Add priceability module

[drknzz]

• Use mip>=1.16rc0
  • mip==1.15 had inconsistent mapping between gurobi optimization status and mip library optimization status (see v1.15, v1.16)
• Add priceability module

[Simon-Rey]

Dear Kamil,

Thank you so much for the PR. It is much appreciated. Just two comments:

• The tests fail for Python 3.10 and 3.9, check what the problem is.
• I do not want to use the "typing" module which may be deprecated in the future (see e.g., https://stackoverflow.com/questions/66738753/python-typing-deprecation). You can simply use the built-in types for typing list and dict.
• It would be nice to include how to use your code in the docs. That should go to the file pabutools/docs-source/source/usage/analysis.rst. I made you a priceability section there. You can take example from other parts of the docs to see what to put.

I did not go through the code in details but I trust you that it works as expected.

Simon.

[Simon-Rey]

By the way, the docs thing does not have to be now, but it's good to do it at some points. You can wait to have a stable version of your function to do so.

[drknzz]

Removed the deprecated typing imports 👍

As for the tests, it seems that there have been some precision problem when running the model with the cbc solver. Locally I am using gurobi instead, so that's why I haven't caught it previously. I've tweaked some parameters in mip and it seems to be working with the cbc solver.

[drknzz]

• It would be nice to include how to use your code in the docs. That should go to the file pabutools/docs-source/source/usage/analysis.rst. I made you a priceability section there. You can take example from other parts of the docs to see what to put.

Sure thing, there should be some guide on the function. Will submit a separate PR, after the relaxation part, as suggested.

[Simon-Rey]

Great, that's all fixed now.

Another thing I'm only seeing now: for both validate_price_system and priceable you have a * argument. What's the motivation there? I do think them being used and in general I think it's cleaner to restrict to the arguments you need.

[drknzz]

Great, that's all fixed now.

Another thing I'm only seeing now: for both validate_price_system and priceable you have a * argument. What's the motivation there? I do think them being used and in general I think it's cleaner to restrict to the arguments you need.

I don't think I quite follow. The *, makes it so the following arguments are keyword-only, so one has to be explicit when passing them. They can be, and are meant, to be used as usual. You cannot pass more arguments than the specified positional arguments when using * (as they are not bound to a variable, like in *args).

i.e. calling

fun(1, 2, 3, c=4)
fun(1, 2, 3)

for a function signature

def fun(a, b, *, c=10)

wouldn't work

[Simon-Rey]

Ooh I see, I did not know that. Thanks for the tip :)