that it could be done in a similar way to that which you've done here.
I wonder too if it might not be worthwhile having a fallback brute force method for semigroups that are not too big?
gap/attributes/attr.gi
Outdated
| return gens; | ||
| fi; | ||
|
|
||
| ErrorNoReturn("not yet implemented,"); |
There was a problem hiding this comment.
This error message does not have the correct format.
doc/attr.xml
Outdated
| @@ -337,8 +337,9 @@ gap> Length(SmallGeneratingSet(S)); | |||
| <Returns>A minimal generating set for a semigroup.</Returns> | |||
| <Description> | |||
|
|
|||
| <B>Warning:</B> currently, no methods are installed to compute these | |||
| attributes.<P/> | |||
There was a problem hiding this comment.
I'm confused, this documentation existed before, but there are no methods installed? Would it not make more sense to not document (or indeed declare) attributes where there are no methods installed? Maybe I misunderstood something here, but this seems like an oddity to me.
There was a problem hiding this comment.
And the example below seems to contradict the assertion that there are no methods installed for MinimalInverseMonoidGeneratingSet. The distinction between there being a method that "computes" a minimal generating set, and one that simply knows (like a monogenic inverse semigroup) a minimal generating set is probably too subtle. I suggest improving this warning to make it clearer what is going on.
wilfwilson
force-pushed
the
minimal-genset
branch
from
1ef10f7
to
1b8451e
Compare
|
@wilfwilson any comments on my comments? |
|
Yes, I'll address it and hopefully make the changes at the next Wednesday meeting. A minimal monoid generating set is simply a minimal semigroup generating set, except you throw away the identity if present (except possibly for partial perms...). I'll have to think and talk to you about minimal inverse semigroup generating sets. I'm happy to not document the attributes that genuinely have no methods installed. But it is useful to declare them so that we can set them when we create things that we know have minimal (whatever) generating sets: eg if we create a monogenic inverse monoid, for instance. |
wilfwilson
force-pushed
the
minimal-genset
branch
from
1b8451e
to
afcba03
Compare
|
I think I've addressed your changes. I've added methods for |
wilfwilson
force-pushed
the
minimal-genset
branch
3 times, most recently
from
a4dd55f
to
8832ae5
Compare
|
The Travis tests are failing because of the coverage job, even though it shouldn't be. I think it's a problem with the profiling package. I'll investigate. |
wilfwilson
force-pushed
the
minimal-genset
branch
from
b939518
to
269f00f
Compare
|
I've changed Travis to use the master version of |
|
@wilfwilson happy to merge this if you rebase |
wilfwilson
force-pushed
the
minimal-genset
branch
from
269f00f
to
44f36c1
Compare
|
Done, thanks @james-d-mitchell. |
It is going to be very difficult to implement completely general methods for
MinimalSemigroupGeneratingSetthat have reasonable performance. However, I think we should still write methods that apply in special cases.Here I've added several of those special cases. The
IndecomposableElementsattribute added in PR #347 was useful for this.The biggest class of semigroups to which the new method applies is probably finite J-trivial semigroups.