Duplicate Variables Bandaid

README.md

@@ -16,7 +16,7 @@ Some of the main features of MIP are:
   [MathProg](https://en.wikibooks.org/wiki/GLPK/GMPL_(MathProg)): 
   operator overloading makes it easy to write linear expressions in Python;
 
-* full featured:
+* full featured: 
     - cut generators and lazy constraints: work with strong formulations with a
     large number of constraints by generating only the required inequalities
     during the branch and cut search;

mip/entities.py

@@ -65,6 +65,9 @@ class LinExpr:
      a = 10*x1 + 7*x4
      print(a.x)
 
+    .. warning::
+    Do not pass identical objects in the ``variables`` argument when constructing
+    a LinExpr manually.   
     """
 
     __slots__ = ["__const", "__expr", "__sense"]
@@ -542,6 +545,8 @@ def __add__(
         self, other: Union["mip.Var", LinExpr, numbers.Real]
     ) -> Union["mip.Var", LinExpr]:
         if isinstance(other, Var):
+            if id(self) == id(other):
+                return LinExpr([self], [2])
             return LinExpr([self, other], [1, 1])
         if isinstance(other, LinExpr):
             return other.__add__(self)
@@ -561,6 +566,8 @@ def __sub__(
         self, other: Union["mip.Var", LinExpr, numbers.Real]
     ) -> Union["mip.Var", LinExpr]:
         if isinstance(other, Var):
+            if id(self) == id(other):
+                return LinExpr([self], [0])
             return LinExpr([self, other], [1, -1])
         if isinstance(other, LinExpr):
             return (-other).__add__(self)
@@ -575,6 +582,8 @@ def __rsub__(
         self, other: Union["mip.Var", LinExpr, numbers.Real]
     ) -> Union["mip.Var", LinExpr]:
         if isinstance(other, Var):
+            if id(self) == id(other):
+                return LinExpr([self], [0])
             return LinExpr([self, other], [-1, 1])
         if isinstance(other, LinExpr):
             return other.__sub__(self)
@@ -603,6 +612,8 @@ def __neg__(self) -> LinExpr:
 
     def __eq__(self, other) -> LinExpr:
         if isinstance(other, Var):
+            if id(self) == id(other):
+                return LinExpr([self], [0], sense="=")
             return LinExpr([self, other], [1, -1], sense="=")
         if isinstance(other, LinExpr):
             return LinExpr([self], [1]) == other
@@ -613,6 +624,8 @@ def __eq__(self, other) -> LinExpr:
 
     def __le__(self, other: Union["mip.Var", LinExpr, numbers.Real]) -> LinExpr:
         if isinstance(other, Var):
+            if id(self) == id(other):
+                return LinExpr([self], [0], sense="<")
             return LinExpr([self, other], [1, -1], sense="<")
         if isinstance(other, LinExpr):
             return LinExpr([self], [1]) <= other
@@ -623,6 +636,8 @@ def __le__(self, other: Union["mip.Var", LinExpr, numbers.Real]) -> LinExpr:
 
     def __ge__(self, other: Union["mip.Var", LinExpr, numbers.Real]) -> LinExpr:
         if isinstance(other, Var):
+            if id(self) == id(other):
+                return LinExpr([self], [0], sense=">")
             return LinExpr([self, other], [1, -1], sense=">")
         if isinstance(other, LinExpr):
             return LinExpr([self], [1]) >= other

test/mip_test.py

@@ -4,6 +4,7 @@
 from os import environ
 
 import networkx as nx
+from mip.entities import LinExpr
 import mip.gurobi
 import mip.highs
 from mip import Model, xsum, OptimizationStatus, MAXIMIZE, BINARY, INTEGER
@@ -578,6 +579,28 @@ def test_obj_const2(self, solver: str):
         assert model.objective_const == 1
 
 
+@skip_on(NotImplementedError)
+@pytest.mark.parametrize("solver", SOLVERS)
+@pytest.mark.parametrize("constraint, lb, ub", [
+    (lambda x: x + x >= 3, 1, 2),
+    (lambda x: x - x >= 0, 1, 2),
+    (lambda x: x == x, 1, 2),
+    (lambda x: x >= x, 1, 2),
+    (lambda x: x <= x, -2, -1),
+    # (lambda x: LinExpr([x, x, x], [2, -1, -1], sense="="), 1, 2),
+])
+def test_identical_vars(solver: str, constraint, lb, ub):
+    """Try if constraints are correctly added when variables are identical"""
+    m = Model(solver_name=solver)
+    x = m.add_var(name="x", lb=lb, ub=ub, obj=1)
+
+    m.add_constr(constraint(x))
+
+    m.optimize()
+    assert m.status == OptimizationStatus.OPTIMAL
+    assert lb - TOL <= x.x <= ub + TOL
+
+
 @skip_on(NotImplementedError)
 @pytest.mark.parametrize("val", range(1, 4))
 @pytest.mark.parametrize("solver", SOLVERS)