Bi-Directional Dictionary

[Martin Albrecht]

Hi there,

I'm working on a C interface to SCIP

http://scip.zib.de

and for that I need a bi-directional dictionary, that is a dictionary where
you can query in both directions X->Y and Y->X. Of course, a simple
implementation is quickly done using two dictionaries but I was wondering:
since this is a general data structure, maybe someone else already implemented
it nicely in Sage under a different name? ... building the car instead of re-
implementing bi-directional dictionaries.

Cheers,
Martin

--
name: Martin Albrecht
_pgp: http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x8EF0DC99
_otr: 47F43D1A 5D68C36F 468BAEBA 640E8856 D7951CCF
_www: http://martinralbrecht.wordpress.com/
_jab: martinr...@jabber.ccc.de

[Nils Bruin]

This might do what you want (but it's not in sage as far as I know):

http://pypi.python.org/pypi/bidict/

[Nathann Cohen]

> Hi there,
>
> I'm working on a C interface to SCIP

Hello Martin !!!

When you say C interface to SCIP, are you talking about a Cython
interface ? Are you talking about an interface between SCIP and Sage ?
I have just been finishing a patch using constraint generation, and I
was thinking about changing a lot of things in the
MixedIntegerLinearProgram constraint... I wanted to build a Cython
class (a backend) for each solver (CPLEX, Coin, GLPK, gurobi perhaps
too), so that it may be possible to use them directly through C calls,
and perhaps having one "natural" way to work on constraint/variable
generation through Sage.

If you think what you are doing and what I am -- just planning -- may
overlap, perhaps we should discuss it ?.... :-)

Nathann

[Martin Albrecht]

Hi Nathann,

> When you say C interface to SCIP, are you talking about a Cython
> interface ?

Indeed, a Cython interface to SCIP.

> Are you talking about an interface between SCIP and Sage ?

So far my interface supports:
- linear, XOR and AND constraints (SCIP is a constraint integer programming
solver, supporting many different kinds of constraints)
- parameter getting/setting ("limits/time" etc.)
- conversion from MixedIntegerLinearProgram
- conversion from MPolynomialSystem_gf2 (using AND and XOR or linear
constraints)
- returning all solutions not just the optimal solution

the interface so far is very SCIP-y insofar as it isn't very natural Sage
style: no operator overloading etc.

> I have just been finishing a patch using constraint generation, and I
> was thinking about changing a lot of things in the
> MixedIntegerLinearProgram constraint... I wanted to build a Cython
> class (a backend) for each solver (CPLEX, Coin, GLPK, gurobi perhaps
> too), so that it may be possible to use them directly through C calls,
> and perhaps having one "natural" way to work on constraint/variable
> generation through Sage.

I think that would be a good idea, since they all seem to have specialised
features. For instance, IIRC Harald told me that Gurobi allows to return those
constraints which conflict in an infeasible problem.