Opened 12 years ago
Closed 5 years ago
#6101 closed enhancement (fixed)
computation of induced morphism on homology and cohomology of simplicial complex morphisms
Reported by:  bantieau  Owned by:  bantieau 

Priority:  major  Milestone:  sage6.10 
Component:  algebraic topology  Keywords:  
Cc:  jhpalmieri, nthiery, SimonKing  Merged in:  
Authors:  John Palmieri  Reviewers:  Travis Scrimshaw 
Report Upstream:  N/A  Work issues:  
Branch:  e9150ca (Commits)  Commit:  e9150caff3d157cadd739042461d972c9d5952f1 
Dependencies:  #19179, #6102, #18175  Stopgaps: 
Description (last modified by )
Change History (44)
comment:1 Changed 8 years ago by
 Milestone changed from sage5.11 to sage5.12
comment:2 Changed 7 years ago by
 Milestone changed from sage6.1 to sage6.2
comment:3 Changed 7 years ago by
 Milestone changed from sage6.2 to sage6.3
comment:4 Changed 7 years ago by
 Milestone changed from sage6.3 to sage6.4
comment:5 Changed 5 years ago by
 Dependencies set to #19179
 Report Upstream set to N/A
comment:6 Changed 5 years ago by
 Dependencies changed from #19179 to #19179, #6102
comment:7 Changed 5 years ago by
 Branch set to u/jhpalmieri/inducedmaps
comment:8 Changed 5 years ago by
 Commit set to 698ba79f9837c3c2e83bd358441b4cd4a4073392
 Status changed from new to needs_review
comment:9 Changed 5 years ago by
 Priority changed from minor to major
comment:10 Changed 5 years ago by
Hi John,
This seems pretty interesting. I just started to play around with it and was struck by a couple of minor usability issues:
sage: X=simplicial_complexes.BrucknerGrunbaumSphere() sage: phi, M = algebraic_topological_model(X, GF(2)) sage: phi Chain homotopy between Chain complex morphism from Chain complex with at most 4 nonzero terms over Finite Field of size 2 to Chain complex with at most 4 nonzero terms over Finite Field of size 2 and Chain complex morphism from Chain complex with at most 4 nonzero terms over Finite Field of size 2 to Chain complex with at most 4 nonzero terms over Finite Field of size 2 sage: phi._codomain() Trivial chain sage: phi._domain() Trivial chain sage: phi._codomain() is phi._domain() False
As you can see, I made the basic mistake of thinking of "phi._domain" as private functions, when they are in fact just attributes:
sage: phi._domain Chain complex with at most 4 nonzero terms over Finite Field of size 2 sage: phi._domain is phi._f._domain True
The Sage maps that I'm used to (mainly morphisms between CombinatorialFreeModules?) have public "domain" and "codomain" methods; this might have spared me a 5 minute detour in this case. I have no good idea, though, what to call the $H_0$ and $H_1$ of a homotopy $H_t$...
Also, I wonder if phi shouldn't better be printed like
sage: phi Chain homotopy between Chain complex morphism From: Chain complex with at most 4 nonzero terms over Finite Field of size 2 To: Chain complex with at most 4 nonzero terms over Finite Field of size 2 and Chain complex morphism From: Chain complex with at most 4 nonzero terms over Finite Field of size 2 To: Chain complex with at most 4 nonzero terms over Finite Field of size 2
At least this is the way morphisms between combinatorial free modules are printed:
sage: A,B=[CombinatorialFreeModule(GF(2),x) for x in [(1,2,3),(2,3,4)]] sage: f=A.module_morphism(codomain=B,on_basis = lambda x:x+1) ; f Generic morphism: From: Free module generated by {1, 2, 3} over Finite Field of size 2 To: Free module generated by {2, 3, 4} over Finite Field of size 2
comment:11 followup: ↓ 12 Changed 5 years ago by
doc does not build
WARNING: citation not found: GDR
comment:12 in reply to: ↑ 11 Changed 5 years ago by
Replying to chapoton:
doc does not build
WARNING: citation not found: GDR
Sorry, that should be GDR03
. I'll fix it.
comment:13 Changed 5 years ago by
As Christian implied, you should use .domain()
and .codomain()
in the docstring and try to make thing behave like other morphisms as much as possible (anything inheriting from Morphism
has some method like _repr_defn
or similar).
For the record, this is on my todo list, but I don't know when I might be able to do a full review.
comment:14 Changed 5 years ago by
 Commit changed from 698ba79f9837c3c2e83bd358441b4cd4a4073392 to b925f7c8f3b82027ae18e05f0bd93e0e9969ae04
Branch pushed to git repo; I updated commit sha1. New commits:
b925f7c  trac 6101: add domain and codomain methods.

comment:15 followup: ↓ 21 Changed 5 years ago by
Good suggestions, Christian. I've made some changes to try to address them.
Regarding $H_0$ and $H_1$ (currently ._f
and ._g
), I don't know good names either, but I also almost never needed to use those attributes (only when constructing the dual chain homotopy).
comment:16 Changed 5 years ago by
 Commit changed from b925f7c8f3b82027ae18e05f0bd93e0e9969ae04 to 36a0db648c8fc4c5e0c0ef35059dd13ad6ba0575
Branch pushed to git repo; I updated commit sha1. New commits:
d65ba8d  trac 19179: change _repr_ for chain maps, chain homotopies

c38751b  Merge branch 'chains' into ATmodel

f70e3a2  trac 19179: doctest fixes for _repr_ changes

92b17e8  Merge branch 'chains' into ATmodel

6cd5b7c  trac 6102: doctest fixes for _repr_ changes

36a0db6  whitespace fix

comment:17 followup: ↓ 19 Changed 5 years ago by
 Description modified (diff)
comment:18 Changed 5 years ago by
 Commit changed from 36a0db648c8fc4c5e0c0ef35059dd13ad6ba0575 to 048061cb41ace2fb9563cf99bb2f7f9fcbe80516
Branch pushed to git repo; I updated commit sha1. New commits:
048061c  trac 6101: add homology_morphism to ref manual

comment:19 in reply to: ↑ 17 Changed 5 years ago by
Replying to jhpalmieri:
(Same commits as before, but I also put them on #19179 and #6102, where they really belong.)
I'm a bit confused by these dependencies  what's the right place to start reviewing things: 6101, 6102, or 19179 ?
comment:20 Changed 5 years ago by
In principle, #19179 is first, then #6102, then this one. However, #19179 has one or two doctests that use things that aren't introduced until #6102, so they will fail. You can read the code there and run everything, though.
#6102 should build and pass doctests, so it would come next (or jointly with #19179). This ticket comes last. The new code for this ticket should be all in the commit labeled "trac 6101: maps in (co)homology induced by simplicial maps".
I suppose I should change the _repr_
method for induced homology maps here, but not right now.
comment:21 in reply to: ↑ 15 Changed 5 years ago by
Replying to jhpalmieri:
Good suggestions, Christian. I've made some changes to try to address them.
Regarding $H_0$ and $H_1$ (currently
._f
and._g
), I don't know good names either, but I also almost never needed to use those attributes (only when constructing the dual chain homotopy).
I wonder if we might use bracketnotation, so that a homotopy H
goes from H[0]
to H[1]
? But I don't know if there are precedents in Sage for custom __getitem__
, __setitem__
methods. And I'm also not sure, whether there's a real need for this...
comment:22 Changed 5 years ago by
 Commit changed from 048061cb41ace2fb9563cf99bb2f7f9fcbe80516 to 470ca9b742e1efc2159b004f1b4edd9fd038eea9
comment:23 Changed 5 years ago by
 Commit changed from 470ca9b742e1efc2159b004f1b4edd9fd038eea9 to 6e3e801ccb73a6dca8c5e4bd05da0e87a511d437
comment:24 Changed 5 years ago by
Okay, last up is this ticket. A first question from a quick lookover is could we use Morphism
as the base class?
comment:25 followup: ↓ 26 Changed 5 years ago by
 Status changed from needs_review to needs_work
This needs to be rebased on top of the most recent #6102. I'm working on it. I'll also look into using Morphism
. Does that mean that we also change the alreadyexisting ChainComplexMorphism
and SimplicialComplexMorphism
?
comment:26 in reply to: ↑ 25 Changed 5 years ago by
Replying to jhpalmieri:
I'll also look into using
Morphism
. Does that mean that we also change the alreadyexistingChainComplexMorphism
andSimplicialComplexMorphism
?
It gives us a little bit of extra features (like register_as_coercion()
) and more consistent class hierarchy, but it isn't a requirement. It might also make things easier once homsets become proper parents too, but that is so far down the road that it's not worth it the worry either.
comment:27 Changed 5 years ago by
 Commit changed from 6e3e801ccb73a6dca8c5e4bd05da0e87a511d437 to ec81de20addced6a723f51413c3ffb31eaf4bc13
Branch pushed to git repo; I updated commit sha1. Last 10 new commits:
0d2ea83  trac 6102:

a0f20c2  Merge branch 'u/jhpalmieri/ATmodel' of trac.sagemath.org:sage into u/jhpalmieri/ATmodel

716469a  Some smaller reviewer tweaks.

d00d31b  trac 6102: cup products for Delta complexes

12d4cc7  trac 6102: fix typo "left" > "right"

8b59ed3  Making (co)homology into a graded module (algebra).

3751753  Making Sq work for inhomogeneous elements.

9bfc2d2  trac 6102: documentation, minor cleanup

8e761de  Merge branch 't/6102/ATmodel' into t/6101/inducedmaps

ec81de2  trac 6101: make morphisms inherit from 'Morphism'

comment:28 Changed 5 years ago by
Okay, here's an attempt at using Morphism
. I also used Parent
for SimplicialComplex
(instead of CategoryObject
). The only issue with this is that the test suite for simplicial complexes fails unless I skip _test_category
.
comment:29 Changed 5 years ago by
 Milestone changed from sage6.4 to sage6.10
Would you be opposed to basing this on #18175 (which should be positively reviewed soon I hope)? This might solve the issues with _test_category
, but I suspect this is more due to a lack of an element class. If you were to use the category of (abstract) simplical complexes from #18175, the intent was that elements correspond to faces.
I should also ask how you feel about using morphism, does it feel any more natural to do things this way or do you think that continuing to work on this isn't worth the time (at this stage)?
comment:30 Changed 5 years ago by
If I try to merge #18175, the test suite fails more tests than before:
The following tests failed: _test_an_element, _test_elements, _test_elements_eq_reflexive, _test_elements_eq_symmetric, _test_elements_eq_transitive, _test_elements_neq, _test_some_elements
because
NotImplementedError: please implement _an_element_ for Simplicial complex ...
I'm not sure what _an_element_
should return: a simplex (since a simplicial complex is a set of simplices)? a vertex (since their images determine maps)? a facet (since they determine the complex)?
As far as morphisms go, it was not hard to switch to them (assuming I did everything right).
comment:31 Changed 5 years ago by
I would just have it return the first facet (and the empty simplex for the empty complex), but it just has to be any element of the parent object (which could include the empty simplex).
comment:32 Changed 5 years ago by
Going this route tempts me to make Simplex
inherit from Element
, but you might want to construct abstract instances of Simplex
which are not elements of any particular complex. So I can't use the Element.__init__
method, which requires setting a parent. As a result, I don't think I can use _element_constructor_
for simplicial complexes. Instead, I can do something like this (plus doctests):

src/sage/homology/simplicial_complex.py
diff git a/src/sage/homology/simplicial_complex.py b/src/sage/homology/simplicial_complex.py index 8d00cd6..366c1db 100644
a b class SimplicialComplex(Parent, GenericCellComplex): 1077 1077 """ 1078 1078 return self._vertex_set 1079 1079 1080 def _an_element_(self): 1081 return self.facets()[0] 1082 1083 def __contains__(self, x): 1084 if not isinstance(x, Simplex): 1085 return False 1086 dim = x.dimension() 1087 return x in self.n_faces(dim) 1088 1089 def __call__(self, simplex): # not _element_constructor_ 1090 if simplex not in self: 1091 raise ValueError 1092 return simplex 1093 1080 1094 def maximal_faces(self): 1081 1095 """ 1082 1096 The maximal faces (a.k.a. facets) of this simplicial complex.
Then the full test suite passes.
comment:33 Changed 5 years ago by
We can always subclass Simplex
as something like SimplexElement
and changing out some calls to Simplex(data)
to self.element_class(self, data)
. I wouldn't really want to overwrite __call__
unless we really have to... I can also do some of the work for this changeover as well.
comment:34 Changed 5 years ago by
 Commit changed from ec81de20addced6a723f51413c3ffb31eaf4bc13 to 953e0a6489a3eca48b13b1e91ece798c51a6fe5c
Branch pushed to git repo; I updated commit sha1. Last 10 new commits:
3647305  Some cleanup and adding more category information to particular sets.

f810aa6  Reworking the category of manifolds.

375ff46  Fixing doctest failures and letting a few other rings know they are metric spaces.

8b851a0  Merge branch 'develop' into public/categories/topological_metric_spaces18175

f8f5b93  Fixing last remaining doctests.

041a5d1  Adding padics to metric spaces and some cleanup.

bfa0cdf  One last doc tweak.

d13c368  Fixing doc of metric spaces.

9d38c71  trac 6101: Merge branch 't/18175/public/categories/topological_metric_spaces18175' into t/6101/inducedmaps

953e0a6  trac 6101: add _an_element_, etc., to SimplicialComplex

comment:35 Changed 5 years ago by
 Dependencies changed from #19179, #6102 to #19179, #6102, #18175
 Status changed from needs_work to needs_review
I tried implementing SimplexElement
and it turned into a real mess, so I gave up. Instead, here is my version using __call__
.
comment:36 Changed 5 years ago by
Oh, and I would welcome any help with this or other approaches.
comment:37 Changed 5 years ago by
I will work on this later this week.
comment:38 Changed 5 years ago by
 Branch changed from u/jhpalmieri/inducedmaps to public/algtop/induced_maps6101
 Cc nthiery SimonKing added
 Commit changed from 953e0a6489a3eca48b13b1e91ece798c51a6fe5c to b6c565763a2df99471bb6b574f60d15450b44c2e
 Reviewers set to Travis Scrimshaw
Man...non hashable parents really throw a big wrench into the coercion framework. In some ways that is not surprising, as they get put into a big dict
to store the necessary coercions, but I would have liked it to better handle nonhashable parents.
Well, one of the problems that you had was Element
tries to use coercion on elements which do not define rich comparisons, which caused problems until I changed Simplex.__cmp__
to rich comparisons (I <3 functools.total_ordering
). However, even with that change, I could not get the combination of class Element(Simplex, Element)
in SimplicialComplex
to pickle (granted, I did not try too hard). I think the problem is that SimplicialComplex
is defined by a set of its elements, and we get a loop of sorts with the pickling...
Therefore I decided to just add a warning message to SimplicialComplex
instead. Simon, Nicolas, the reason I am ccing you is to make sure that this abuse of Parent
is acceptable to you (or if you had any thoughts about the pickling issues). At least until a followup ticket takes care of things.
New commits:
b6c5657  Add a warning about the abuse of Parent in simplicial complex.

comment:39 Changed 5 years ago by
 Commit changed from b6c565763a2df99471bb6b574f60d15450b44c2e to e9150caff3d157cadd739042461d972c9d5952f1
Branch pushed to git repo; I updated commit sha1. New commits:
e9150ca  Merge branch 'develop' into public/algtop/induced_maps6101

comment:40 Changed 5 years ago by
John, if nobody objects to the current setup in a few days, then I think we should set this to positive review.
comment:41 Changed 5 years ago by
Sounds great. I certainly don't have any objection to your warning.
comment:42 Changed 5 years ago by
Okay, I think we've let enough time for a response pass. Feel free to set a positive review if you're still in agreement.
comment:43 Changed 5 years ago by
 Status changed from needs_review to positive_review
comment:44 Changed 5 years ago by
 Branch changed from public/algtop/induced_maps6101 to e9150caff3d157cadd739042461d972c9d5952f1
 Resolution set to fixed
 Status changed from positive_review to closed
New commits:
trac 19179: chain homotopies, chain contractions, and duals of chain maps
trac 6102: cup products, cohomology rings, mod 2 cohomology operations
change AssertionError to other errors
Merge branch 'chains' into ATmodel
change AttributeErrors to other errors
trac 6101: maps in (co)homology induced by simplicial maps