Confusion between values and indices

[matbesancon]

It might be due to the solver being tested first on Sudoku instances. When testing the solver on problems with non-integer coefficients, some errors occur, which can be fixed by widening the types.

However, in some places of the code, the value and index spaces are mixed up. An example with the has function:

function has(v::CS.Variable, x::Int)
    if x > v.max || x < v.min
        return false
    end
    ind = v.indices[x + v.offset]
    return v.first_ptr <= ind <= v.last_ptr
end

In the first part, x is used as a value, checking if the variable is between min and max. But in the second part, it is used as an index as x + v.offset. Using values and indices interchangeably will lead to bugs later (what if you have infinite, if you have float coefficients, etc).

[Wikunia]

Floating coefficients shouldn't affect the discrete variables, right? I think for non discrete or unbounded variables (which don't exist yet) a different variable type might need to exist.
Not sure how to rewrite it otherwise without loosing performance on these operations.

[matbesancon]

Yes they can affect them, even for pure integer variables, you can have floating constraints a^T x <= b with a non-integral

…

On Sun, Jan 12, 2020, 14:42 Ole Kröger ***@***.***> wrote: Floating coefficients shouldn't affect the discrete variables, right? I think for non discrete or unbounded variables (which don't exist yet) a different variable type might need to exist. Not sure how to rewrite it otherwise without loosing performance on these operations. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#56?email_source=notifications&email_token=AB2FDMRLRCA4VTX3TWYQLHDQ5MM23A5CNFSM4KFXV3XKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIW2I5Y#issuecomment-573416567>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB2FDMTCDLSAYIWAXHHCSPTQ5MM23ANCNFSM4KFXV3XA> .

[matbesancon]

The problem is explained in more detail in https://en.wikipedia.org/wiki/Independent_set_(graph_theory)

[Wikunia]

I think this can be closed if I take care that only discrete variables get checked here, right?

[matbesancon]

Yup all right :)

…

On Thu, Jan 16, 2020, 19:56 Ole Kröger ***@***.***> wrote: I think this can be closed if I take care that only discrete variables get checked here, right? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#56?email_source=notifications&email_token=AB2FDMXGHYX72W7DDFJA3FLQ6CUWDA5CNFSM4KFXV3XKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEJFEV4Y#issuecomment-575294195>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB2FDMWOQKAGKFMRRY776YLQ6CUWDANCNFSM4KFXV3XA> .