sage.numerical.linear_functions.LinearConstraint is not a hashable type

[sagetrac-cbowen]

I'd like to have sage.numerical.linear_functions.LinearConstraint to be a hashable type.
Currently it is not:

sage: p = MixedIntegerLinearProgram()
sage: b = p.new_variable()
sage: b[0] <= b[1] <= 2
x_0 <= x_1 <= 2
sage: hash(b[0] <= b[1] <= 2)
...
TypeError: unhashable type: 'sage.numerical.linear_functions.LinearConstraint'

This is for the following reasons:

1. When adding constraints to an instance of a MixedIntegerLinearProgram, it is not necessary to add the same or equivalent constraint multiple times. A hash could be used to determine distinct linear constraints that are added into a MILP instance.
2. By having hashable linear constraints, one could allow for linear constraints to be added into sets (http://docs.python.org/2/library/sets.html) or used as keys in dictionaries.

Thanks,
Clinton

CC: @nathanncohen @vbraun @dimpase

Component: linear programming

Keywords: linear constraint, mixed integer linear program

Reviewer: Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/15583

[mkoeppe]

comment:5

Added example to the description.

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1,9 +1,19 @@
-Hello,
+I'd like to have `sage.numerical.linear_functions.LinearConstraint` to be a hashable type.
+Currently it is not:
 
-I'd like to have sage.numerical.linear_functions.LinearConstraint to be a hashable type for the following reasons:
+```
+sage: p = MixedIntegerLinearProgram()
+sage: b = p.new_variable()
+sage: b[0] <= b[1] <= 2
+x_0 <= x_1 <= 2
+sage: hash(b[0] <= b[1] <= 2)
+...
+TypeError: unhashable type: 'sage.numerical.linear_functions.LinearConstraint'
+```
 
+This is for the following reasons:
 
-1)  When adding constraints to an instance of a MixedIntegerLinearProgram, it is not necessary to add the same or equivalent constraint multiple times.  A hash could be used to determine distinct linear constraints that are added into a MILP instance.  
+1)  When adding constraints to an instance of a `MixedIntegerLinearProgram`, it is not necessary to add the same or equivalent constraint multiple times.  A hash could be used to determine distinct linear constraints that are added into a MILP instance.  
 2) By having hashable linear constraints, one could allow for linear constraints to be added into sets (http://docs.python.org/2/library/sets.html) or used as keys in dictionaries.
 
 

[sagetrac-brown]

comment:7

sage.numerical.linear_functions.LinearConstraint is now a hashable type.

sage: p = MixedIntegerLinearProgram()
sage: b = p.new_variable()
sage: c = (b[0] <= b[1] <= 2)
sage: c.__hash__()
8748240900236
sage: c.__hash__
<method-wrapper '__hash__' of sage.numerical.linear_functions.LinearConstraint object at 0x7f4dbca908c0>

I think we have https://github.com/sagemath/sagetrac-mirror/commits/682ba3e7b8aaf84ba181ac16b8630aefb695c940 to thank for the update.