CombinatorialPolyhedron: replace attributes by methods, make names more consistent with Polyhedron

[gh-kliem]

Replace attributes in CombinatorialPolyhedron by methods, such that they can potentially be lazily evaluated.

This is motivated by #10777.

While at it, we will change some names of methods for consistency:

We make CombinatorialPolyhedron more consistent with Polyhedron_base, most importantly:

• edge_graph -> vertex_graph,
• ridge_graph -> facet_graph,
• unbounded(self) -> is_bounded(self),
• length_Hrep -> n_Hrepresentation,
• length_Hrep -> n_Hrepresentation.

For CombinatorialFace we do the following changes:

• face.Vrep(True) -> face.ambient_Vrepresentation(), face.Vrep(False) -> face.ambient_V_indices(),
• n_Vrepresentation -> n_ambient_Vrepresentation,
• the above applied to Hrep.

We also delete .cc in the module name of bit_vector_operations. The .cc lead to an empty folder in the source directory. Without this folder does not reappear.

In addition we do some minor bug fixing and some improvement in the documentation.

[gh-kliem]

Some attributes like bitrep_vertices might have been an unfortunate choice:

As they are accessed by other classes, I'm forced to compute e.g. bitrep_vertices on initialization. If CombinatorialPolyhedron was ever to be a parent of Polyhedron_base (see #10777), this would force a polyhedron to compute the incidence matrix right away.

From my understanding, the decorator @lazy_attribute is not available for extension types directly. I'm thinking of changing those attributes to methods, such that there is a method bitrep_facets and an attribute _bitrep_facets. This way one can later change this to be lazily evaluated.

Any thoughts?

[gh-kliem]

New commits:

​d6d663a	combinatorial polyhedron uses far face as a bug fix
​14d17a8	added doctest reporting the bug fix
​806a217	correct link to trac ticket, if unbounded is False -> if not unbounded
​9194627	replace attributes by methods
​0aed9fd	removed empty folder being created in source

[git]

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

​d597ed3	replace attributes by methods
​2fc4fe0	removed empty folder being created in source

[tscrim]

If a @lazy_attribute doesn't work, then just make it a @cached_method (with no arguments).

[gh-kliem]

Replying to tscrim:

If a @lazy_attribute doesn't work, then just make it a @cached_method (with no arguments).

Does either one work with cdef?

Here, I want to give a small example of what I want to do:

sage: cython('''
....: cdef class BaseClass:
....:     cdef public int number
....: 
....:     def __init__(self):
....:         self.number = 2
....: ''')
sage: class NewClass(BaseClass):
....:     def __init__(self):
....:         pass
....:     
sage: example = NewClass()
sage: example.number
0

I'm trying to somehow make number be a lazy attribute. The only solution I came up with, is to replace it by a method. Is there another way?

The background is that CombinatorialPolyhedron does require Polyhedron to compute the incidence matrix for correct initialization. If CombinatorialPolyhedron is to be a base class of Polyhedron_base at some point (#10777), it needs to be lazily initialized.

[gh-kliem]

Is this the right approach, am I missing something? Is it acceptable to do those name changes? (CombinatorialPolyhedron still was not part of a stable release)

If so, I think the review is rather trivial and I can try to find someone else to do it.

[jipilab]

Some comments:

Since these methods will then occur in the namespace, perhaps it would make sense to make it consistent with the names of the methods for Polyhedron. For example, unbounded -> is_(un)bounded. Probably it would make more sense to store is_bounded rather than unbounded....

Similar for V and H, having a method with only one letter is not so revealing...

I guess it is important to be careful with the naming at this precise moment as after, changing the name of methods will force the usage of deprecation warnings and that's annoying. (I believe that some methods should be changed to be consistent with Polyhedron...).

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. Last 10 new commits:

​037b3ed	ridge_graph -> facet_graph; some bug fixing
​07e4ea7	f_vector is a vector now
​c18a757	abbreavations: Hrepr -> Hrep, Vrepr -> Vrep etc.
​4d165f5	unbounded -> bounded, bounded(self) -> is_bounded(self)
​bbd5c74	length_Hrep -> n_Hrepresentation; likewise for Vrep
​21b9e03	removed attribute Vinv, as its not being used
​c844ec5	made specification to n_Hrepresentation in combinatorial_face
​7403c4b	Hrep -> facet_names
​b8cd255	V -> Vrep
​5f0f704	added docstrings to the new methods

[git]

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

​7ef7332

meaningful test for graph without edges

[git]

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

​d02dee5

typos; length_atom_rep -> n_atom_rep

[gh-kliem]

Ok, I think naming is consistent with Polyhedron_base now.

It would be great to have this in sage 8.9, so that we can avoid any deprecation warning.

[jipilab]

Well... 8.9.rc0 is already out... so time is running out... :-(

On which version of Sage is this based? Could you rebase on 8.9.rc0 so that we have a clear view of what is done in this ticket. Or is it actually already only stuff from this ticket?

• I would vote for Vrep->Vrepresentation same for H. This is what I meant originally with be consistent with Polyhedron, I should have specified it, my bad.

• You should also mention in the description of this ticket that you remove an empty folder. This is done in one of the commits. (This is a good habit to take for such small changes, otherwise, the penalty is: "please open a ticket just for that"...).

[gh-kliem]

Replying to jipilab:

Well... 8.9.rc0 is already out... so time is running out... :-(

On which version of Sage is this based? Could you rebase on 8.9.rc0 so that we have a clear view of what is done in this ticket. Or is it actually already only stuff from this ticket?

This ticket is huge, but almost all changes are trivial. Sorry, but that's what happens, when you change some names.

• I would vote for Vrep->Vrepresentation same for H. This is what I meant originally with be consistent with Polyhedron, I should have specified it, my bad.

In Polyhedron there is Vrepresentation and Vrep-generator. So in a way, both things are used.

Now for CombinatorialPolyhedron there is only Vrepresentation in the namespace. _Vrep is an internal cython variable, while Vrep is the cython method that accesses _Vrep. I tried to use cpdef for both of them, but that didn't work (closures inside cpdef functions not yet supported, seems to be connected with smallInteger). So as long as it is that way, I would propse leaving the python method named Vrepresentation and the cython method Vrepr. But I'm open for different suggestions.

Now with CombinatorialFace things are a bit different. At the moment there is Vrep, Hrep, n_Vrepresentation and n_Hrepresentation (not consistent, I just noticed). Vrep allows giving both the indices and the names of the Vrepresentation. So in a way it combines ambient_Vrepresentation and ambient_V_indices of PolyhedronFace. What do you suggest?

One option would be the following for CombinatorialFace:

• Vrep -> ambient_Vrep
• make ambient_Vrepresentation an alias for ambient_Vrep(names=True)
• make ambient_V_indices an alias for ambient_V_indices(names=False)
• n_Vrepresentation -> n_ambient_Vreprentation
• same for Hrep

Unfortunatly, those changes are going to make the ticket even larger.

What do you think? If I have a clear idea what to do, I think I can make the remaining changes tonight.

• You should also mention in the description of this ticket that you remove an empty folder. This is done in one of the commits. (This is a good habit to take for such small changes, otherwise, the penalty is: "please open a ticket just for that"...).

[jipilab]

Replying to gh-kliem:

Replying to jipilab:

Well... 8.9.rc0 is already out... so time is running out... :-(

On which version of Sage is this based? Could you rebase on 8.9.rc0 so that we have a clear view of what is done in this ticket. Or is it actually already only stuff from this ticket?

This ticket is huge, but almost all changes are trivial. Sorry, but that's what happens, when you change some names.

I was just making sure. Let's try to make it happen.

• I would vote for Vrep->Vrepresentation same for H. This is what I meant originally with be consistent with Polyhedron, I should have specified it, my bad.

In Polyhedron there is Vrepresentation and Vrep-generator. So in a way, both things are used.

V_rep is shortened because _generator appears after. So I would vote for completing the name to Vrepresentation.

Now for CombinatorialPolyhedron there is only Vrepresentation in the namespace. _Vrep is an internal cython variable, while Vrep is the cython method that accesses _Vrep. I tried to use cpdef for both of them, but that didn't work (closures inside cpdef functions not yet supported, seems to be connected with smallInteger). So as long as it is that way, I would propse leaving the python method named Vrepresentation and the cython method Vrepr. But I'm open for different suggestions.

Got it! Looks good.

Now with CombinatorialFace things are a bit different. At the moment there is Vrep, Hrep, n_Vrepresentation and n_Hrepresentation (not consistent, I just noticed). Vrep allows giving both the indices and the names of the Vrepresentation. So in a way it combines ambient_Vrepresentation and ambient_V_indices of PolyhedronFace. What do you suggest?

Ok. I see. I would do Vrep to Vrepresentation. Then, if CombinatorialFace has a different behavior than the PolyhedronFace I believe it is fine.

One option would be the following for CombinatorialFace:

• Vrep -> ambient_Vrep
• make ambient_Vrepresentation an alias for ambient_Vrep(names=True)
• make ambient_V_indices an alias for ambient_V_indices(names=False)
• n_Vrepresentation -> n_ambient_Vreprentation
• same for Hrep

It seems like making the current Vrep private and have ambient_Vrepresentation (with the whole word) call this private function Vrep(names=True) and likewise for ambient_V_indices be more reasonable.

Why have 3 functions when we can have 2? Especially, the 2 functions behave as they do in PolyhedronFace... so I do not see the need to provide Vrep in the namespace.

• You should also mention in the description of this ticket that you remove an empty folder. This is done in one of the commits. (This is a good habit to take for such small changes, otherwise, the penalty is: "please open a ticket just for that"...).

Please apply this change to the description.

[gh-kliem]

Replying to jipilab:

Replying to gh-kliem:

Replying to jipilab:

Well... 8.9.rc0 is already out... so time is running out... :-(

On which version of Sage is this based? Could you rebase on 8.9.rc0 so that we have a clear view of what is done in this ticket. Or is it actually already only stuff from this ticket?

This ticket is huge, but almost all changes are trivial. Sorry, but that's what happens, when you change some names.

I was just making sure. Let's try to make it happen.

• I would vote for Vrep->Vrepresentation same for H. This is what I meant originally with be consistent with Polyhedron, I should have specified it, my bad.

In Polyhedron there is Vrepresentation and Vrep-generator. So in a way, both things are used.

V_rep is shortened because _generator appears after. So I would vote for completing the name to Vrepresentation.

Now for CombinatorialPolyhedron there is only Vrepresentation in the namespace. _Vrep is an internal cython variable, while Vrep is the cython method that accesses _Vrep. I tried to use cpdef for both of them, but that didn't work (closures inside cpdef functions not yet supported, seems to be connected with smallInteger). So as long as it is that way, I would propse leaving the python method named Vrepresentation and the cython method Vrepr. But I'm open for different suggestions.

Got it! Looks good.

Now with CombinatorialFace things are a bit different. At the moment there is Vrep, Hrep, n_Vrepresentation and n_Hrepresentation (not consistent, I just noticed). Vrep allows giving both the indices and the names of the Vrepresentation. So in a way it combines ambient_Vrepresentation and ambient_V_indices of PolyhedronFace. What do you suggest?

Ok. I see. I would do Vrep to Vrepresentation. Then, if CombinatorialFace has a different behavior than the PolyhedronFace I believe it is fine.

One option would be the following for CombinatorialFace:

• Vrep -> ambient_Vrep
• make ambient_Vrepresentation an alias for ambient_Vrep(names=True)
• make ambient_V_indices an alias for ambient_V_indices(names=False)
• n_Vrepresentation -> n_ambient_Vreprentation
• same for Hrep

It seems like making the current Vrep private and have ambient_Vrepresentation (with the whole word) call this private function Vrep(names=True) and likewise for ambient_V_indices be more reasonable.

Why have 3 functions when we can have 2? Especially, the 2 functions behave as they do in PolyhedronFace... so I do not see the need to provide Vrep in the namespace.

• You should also mention in the description of this ticket that you remove an empty folder. This is done in one of the commits. (This is a good habit to take for such small changes, otherwise, the penalty is: "please open a ticket just for that"...).

Please apply this change to the description.

I will. It's just one of those things that where perfectly clear to me, so I do didn't comment.

One more thing: Lots of changes are due to the changes of abbreciations Vrepr -> Vrep etc. I noticed you seem to at least use both. I could undo this change and this might make the ticket shorter.

[git]

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

​7b6bd9e	face.Vrep(True) -> face.ambient_Vrepresentation(), face.Vrep(False) -> face.ambient_V_indices(); same for Hrep
​9d506b8	for CombinatorialFace: n_Vrepresentation -> n_ambient_Vrepresentation; same for Hrep