def _get_primal(self):
if self.type == "continuous":
primal_from_solver = glp_get_col_prim(self.problem.problem, self.index)
elif self.type in ("binary", "integer"):
primal_from_solver = glp_mip_col_val(self.problem.problem, self.index)
else:
raise TypeError("Unknown variable type")
return primal_from_solver
def _get_primal(self):
if self.problem._glpk_is_mip():
primal_from_solver = glp_mip_col_val(self.problem.problem,
self._index)
else:
primal_from_solver = glp_get_col_prim(self.problem.problem,
self._index)
return primal_from_solver
def _get_primal(self):
if self.type == "continuous":
primal_from_solver = glp_get_col_prim(self.problem.problem,
self._index)
elif self.type in ("binary", "integer"):
primal_from_solver = glp_mip_col_val(self.problem.problem,
self._index)
else:
raise AssertionError("Unknown variable type")
return primal_from_solver
def get_clocks(problem: SwigPyObject) -> Iterable[Tuple[str, float]]:
n_recipes = 0
for j in range(1, 1 + lp.glp_get_num_cols(problem)):
clock = lp.glp_mip_col_val(problem, j)
if clock:
name = lp.glp_get_col_name(problem, j)
yield name, clock
n_recipes += 1
logger.info(f'{n_recipes} recipes in rate solution.')
def reduced_costs(self):
reduced_costs = collections.OrderedDict()
for index, variable in enumerate(self.variables):
if variable.type == "continuous":
value = glp_get_col_dual(self.problem, index + 1)
elif variable.type in ["binary", "integer"]:
value = glp_mip_col_val(self.problem, index + 1)
else:
raise AssertionError("Unknown variable type")
reduced_costs[variable.name] = value
return reduced_costs
def primal_values(self):
primal_values = collections.OrderedDict()
is_mip = self._glpk_is_mip()
for index, variable in enumerate(self.variables):
if is_mip:
value = glp_mip_col_val(self.problem, index + 1)
else:
value = glp_get_col_prim(self.problem, index + 1)
primal_values[variable.name] = variable._round_primal_to_bounds(
value)
return primal_values
def reduced_costs(self):
reduced_costs = collections.OrderedDict()
for index, variable in enumerate(self.variables):
if variable.type == "continuous":
value = glp_get_col_dual(self.problem, index + 1)
elif variable.type in ["binary", "integer"]:
value = glp_mip_col_val(self.problem, index + 1)
else:
raise AssertionError("Unknown variable type")
reduced_costs[variable.name] = value
return reduced_costs
def primal_values(self):
primal_values = collections.OrderedDict()
for index, variable in enumerate(self.variables):
if variable.type == "continuous":
value = glp_get_col_prim(self.problem, index + 1)
elif variable.type in ["binary", "integer"]:
value = glp_mip_col_val(self.problem, index + 1)
else:
raise AssertionError("Unknown variable type")
primal_values[variable.name] = variable._round_primal_to_bounds(value)
return primal_values
def primal_values(self):
primal_values = collections.OrderedDict()
for index, variable in enumerate(self.variables):
if variable.type == "continuous":
value = glp_get_col_prim(self.problem, index + 1)
elif variable.type in ["binary", "integer"]:
value = glp_mip_col_val(self.problem, index + 1)
else:
raise AssertionError("Unknown variable type")
primal_values[variable.name] = variable._round_primal_to_bounds(
value)
return primal_values
def _solve(self):
import swiglpk as glpk
# An alias to the internal problem instance.
p = self.int
continuous = self.ext.is_continuous()
# Select LP solver (Simplex or Interior Point Method).
if continuous:
if self.ext.options["lp_root_method"] == "interior":
interior = True
else:
# Default to Simplex.
interior = False
simplex = not interior
else:
simplex = interior = False
# Select appropriate options container.
if simplex:
options = glpk.glp_smcp()
glpk.glp_init_smcp(options)
elif interior:
options = glpk.glp_iptcp()
glpk.glp_init_iptcp(options)
else:
options = glpk.glp_iocp()
glpk.glp_init_iocp(options)
# Handle "verbose" option.
verbosity = self.verbosity()
if verbosity < 0:
options.msg_lev = glpk.GLP_MSG_OFF
elif verbosity == 0:
options.msg_lev = glpk.GLP_MSG_ERR
elif verbosity == 1:
options.msg_lev = glpk.GLP_MSG_ON
elif verbosity >= 2:
options.msg_lev = glpk.GLP_MSG_ALL
# Handle "tol" option.
# Note that GLPK knows three different tolerances for Simplex but none
# for the Interior Point Method, while PICOS states that "tol" is meant
# only for the IPM.
# XXX: The option is unsupported but does not default to None, so we
# cannot warn the user.
pass
# Handle "maxit" option.
if not simplex:
self._handle_unsupported_option("maxit",
"GLPK supports the 'maxit' option only with Simplex.")
elif self.ext.options["maxit"] is not None:
options.it_lim = int(self.ext.options["maxit"])
# Handle "lp_root_method" option.
# Note that the PICOS option is explicitly also meant for the MIP
# preprocessing step but GLPK does not support it in that scenario.
if not continuous:
self._handle_unsupported_option("lp_root_method",
"GLPK supports the 'lp_root_method' option only for LPs.")
elif self.ext.options["lp_root_method"] is not None:
if self.ext.options["lp_root_method"] == "interior":
# Handled above.
pass
elif self.ext.options["lp_root_method"] == "psimplex":
assert simplex
options.meth = glpk.GLP_PRIMAL
elif self.ext.options["lp_root_method"] == "dsimplex":
assert simplex
options.meth = glpk.GLP_DUAL
else:
self._handle_bad_option_value("lp_root_method")
# Handle "timelimit" option.
if interior:
self._handle_unsupported_option("timelimit",
"GLPK does not support the 'timelimit' option with the "
"Interior Point Method.")
elif self.ext.options["timelimit"] is not None:
options.tm_lim = 1000 * int(self.ext.options["timelimit"])
# Handle "gaplim" option.
# Note that the option is silently ignored if passed alongside an LP;
# while the solver does not allow us to pass the option in that case, it
# is still technically a valid option as every LP is also a MIP.
# TODO: Find out if "mip_gap" is really equivalent to "gaplim".
if self.ext.options["gaplim"] is not None:
if not continuous:
options.mip_gap = float(self.ext.options["gaplim"])
# Handle unsupported options.
self._handle_unsupported_options(
"lp_node_method", "treememory", "nbsol", "hotstart")
# TODO: Add GLPK-sepcific options. Candidates are:
# For both Simplex and MIPs:
# tol_*, out_*
# For Simplex:
# pricing, r_test, obj_*
# For the Interior Point Method:
# ord_alg
# For MIPs:
# *_tech, *_heur, ps_tm_lim, *_cuts, cb_size, binarize
# Attempt to solve the problem.
with self._header():
with self._stopwatch():
if simplex:
# TODO: Support glp_exact.
error = glpk.glp_simplex(p, options)
elif interior:
error = glpk.glp_interior(p, options)
else:
options.presolve = glpk.GLP_ON
error = glpk.glp_intopt(p, options)
# Conert error codes to text output.
# Note that by printing it above the footer, this output is made to
# look like it's coming from GLPK, which is technically wrong but
# semantically correct.
if error == glpk.GLP_EBADB:
self._warn("Unable to start the search, because the initial "
"basis specified in the problem object is invalid.")
elif error == glpk.GLP_ESING:
self._warn("Unable to start the search, because the basis "
"matrix corresponding to the initial basis is singular "
"within the working precision.")
elif error == glpk.GLP_ECOND:
self._warn("Unable to start the search, because the basis "
"matrix corresponding to the initial basis is "
"ill-conditioned.")
elif error == glpk.GLP_EBOUND:
self._warn("Unable to start the search, because some double-"
"bounded variables have incorrect bounds.")
elif error == glpk.GLP_EFAIL:
self._warn("The search was prematurely terminated due to a "
"solver failure.")
elif error == glpk.GLP_EOBJLL:
self._warn("The search was prematurely terminated, because the "
"objective function being maximized has reached its lower "
"limit and continues decreasing.")
elif error == glpk.GLP_EOBJUL:
self._warn("The search was prematurely terminated, because the "
"objective function being minimized has reached its upper "
"limit and continues increasing.")
elif error == glpk.GLP_EITLIM:
self._warn("The search was prematurely terminated, because the "
"simplex iteration limit has been exceeded.")
elif error == glpk.GLP_ETMLIM:
self._warn("The search was prematurely terminated, because the "
"time limit has been exceeded.")
elif error == glpk.GLP_ENOPFS:
self._verbose("The LP has no primal feasible solution.")
elif error == glpk.GLP_ENODFS:
self._verbose("The LP has no dual feasible solution.")
elif error != 0:
self._warn("GLPK error {:d}.".format(error))
# Retrieve primals.
if self.ext.options["noprimals"]:
primals = None
else:
primals = {}
for variable in self.ext.variables.values():
value = []
for localIndex, picosIndex \
in enumerate(range(variable.startIndex, variable.endIndex)):
glpkIndex = self._picos2glpk_variable_index(picosIndex)
if simplex:
localValue = glpk.glp_get_col_prim(p, glpkIndex);
elif interior:
localValue = glpk.glp_ipt_col_prim(p, glpkIndex);
else:
localValue = glpk.glp_mip_col_val(p, glpkIndex);
value.append(localValue)
primals[variable.name] = value
# Retrieve duals.
# XXX: Returns the duals as a flat cvx.matrix to be consistent with
# other solvers. This feels incorrect when the constraint was given
# as a proper two dimensional matrix.
if self.ext.options["noduals"] or not continuous:
duals = None
else:
duals = []
rowOffset = 1
for constraintNum, constraint in enumerate(self.ext.constraints):
assert isinstance(constraint, AffineConstraint)
numRows = len(constraint)
values = []
for localConIndex in range(numRows):
glpkConIndex = rowOffset + localConIndex
if simplex:
localValue = glpk.glp_get_row_dual(p, glpkConIndex);
elif interior:
localValue = glpk.glp_ipt_row_dual(p, glpkConIndex);
else:
assert False
values.append(localValue)
if constraint.is_decreasing():
duals.append(-cvxopt.matrix(values))
else:
duals.append(cvxopt.matrix(values))
rowOffset += numRows
if glpk.glp_get_obj_dir(p) == glpk.GLP_MIN:
duals = [-d for d in duals]
# Retrieve objective value.
if simplex:
objectiveValue = glpk.glp_get_obj_val(p)
elif interior:
objectiveValue = glpk.glp_ipt_obj_val(p)
else:
objectiveValue = glpk.glp_mip_obj_val(p)
# Retrieve solution metadata.
meta = {}
if simplex:
# Set common entry "status".
status = glpk.glp_get_status(p)
if status is glpk.GLP_OPT:
meta["status"] = "optimal"
elif status is glpk.GLP_FEAS:
meta["status"] = "feasible"
elif status in (glpk.GLP_INFEAS, glpk.GLP_NOFEAS):
meta["status"] = "infeasible"
elif status is glpk.GLP_UNBND:
meta["status"] = "unbounded"
elif status is glpk.GLP_UNDEF:
meta["status"] = "undefined"
else:
meta["status"] = "unknown"
# Set GLPK-specific entry "primal_status".
primalStatus = glpk.glp_get_prim_stat(p)
if primalStatus is glpk.GLP_FEAS:
meta["primal_status"] = "feasible"
elif primalStatus in (glpk.GLP_INFEAS, glpk.GLP_NOFEAS):
meta["primal_status"] = "infeasible"
elif primalStatus is glpk.GLP_UNDEF:
meta["primal_status"] = "undefined"
else:
meta["primal_status"] = "unknown"
# Set GLPK-specific entry "dual_status".
dualStatus = glpk.glp_get_dual_stat(p)
if dualStatus is glpk.GLP_FEAS:
meta["dual_status"] = "feasible"
elif dualStatus in (glpk.GLP_INFEAS, glpk.GLP_NOFEAS):
meta["dual_status"] = "infeasible"
elif dualStatus is glpk.GLP_UNDEF:
meta["dual_status"] = "undefined"
else:
meta["dual_status"] = "unknown"
elif interior:
# Set common entry "status".
status = glpk.glp_ipt_status(p)
if status is glpk.GLP_OPT:
meta["status"] = "optimal"
elif status in (glpk.GLP_INFEAS, glpk.GLP_NOFEAS):
meta["status"] = "infeasible"
elif status is glpk.GLP_UNDEF:
meta["status"] = "undefined"
else:
meta["status"] = "unknown"
else:
# Set common entry "status".
status = glpk.glp_mip_status(p)
if status is glpk.GLP_OPT:
meta["status"] = "optimal"
elif status is glpk.GLP_FEAS:
meta["status"] = "feasible"
elif status is glpk.GLP_NOFEAS:
meta["status"] = "infeasible"
elif status is glpk.GLP_UNDEF:
meta["status"] = "undefined"
else:
meta["status"] = "unknown"
return (primals, duals, objectiveValue, meta)
def solve(nutrition_target, foods):
'''
Calculate food amounts to reach the nutrition target
Parameters
----------
nutrition_target : soylent_recipes.nutrition_target.NormalizedNutritionTarget
The desired nutrition
foods : np.array
The foods to use to achieve the nutrition target. Contains exactly the
nutrients required by the nutrition target in the exact same order. Rows
represent foods, columns represent nutrients.
Returns
-------
amounts : np.array(int) or None
The amounts of each food to use to optimally achieve the nutrition
target. ``amounts[i]`` is the amount of the i-th food to use. If the
nutrition target cannot be achieved, returns None.
'''
# Implementation: using the GLPK C library via ecyglpki Python library binding
# GLPK documentation: download it and look inside the package (http://ftp.gnu.org/gnu/glpk/)
# GLPK wikibook: https://en.wikibooks.org/wiki/GLPK
#
# GPLK lingo: rows and columns refer to Ax=b where b_i are auxiliary
# variables, x_i are structural variables. Setting constraints on rows, set
# constraints on b_i, while column constraints are applied to x_i.
# Note: glpk is powerful. We're using mostly the default settings.
# Performance likely can be improved by tinkering with the settings; or even
# by providing the solution to the least squares equivalent, with amounts
# rounded afterwards, as starting point could improve performance.
nutrition_target = nutrition_target.values
problem = glp.glp_create_prob()
try:
glp.glp_add_rows(problem, len(nutrition_target))
glp.glp_add_cols(problem, len(foods))
# Configure columns/amounts
for i in range(len(foods)):
glp.glp_set_col_kind(problem, i+1, glp.GLP_IV) # int
glp.glp_set_col_bnds(problem, i+1, glp.GLP_LO, 0.0, np.nan) # >=0
# Configure rows/nutrients
for i, extrema in enumerate(nutrition_target):
if np.isnan(extrema[0]):
bounds_type = glp.GLP_UP
elif np.isnan(extrema[1]):
bounds_type = glp.GLP_LO
else:
# Note: a nutrition target has either min, max or both and min!=max
bounds_type = glp.GLP_DB
glp.glp_set_row_bnds(problem, i+1, bounds_type, *extrema)
# Load A of our Ax=b
non_zero_count = foods.size
row_indices = glp.intArray(non_zero_count+1) # +1 because (insane) 1-indexing
column_indices = glp.intArray(non_zero_count+1)
values = glp.doubleArray(non_zero_count+1)
for i, ((row, column), value) in enumerate(np.ndenumerate(foods.transpose())):
row_indices[i+1] = row+1
column_indices[i+1] = column+1
values[i+1] = value
glp.glp_load_matrix(problem, non_zero_count, row_indices, column_indices, values)
# Solve
int_opt_args = glp.glp_iocp()
glp.glp_init_iocp(int_opt_args)
int_opt_args.presolve = glp.GLP_ON # without this, you have to provide an LP relaxation basis
int_opt_args.msg_lev = glp.GLP_MSG_OFF # be quiet, no stdout
glp.glp_intopt(problem, int_opt_args) # returns an error code; can safely ignore
# Check we've got a valid solution
#
# Note: glp_intopt returns whether the algorithm completed successfully.
# This does not imply you've got a good solution, it could even be
# infeasible. glp_mip_status returns whether the solution is optimal,
# feasible, infeasible or undefined. An optimal/feasible solution is not
# necessarily a good solution. An optimal solution may even violate
# bounds constraints. The thing you actually need to use is
# glp_check_kkt and check that the solution satisfies KKT.PB (all within
# bounds)
max_error = glp.doubleArray(1)
glp.glp_check_kkt(problem, glp.GLP_MIP, glp.GLP_KKT_PB, max_error, None, None, None)
if not np.isclose(max_error[0], 0.0):
# A row/column value exceeds its bounds
return None
# Return solution
amounts = np.fromiter((glp.glp_mip_col_val(problem, i+1) for i in range(len(foods))), int)
return amounts
finally:
glp.glp_delete_prob(problem)
def _get_var_value(self, variable):
self._check_valid()
if variable not in self._problem._variables:
raise ValueError('Unknown variable: {}'.format(variable))
return swiglpk.glp_mip_col_val(self._problem._p,
self._problem._variables[variable])
def _get_var_value(self, variable):
self._check_valid()
if variable not in self._problem._variables:
raise ValueError('Unknown variable: {}'.format(variable))
return swiglpk.glp_mip_col_val(
self._problem._p, self._problem._variables[variable])
def _get_primal(self):
if self.problem._glpk_is_mip():
primal_from_solver = glp_mip_col_val(self.problem.problem, self._index)
else:
primal_from_solver = glp_get_col_prim(self.problem.problem, self._index)
return primal_from_solver
def _get_value(self, variable):
"""Return value of variable in solution."""
return swiglpk.glp_mip_col_val(self._problem._p,
self._problem._variables[variable])
