julian-upc / 2015-algorithms Goto Github PK

Dear all,

• In dancing_links/c++ you can find the material we discussed today;
• in coxeter_group_orbits/ you can find a specification and some skeleton files for the new project.

Have fun!

If I understand the specification correctly, the normal vectors for An have dimension n+1.
Does the input vector still have dimension n? How would that work? Or is it a special case and in this case the vector also has dimension n+1?

According to which axis shall we flip pentominos?
Also, when a method moves a pentomino out of our range, what shall we do with that, just leave it where it is? Give an error?

We need to have a conversation about how everybody can turn in their code without stepping all over other people's code. Having everybody turn in their code into the same directory clearly won't work, so how about setting up some subdirectories?

The test uses this function, that our pentomino class does not have, shall we add it?
Also, what should it do? Shall we just hardcode the right numbers, or shall it create TileSets and calculate the size?

So I would like to start a discussion about the TileSet definition.
As I see it a pentomino can be in a TileSet multiple times, exactly once per orientation it has.
But two pentominos that have same orientation but different coos, can those be both in one TileSet?

How exactly is the pattern to name the cells in the matrix.
I thought it was something like Rowname + Columnname, where the Rowname is like I[4] if this is the forth row added for the pentomino I, but that seems to not fit the pattern used in the tests.

The test, that wants to check whether the 'Y' pentomino can be added in each orientation does not make deepcopies and therefore also rotates and flips the pentomino already in the set. Thereby the size o fthe set will always be 1.

What exactly is this function supposed to do?
There is no testcase from which I could get ideas about this.

The test wants I to have the same coordinates after turning it twice, that demands one specific turning point.
If we normalize after turning anyway to get legal coordinates, why does it matter whether I moves when turned twice?

Ich bin mir nicht sicher, was hash(self) als Funktion genau ausgeben soll, nehme aber an, dass ein int erfordert wird. Stimmt das so?

It seems the function insert from std::set uses the "<" operator to insert objects and prohibit multiple entries of the same object.
Is there another good container or solution to fix this?

How should we handle invalid placements, meaning placements including names we do not have as columns?

We have in the init-method the code-line self.dim = len(coos[0]).
Does that mean we have to consider that we might get more then 2 dimensions later on?
That would make turning much more difficult.