varEq on a BinVar is ignored.

[ghost]

Suppose we have an LP v c with setVarKind x BinVar at the top. Later on, I might like to set varEq x 0 or even to 1, but never a non-binary value, to rule x out or make it certain under chosen conditions. Currently this varEq is ignored and the only constraint that goes through to the linear programme is that 0 <= x <= 1 and that x is general.

[ghost]

Looking into this further (Data.LinearProgram.GLPK.Internal), this seems to be more of a feature of GLPK. What's your opinion on this?

[jyp]

I must say I don't really understand your issue. Further I have not used this package for a couple of years so your guess is probably as good as mine.

…

On Wed, Aug 1, 2018 at 4:05 PM, justinpastoncooper ***@***.*** > wrote: Looking into this further (Data.LinearProgram.GLPK.Internal), this seems to be more of a feature of GLPK. What's your opinion on this? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#10 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABsY4DzDop1NM_TkhXUZL9Ml9ozoNmdks5uMbWzgaJpZM4Vqhz7> .

[ghost]

Maybe I didn't explain in the best way. setVarKind x BinVar followed by varEq x 0 is equivalent to setVarKind x BinVar alone. The varEq x 0 is ignored. I first assumed that this was a problem with glpk-hs, but it turns out that it is a "feature" of glpk itself.

[jyp]

Thanks for documenting the issue. Perhaps you'll want to push this upstream.

[ghost]

Sorry, I only just saw your message. Maybe it's better to retain the current behaviour so that it mirrors the glpk library?

[jyp]

This can either be fixed upstream, or we can add a workaround (for example by adding a prefix to all variables). In the latter case I'll be happy to get a pull request.

[ghost]

I sadly don't have access to the computer I wrote the fix on right now. I will get back to this when I do.