esabo / codingtheory Goto Github PK

RotatedSurfaceCode(5) works
RotatedSurfaceCode(4) doesn't

It would be nice if Pkg.add(url="https://github.com/esabo/CodingTheory") worked directly. We'll need this before we can consider adding this package to the Julia registry anyway.

Biggest roadblock to easily adding the package as above is the current version of Oscar changed so many things. We should add a [compat] entry for it to Project.toml until we find time to make it work with the newer version. Probably need compat for a few other packages. #21 will be solved if we do this I think.

Happens when calling LDPCCode() with any matrix given as an argument.

Also, the examples haven't been updated yet either.

Replicatable on both the master and QWRv1 on Julia v1.10

The Bacon-Shor constructor may not produce the correct lattice for some values of m != n. Reasoning unclear. Probably a typo.

I understand the reasoning behind having types be in one file, but I was wondering if we could break up the types by topic and place them in their relevant subfolders considers that the type break up nicely. I am happy to do this and open a PR with this. I was thinking

src
. Classical
. . types.jl
. LDPC
. . types/
. . . types_channels.jl
. . . types_code.jl
. Quantum
. . types.jl

Open to suggestions. Am happy to take this up as well

Recent update to is_isomorphic has broken the subsystem code constructor.

The Reed-Solomon constructor is returning the wrong generator polynomial, yet the same parameters work correctly in the BCH constructor.

julia> Gray_code_minimum_distance(LinearCode(ToricCode(3).X_stabs))
(0, [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0])

Either the information sets are reduced to include a zero row, which is then picked out by a weight one element in the enumeration, or the enumeration vector happens to exactly contain the redundancies of the generators. In both cases, the all-zero vector appears to be a codeword.

Can probably be fixed easily by simply always using the standard form or calling _remove_empty(, :rows) for each information set depending on what the problem is. Using standard form should fix it in both cases.

julia> T = ToricCode(3); all(is_logical(T, T.logs_mat[i, :]) for i in nrows(T.logs_mat))
true

julia> T = ToricCode(4); all(is_logical(T, T.logs_mat[i, :]) for i in nrows(T.logs_mat))
false

Probably related (via copy/paste) to #10.