the working structjump code

[fqiang]

• added the othermap instead of othervar. used to track of the mapping of parent -> child dummy variable indices.
• added function for computing list of scenario IDs on a local process.
• added some StructJuMP example models that can be solved by current supported solvers, Ipopt and PIPS.

[mlubin]

@joehuchette, now that PIPS is publicly released, could we set up travis CI here?

[joehuchette]

Since it's PIPS specific, it seems like src/pips_structure_interface.jl should go in PIPS.jl or somewhere similar?

[fqiang]

@joehuchette Julia Structure interface for PIPS-NLP is at repository in PIPS, which I think serves the same purpose as the PIPS.jl. pips_structure_interface.jl is a glue code between them. I think it can go with either one. However I think it is a good demonstration of the usage of the StructJuMP.

[fqiang]

@joehuchette maybe put it in another folder, but I am not sure what's the best way to present this.

[joehuchette]

I really think the PIPS specific code needs to go somewhere else; this repo should only include solver-agnostic modeling code. PIPS.jl seems like the natural place to put it.

[joehuchette]

Thanks @fqiang, looks good now. I have a few small comments. Can you please address those, and then squash this down to a single commit?

[fqiang]

@joehuchette Thanks. Just merge the commits into 35e7412. Let me know if that works.

[joehuchette]

Didn't seem to work, there are still 73 commits, and merge conflicts with master.

[fqiang]

resolved the conflict, do you know the command to merge commits into one. or I will need to do another PR to achieve this? @joehuchette

[joehuchette]

You can use git rebase -i master. For each commit besides the first, change pick to f and step through the rebase. You can just open another PR if you run into issues, though.

[fqiang]

close this one to open a new PR.