Using sum_constraints with "later" dimensions breaks opt.ask()

[dk-teknologisk-mon]

Sum_equals constraints break if you use a set of dimensions that do not start from the first one and continue upwards. E.g. constraining dimensions [0, 1, 2] works, but constraining dimensions [1, 2, 3] does not. The error is caused by the step that checks whether candidate points in the constrained dimensions actually lie within the space.

To reproduce

##% Import packages
from ProcessOptimizer import Optimizer
from ProcessOptimizer.space.constraints import SumEquals

# Step 1: This code works just fine
seed = 42

dimensions = [
    (200.0, 450.0),
    (50.0, 450.0),
    (0.0, 450.0),
    (10.0, 60.0),
]
# Build optimizer
opt = Optimizer(
    dimensions=dimensions,
    lhs=False,
    acq_func="EI",
    n_initial_points=5,
    random_state=seed,
)

# Create a constraint among the three factors with the 450 upper limit
constraints = [SumEquals(dimensions=[0, 1, 2], value=450)]
opt.set_constraints(constraints)

x = opt.ask(
    5,
    strategy="cl_min",
)

# ==============================
# Step 2: This code breaks
seed = 42

dimensions = [
    (10.0, 60.0), # <-- We have moved this dimension from last, to first
    (200.0, 450.0),
    (50.0, 450.0),
    (0.0, 450.0),
]
# Build optimizer
opt = Optimizer(
    dimensions=dimensions,
    lhs=False,
    acq_func="EI",
    n_initial_points=5,
    random_state=seed,
)

# Create the constraint, again just with the three factors with the 450 upper limit
constraints = [SumEquals(dimensions=[1, 2, 3], value=450)]
opt.set_constraints(constraints)

x = opt.ask(
    5,
    strategy="cl_min",
)

The code above throws an error in a later step, where the coordinates are transformed, with an error indicating that some points lie outside the defined space.

[dk-teknologisk-mon]

I have already identified and corrected the problem, working on a pull request now.