support group extensions of matrix groups #4013
Conversation
- admit `Oscar.GAPGroupElem` not only `Oscar.BasicGAPGroupElem`
in evaluating 2-cochains (we need in particular `Oscar.MatrixGroupElem`)
- add a `isomorphism(::Type{FPGroup}, A::AbstractAlgebra.Generic.FreeModule)`
method, which is needed by `extension` if we want to support G-modules
created from matrix groups acting on vector spaces
- add tests
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #4013 +/- ##
==========================================
+ Coverage 84.59% 84.68% +0.09%
==========================================
Files 596 600 +4
Lines 81971 82640 +669
==========================================
+ Hits 69341 69987 +646
- Misses 12630 12653 +23
|
|
Mostly happy, but in Cohomology.m we already have I think we need to consolidate this (and make a consistent interface?) At least as a promise, the "old" function is more general as in accepts not only prime fields, and not only free_modules. |
|
What about also adding (maybe later)
|
|
Concerning a consistent interface, the code anyhow needs Concerning |
|
The existing functions will have to be renamed into |
|
O.k., I can adjust the pull request according to the above discussion. |
- rename `natural_gmodule` to `regular_gmodule`, extend its documentation, export it - introduce `natural_module`, export it - add tests for `natural_gmodule` and `regular_gmodule`
|
Concerning the idea
There is already a matching |
and support `is_elementary_abelian` for `FinGenAbGroup`
- generalize `isomorphism(FPGroup, M::...; on_gens=true)` and
`isomorphism(PcGroup, M::...: on_gens=true)`
to `M::AbstractAlgebra.Generic.FreeModule{FqFieldElem}`
as in the functions `fp_group_with_isomorphism`,
`pc_group_with_isomorphism` in `experimental/GModule`,
delegate from the latter to the former
- remove the variant
`fp_group_with_isomorphism(g::Vector{<:Oscar.GAPGroupElem})`,
since it did in fact not depend on `g` but on `parent(g[1])`.
- move the general `relators(G::GAPGroup)` methods to `src/Groups`,
changed `relations` to delegate to `relators`
- change `free_group(G::FPGroup)` to return `G` if `G` is free
- move `underlying_word(g::FPGroupElem)` to `src/Groups`,
changed it to return `g` if its parent is free
- add `isomorphism(T, G::T)` methods that return an identity map
- improve the `isomorphism(::Type{FPGroup}, A::FinGenAbGroup)`
in the sense that only commutator relators for one half of the matrix
are needed
- add some tests
- removed some tests:
Due to the fact that `relators` is now defined for more types of groups,
one may get fewer errors or different errors than before (and it may take
longer until one gets these errors)
|
on first glance: looks good! |
|
I'll discuss this next week with @fieker and @ThomasBreuer to see if we can merge (and I'll try to get in a review I before that, too) |
* support group extensions of matrix groups
- admit `Oscar.GAPGroupElem` not only `Oscar.BasicGAPGroupElem`
in evaluating 2-cochains (we need in particular `Oscar.MatrixGroupElem`)
- add a `isomorphism(::Type{FPGroup}, A::AbstractAlgebra.Generic.FreeModule)`
method, which is needed by `extension` if we want to support G-modules
created from matrix groups acting on vector spaces
- add tests
* add a test in order to increase coverage
* `natural_gmodule` and `regular_gmodule`
- rename `natural_gmodule` to `regular_gmodule`,
extend its documentation, export it
- introduce `natural_module`, export it
- add tests for `natural_gmodule` and `regular_gmodule`
* add `elementary_abelian_group`
and support `is_elementary_abelian` for `FinGenAbGroup`
* next iteration, motivated by the comments
- generalize `isomorphism(FPGroup, M::...; on_gens=true)` and
`isomorphism(PcGroup, M::...: on_gens=true)`
to `M::AbstractAlgebra.Generic.FreeModule{FqFieldElem}`
as in the functions `fp_group_with_isomorphism`,
`pc_group_with_isomorphism` in `experimental/GModule`,
delegate from the latter to the former
- remove the variant
`fp_group_with_isomorphism(g::Vector{<:Oscar.GAPGroupElem})`,
since it did in fact not depend on `g` but on `parent(g[1])`.
- move the general `relators(G::GAPGroup)` methods to `src/Groups`,
changed `relations` to delegate to `relators`
- change `free_group(G::FPGroup)` to return `G` if `G` is free
- move `underlying_word(g::FPGroupElem)` to `src/Groups`,
changed it to return `g` if its parent is free
- add `isomorphism(T, G::T)` methods that return an identity map
- improve the `isomorphism(::Type{FPGroup}, A::FinGenAbGroup)`
in the sense that only commutator relators for one half of the matrix
are needed
- add some tests
- removed some tests:
Due to the fact that `relators` is now defined for more types of groups,
one may get fewer errors or different errors than before (and it may take
longer until one gets these errors)
* fix an `elementary_abelian_group` method
admit
Oscar.GAPGroupElemnot onlyOscar.BasicGAPGroupElemin evaluating 2-cochains (we need in particularOscar.MatrixGroupElem)add a
isomorphism(::Type{FPGroup}, A::AbstractAlgebra.Generic.FreeModule)method, which is needed byextensionif we want to support G-modules created from matrix groups acting on vector spacesadd tests