Hi everyone,

I have a docplex model where I create a set of variables using something similar to

m = Model(...)

# the initial set is not relevant I think
x = m.continuous_var_dict((k for k in ...), name="x")​

# get a solution
s = m.solve()

When I query s for the value of x, one entry is missing. I cannot share the list of keys as it's very large, but basically

# shows the correct keys, e.g., 10 keys
print(x.keys())

# some key in x.keys()
k = ...
assert k in x  # this is ok

# I can query the value using .get_value(), this returns a very small (negative?) number,
# e.g., -9.094947017729282e-13 (the value has a 0 lower-bound)
k_value = s.get_value(x[k])

# I can query the dictionary from x
x_values = s.get_value_dict(x)

# ...but the key is missing
assert k in x_values  # this fails

Any idea why this happens? I tried forcing keep_zeros to True in get_value_dict() (which should already be there since it's the default) but it did not change anything.

------------------------------
Mikaël Capelle
------------------------------

#DecisionOptimization

The method signature is

def get_value_dict(self, var_dict, keep_zeros=True, precision=1e-6)

I think its semantic is not obvious and trivial. The implementation is buggy here.
The keep_zeros works only on values which are really 0 or null to keep them or ignore them.
precision is a way to consider small values as 0 and filter them, but it does not look at keep_zeros.
That's the reason why your value is ignored.

Decrease the precision to 1e-13 and it will keep your variable.

We will correct docplex in a next version about this.

------------------------------
Vincent Beraudier
------------------------------