Constraints do not match RHS based on coordinates

[thomgeo]

Version Checks (indicate both or one)

• I have confirmed this bug exists on the lastest release of Linopy.

• I have confirmed this bug exists on the current master branch of Linopy.

Issue Description

When constants are supplied to a linopy model, the model appears to introduce these only based on order, not match them based on the relevant coordinates.

Reproducible Example

import linopy

m = linopy.Model()

idx = ["A1", "A5", "A11", "A100"]

x = m.add_variables(coords=[idx], name="x")

idx2 = ["A100", "A1", "A5", "A11"]

coef = pd.Series({i: int(i[1:]) for i in idx2})
m.add_constraints(x >= coef)

Expected Behavior

linopy should build the constraint so that the variables are matched with the constant of the same index.

[A1]: +1 x[A1]     ≥ 100.0

[A5]: +1 x[A5]     ≥ 1.0

[A11]: +1 x[A11]   ≥ 11.0

[A100]: +1 x[A100] ≥ 5.0

Instead, it returns a constraint of the form:

[A1]: +1 x[A1]     ≥ 1.0

[A5]: +1 x[A5]     ≥ 5.0

[A11]: +1 x[A11]   ≥ 11.0

[A100]: +1 x[A100] ≥ 100.0

Installed Versions

Details

0.5.8

[hugoalgarvio]

Yes, the "lower_bound_constraint" is not aligned with the "coef" index:

coef:
A100 100
A1 1
A5 5
A11 11
dtype: int64

Constraint:
Constraint lower_bound_constraint [dim_0: 4]:

[A1]: +1 x[A1] ≥ 100.0
[A5]: +1 x[A5] ≥ 1.0
[A11]: +1 x[A11] ≥ 5.0
[A100]: +1 x[A100] ≥ 11.0

[FabianHofmann]

@thomgeo yes, you are rainsing a good point here. linopy has the feature that coordinates of objects with the same shape use a override join, saying the coordinates of the second and following objects are ignored. but I more and more see how confusing this is, why I came up with #572 . but I am not there yet. @FBumann raised some points that we should target some standard mathematical conventions as we now it from school commutativity, associative law etc. it seems that the inner join would be the way to go. but this will be breaking, likely to come with v1

for now you can use linopy.align(lhs, rhs) and then assign the constraint. after #572 you can also do lhs.ge(rhs, join="inner")

[FBumann]

@thomgeo If you want to you can join in on the discussion in #591, although I won't continuously working on it for the next 2 weeks