psamm.lpsolver.lp – Linear programming problems

Base objects for representation of LP problems.

A linear programming problem is built from a number of constraints and an objective function. The objective function is a linear expression represented by Expression. The constraints are represented by Relation, created from a linear expression and a relation sense (equals, greater, less).

Expressions are built from variables defined in the Problem instance. In addition, an expression can contain a VariableSet instead of a single variable. This allows many similar expressions to be represented by one Expression instance which means that the LP problem can be constructed faster.

class psamm.lpsolver.lp.Constraint

Represents a constraint within an LP Problem

delete()

Remove constraint from Problem instance

class psamm.lpsolver.lp.Expression(variables={}, offset=0)

Represents a linear expression

The variables can be any hashable objects. If one or more variables are instead VariableSets, this will be taken to represent a set of expressions separately using a different element of the VariableSet.

>>> e = Expression({'x': 2, 'y': 3})
>>> str(e)
'2*x + 3*y'

In order to provide a more natural syntax for creating Relations the binary relation operators have been overloaded to return Relation instances.

>>> rel = Expression({'x': 2}) >= Expression({'y': 3})
>>> str(rel)
'2*x - 3*y >= 0'

Warning

Chained relations cannot be converted to multiple relations, e.g. 4 <= e <= 10 will fail to produce the intended relations!

offset

Value of the offset

value_sets()

Iterator of expression sets

This will yield an iterator of (variable, value)-pairs for each expression in the expression set (each equivalent to values()). If none of the variables is a set variable then a single iterator will be yielded.

values()

Return immutable view of (variable, value)-pairs in expression.

variables()

Return immutable view of variables in expression.

exception psamm.lpsolver.lp.InvalidResultError(msg=None)

Raised when a result that has been invalidated is accessed

class psamm.lpsolver.lp.ObjectiveSense

Enumeration of objective sense values

class psamm.lpsolver.lp.Problem

Representation of LP Problem instance

Variable names in the problem can be any hashable object. It is the responsibility of the solver interface to translate the object into a unique string if required by the underlying LP solver.

add_linear_constraints(*relations)

Add constraints to the problem.

Each constraint is given as a Relation, and the expression in that relation can be a set expression. Returns a sequence of Constraints.

define(*names, **kwargs)

Define a variable in the problem.

Variables must be defined before they can be accessed by var() or set(). This function takes keyword arguments lower and upper to define the bounds of the variable (default: -inf to inf). The keyword argument types can be used to select the type of the variable (Continuous (default), Binary or Integer). Setting any variables different than Continuous will turn the problem into an MILP problem. Raises ValueError if a name is already defined.

expr(values, offset=0)

Return the given dictionary of values as an Expression.

has_variable(name)

Check whether a variable is defined in the problem.

namespace(names=None, **kwargs)

Return namespace for this problem.

If names is given it should be an iterable of names to define in the namespace. Other keyword arguments can be specified which will be used to define the names given as well as being used as default parameters for names that are defined later.

>>> v = prob.namespace(name='v')
>>> v.define([1, 2, 5], lower=0, upper=10)
>>> prob.set_objective(v[1] + 3*v[2] - 5 * v[5])
result

Result of solved problem

set(names)

Return variables as a set expression.

This returns an Expression containing a VariableSet.

set_linear_objective(expression)

Set objective of the problem to the given Expression.

set_objective(expression)

Set objective of the problem to the given Expression.

set_objective_sense(sense)

Set type of problem (minimize or maximize)

solve(sense=None)

Solve problem and return result.

Raises SolverError if the solution is not optimal.

solve_unchecked(sense=None)

Solve problem and return result.

The user must manually check the status of the result to determine whether an optimal solution was found. A SolverError may still be raised if the underlying solver raises an exception.

var(name)

Return variable as an Expression.

class psamm.lpsolver.lp.Product

A tuple used to represent a variable product.

class psamm.lpsolver.lp.RangedProperty(fget=None, fset=None, fdel=None, fmin=None, fmax=None, doc=None)

Numeric property with minimum and maximum values.

The value attribute is used to get/set the actual value of the propery. The min/max attributes are used to get the bounds. The range is not automatically enforced when the value is set.

class psamm.lpsolver.lp.Relation(sense, expression)

Represents a binary relation (equation or inequality)

Relations can be equalities or inequalities. All relations of this type can be represented as a left-hand side expression and the type of relation. In this representation, the right-hand side is always zero.

expression

Left-hand side expression

sense

Type of relation (equality or inequality)

Can be one of Equal, Greater or Less, or one of the strict relations, StrictlyGreater or StrictlyLess.

class psamm.lpsolver.lp.Result

Result of solving an LP problem

The result is tied to the solver instance and is valid at least until the problem is solved again. If the problem has been solved again an InvalidResultError may be raised.

get_value(expression)

Get value of variable or expression in result

Expression can be an object defined as a name in the problem, in which case the corresponding value is simply returned. If expression is an actual Expression object, it will be evaluated using the values from the result.

status

String indicating the status of the problem result

success

Whether solution was optimal

unbounded

Whether solution is unbounded

class psamm.lpsolver.lp.Solver

Factory for LP Problem instances

create_problem()

Create a new Problem instance

exception psamm.lpsolver.lp.SolverError

Error wrapping solver specific errors.

class psamm.lpsolver.lp.VariableNamespace(problem, **kwargs)

Namespace for defining variables.

Namespaces are always unique even if two namespaces have the same name. Variables defined within a namespace will never clash with a global variable name or a name defined in another namespace. Namespaces should be created using the Problem.namespace().

>>> v = prob.namespace(name='v')
>>> v.define([1, 2, 5], lower=0, upper=10)
>>> prob.set_objective(v[1] + 3*v[2] - 5 * v[5])
define(names, **kwargs)

Define variables within the namespace.

This is similar to Problem.define() except that names must be given as an iterable. This method accepts the same keyword arguments as Problem.define().

expr(items)

Return the sum of each name multiplied by a coefficient.

>>> v = prob.namespace(name='v')
>>> v.define(['a', 'b', 'c'], lower=0, upper=10)
>>> prob.set_objective(v.expr([('a', 2), ('b', 1)]))
set(names)

Return a variable set of the given names in the namespace.

>>> v = prob.namespace(name='v')
>>> v.define([1, 2, 5], lower=0, upper=10)
>>> prob.add_linear_constraints(v.set([1, 2]) >= 4)
sum(names)

Return the sum of the given names in the namespace.

>>> v = prob.namespace(name='v')
>>> v.define([1, 2, 5], lower=0, upper=10)
>>> prob.set_objective(v.sum([2, 5]))  # v[2] + v[5]
value(name)

Return value of given variable in namespace.

>>> v = prob.namespace(name='v')
>>> v.define([1, 2, 5], lower=0, upper=10)
>>> prob.solve()
>>> print(v.value(2))
class psamm.lpsolver.lp.VariableSet

A tuple used to represent sets of variables.

class psamm.lpsolver.lp.VariableType

Enumeration of variable types

psamm.lpsolver.lp.ranged_property(min=None, max=None)

Decorator for creating ranged property with fixed bounds.