Allow using ternary operator inside @constraint

[hdavid16]

It would be great to do something like this:

m=Model()
@variable(m,x)
@constraint(m, [i=1:2],
    i == 1 ? exp(x) <= 0 : x^2 <= 0
)

Currently, this gives the error:

Unsupported constraint expression: we don't know how to parse constraints containing expressions of type :if.

This is, however, supported in @expression (the macro below does work):

@constraint(m, [i=1:2],
    i == 1 ? exp(x) : x^2
)

Tested in JuMP v.1.6.0

Current workaround is to use the ternary operator in @expression and then use that expression to build the needed constraints.

[odow]

Multiple problems here:

• @constraint doesn't support exp
• Second example doesn't form a constraint

This is, however, supported in @expression
Current workaround is to use the ternary operator in @expression and then use that expression to build the needed constraints.

Not sure what you mean? This shouldn't work in @expression because it doesn't support exp.

The correct way to write this is:

julia> model = Model()
A JuMP Model
Feasibility problem with:
Variables: 0
Model mode: AUTOMATIC
CachingOptimizer state: NO_OPTIMIZER
Solver name: No optimizer attached.

julia> @variable(model, x)
x

julia> @NLconstraint(model, [i=1:2], ifelse(i == 1, exp(x), x^2) <= 0)
2-element Vector{NonlinearConstraintRef{ScalarShape}}:
 ifelse(1.0 == 1.0, exp(x), x ^ 2.0) - 0.0 ≤ 0
 ifelse(2.0 == 1.0, exp(x), x ^ 2.0) - 0.0 ≤ 0

[hdavid16]

Yes, you are right, exp is nonlinear. Instead of exp(x) <= 0, say x <= 0 instead.

Ok, I see that I can actually use the ternary operator if I wrap it in parenthesis. For example,

@constraint(m,[i=1:3], x <= (i == 3 ? 0 : 2))

The above works as long as the ternary operator part is in parenthesis. Otherwise, it errors.

Thanks for looking into this.