linopt.py
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Tools for fast Linear Problem file writing. This module contains.
- io functions for writing out variables, constraints and objective
into a lp file.
- functions to create lp format based linear expression
- solver functions which read the lp file, run the problem and return the
solution
This module supports the linear optimal power flow calculation without using
pyomo (see module linopt.py)
"""
__author__ = (
"PyPSA Developers, see https://pypsa.readthedocs.io/en/latest/developers.html"
)
__copyright__ = (
"Copyright 2015-2023 PyPSA Developers, see https://pypsa.readthedocs.io/en/latest/developers.html, "
"MIT License"
)
import io
import logging
import os
import re
import subprocess
from importlib.util import find_spec
import numpy as np
import pandas as pd
from packaging.version import Version, parse
from pandas import IndexSlice as idx
from pypsa.descriptors import Dict
logger = logging.getLogger(__name__)
# =============================================================================
# Front end functions
# =============================================================================
def define_variables(n, lower, upper, name, attr="", axes=None, spec="", mask=None):
"""
Defines variable(s) for pypsa-network with given lower bound(s) and upper
bound(s). The variables are stored in the network object under n.vars with
key of the variable name. If multiple variables are defined at ones, at
least one of lower and upper has to be an array (including pandas) of
`shape > (1, )` or axes have to define the dimensions of the variables.
Parameters
----------
n : pypsa.Network
lower : pd.Series/pd.DataFrame/np.array/str/float
lower bound(s) for the variable(s)
upper : pd.Series/pd.DataFrame/np.array/str/float
upper bound(s) for the variable(s)
name : str
general name of the variable (or component which the variable is
referring to). The variable will then be stored under:
* n.vars[name].pnl if the variable is two-dimensional
* n.vars[name].df if the variable is one-dimensional
but can easily be accessed with :func:`get_var(n, name, attr)`
attr : str default ''
Specifying name of the variable, defines under which name the variable(s)
are stored in n.vars[name].pnl if two-dimensional or in n.vars[name].df
if one-dimensional
axes : pd.Index or tuple of pd.Index objects, default None
Specifies the axes and therefore the shape of the variables if bounds
are single strings or floats. This is helpful when multiple variables
have the same upper and lower bound.
mask: pd.DataFrame/np.array
Boolean mask with False values for variables which are skipped.
The shape of the mask has to match the shape the added variables.
Example
--------
Let's say we want to define a demand-side-managed load at each bus of
network n, which has a minimum of 0 and a maximum of 10. We then define
lower bound (lb) and upper bound (ub) and pass it to define_variables
>>> from pypsa.linopt import define_variables, get_var
>>> lb = pd.DataFrame(0, index=n.snapshots, columns=n.buses.index)
>>> ub = pd.DataFrame(10, index=n.snapshots, columns=n.buses.index)
>>> define_variables(n, lb, ub, 'DSM', 'variableload')
Now the variables can be accessed by :func:`pypsa.linopt.get_var` using
>>> variables = get_var(n, 'DSM', 'variableload')
Note that this is usefull for the `extra_functionality` argument.
"""
var = write_bound(n, lower, upper, axes, mask)
set_varref(n, var, name, attr, spec=spec)
return var
def define_binaries(n, axes, name, attr="", spec="", mask=None):
"""
Defines binary-variable(s) for pypsa-network. The variables are stored in
the network object under n.vars with key of the variable name. For each
entry for the pd.Series of pd.DataFrame spanned by the axes argument the
function defines a binary.
Parameters
----------
n : pypsa.Network
axes : pd.Index or tuple of pd.Index objects
Specifies the axes and therefore the shape of the variables.
name : str
general name of the variable (or component which the variable is
referring to). The variable will then be stored under:
* n.vars[name].pnl if the variable is two-dimensional
* n.vars[name].df if the variable is one-dimensional
attr : str default ''
Specifying name of the variable, defines under which name the variable(s)
are stored in n.vars[name].pnl if two-dimensional or in n.vars[name].df
if one-dimensional
mask: pd.DataFrame/np.array
Boolean mask with False values for variables which are skipped.
The shape of the mask has to match the shape given by axes.
See also
---------
define_variables
"""
var = write_binary(n, axes)
set_varref(n, var, name, attr, spec=spec)
return var
def define_constraints(
n, lhs, sense, rhs, name, attr="", axes=None, spec="", mask=None
):
"""
Defines constraint(s) for pypsa-network with given left hand side (lhs),
sense and right hand side (rhs). The constraints are stored in the network
object under n.cons with key of the constraint name. If multiple
constraints are defined at ones, only using np.arrays, then the axes
argument can be used for defining the axes for the constraints (this is
especially recommended for time-dependent constraints). If one of lhs,
sense and rhs is a pd.Series/pd.DataFrame the axes argument is not
necessary.
Parameters
----------
n: pypsa.Network
lhs: pd.Series/pd.DataFrame/np.array/str/float
left hand side of the constraint(s), created with
:func:`pypsa.linot.linexpr`.
sense: pd.Series/pd.DataFrame/np.array/str/float
sense(s) of the constraint(s)
rhs: pd.Series/pd.DataFrame/np.array/str/float
right hand side of the constraint(s), must only contain pure constants,
no variables
name: str
general name of the constraint (or component which the constraint is
referring to). The constraint will then be stored under:
* n.cons[name].pnl if the constraint is two-dimensional
* n.cons[name].df if the constraint is one-dimensional
attr: str default ''
Specifying name of the constraint, defines under which name the
constraint(s) are stored in n.cons[name].pnl if two-dimensional or in
n.cons[name].df if one-dimensional
axes: pd.Index or tuple of pd.Index objects, default None
Specifies the axes if all of lhs, sense and rhs are np.arrays or single
strings or floats.
mask: pd.DataFrame/np.array
Boolean mask with False values for constraints which are skipped.
The shape of the mask has to match the shape of the array that come out
when combining lhs, sense and rhs.
Example
--------
Let's say we want to constraint all gas generators to a maximum of 100 MWh
during the first 10 snapshots. We then firstly get all operational variables
for this subset and constraint there sum to less equal 100.
>>> from pypsa.linopt import get_var, linexpr, define_constraints
>>> gas_i = n.generators.query('carrier == "Natural Gas"').index
>>> gas_vars = get_var(n, 'Generator', 'p').loc[n.snapshots[:10], gas_i]
>>> lhs = linexpr((1, gas_vars)).sum().sum()
>>> define_constraints(n, lhs, '<=', 100, 'Generator', 'gas_power_limit')
Now the constraint references can be accessed by
:func:`pypsa.linopt.get_con` using
>>> cons = get_var(n, 'Generator', 'gas_power_limit')
Under the hood they are stored in n.cons.Generator.pnl.gas_power_limit.
For retrieving their shadow prices add the general name of the constraint
to the keep_shadowprices argument.
Note that this is useful for the `extra_functionality` argument.
"""
con = write_constraint(n, lhs, sense, rhs, axes, mask)
set_conref(n, con, name, attr, spec=spec)
return con
# =============================================================================
# writing functions
# =============================================================================
def _get_handlers(axes, *maybearrays):
axes = [axes] if isinstance(axes, pd.Index) else axes
if axes is None:
axes, shape = broadcasted_axes(*maybearrays)
else:
shape = tuple(map(len, axes))
size = np.prod(shape)
return axes, shape, size
def write_bound(n, lower, upper, axes=None, mask=None):
"""
Writer function for writing out multiple variables at a time.
If lower and upper are floats it demands to give pass axes, a tuple
of (index, columns) or (index), for creating the variable of same
upper and lower bounds. Return a series or frame with variable
references.
"""
axes, shape, size = _get_handlers(axes, lower, upper)
if not size:
return pd.Series(dtype=float)
n._xCounter += size
variables = np.arange(n._xCounter - size, n._xCounter).reshape(shape)
lower, upper = _str_array(lower), _str_array(upper)
exprs = lower + " <= x" + _str_array(variables, True) + " <= " + upper + "\n"
if mask is not None:
exprs = np.where(mask, exprs, "")
variables = np.where(mask, variables, -1)
n.bounds_f.write(join_exprs(exprs))
return to_pandas(variables, *axes)
def write_constraint(n, lhs, sense, rhs, axes=None, mask=None):
"""
Writer function for writing out multiple constraints to the corresponding
constraints file.
If lower and upper are numpy.ndarrays it axes must not be None but a
tuple of (index, columns) or (index). Return a series or frame with
constraint references.
"""
axes, shape, size = _get_handlers(axes, lhs, sense, rhs)
if not size:
return pd.Series()
n._cCounter += size
cons = np.arange(n._cCounter - size, n._cCounter).reshape(shape)
if isinstance(sense, str):
sense = "=" if sense == "==" else sense
lhs, sense, rhs = _str_array(lhs), _str_array(sense), _str_array(rhs)
exprs = "c" + _str_array(cons, True) + ":\n" + lhs + sense + " " + rhs + "\n\n"
if mask is not None:
exprs = np.where(mask, exprs, "")
cons = np.where(mask, cons, -1)
n.constraints_f.write(join_exprs(exprs))
return to_pandas(cons, *axes)
def write_binary(n, axes, mask=None):
"""
Writer function for writing out multiple binary-variables at a time.
According to the axes it writes out binaries for each entry the
pd.Series or pd.DataFrame spanned by axes. Returns a series or frame
with variable references.
"""
axes, shape, size = _get_handlers(axes)
n._xCounter += size
variables = np.arange(n._xCounter - size, n._xCounter).reshape(shape)
exprs = "x" + _str_array(variables, True) + "\n"
if mask is not None:
exprs = np.where(mask, exprs, "")
variables = np.where(mask, variables, -1)
n.binaries_f.write(join_exprs(exprs))
return to_pandas(variables, *axes)
def write_objective(n, terms):
"""
Writer function for writing out one or multiple objective terms.
Parameters
----------
n : pypsa.Network
terms : str/numpy.array/pandas.Series/pandas.DataFrame
String or array of strings which represent new objective terms, built
with :func:`linexpr`
"""
n.objective_f.write(join_exprs(terms))
# =============================================================================
# helpers, helper functions
# =============================================================================
def broadcasted_axes(*dfs):
"""
Helper function which, from a collection of arrays, series, frames and
other values, retrieves the axes of series and frames which result from
broadcasting operations.
It checks whether index and columns of given series and frames,
repespectively, are aligned. Using this function allows to
subsequently use pure numpy operations and keep the axes in the
background.
"""
axes = []
shape = (1,)
if set(map(type, dfs)) == {tuple}:
dfs = sum(dfs, ())
for df in dfs:
shape = np.broadcast_shapes(shape, np.asarray(df).shape)
if isinstance(df, (pd.Series, pd.DataFrame)):
if len(axes):
assert (axes[-1] == df.axes[-1]).all(), (
"Series or DataFrames "
"are not aligned. Please make sure that all indexes and "
"columns of Series and DataFrames going into the linear "
"expression are equally sorted."
)
axes = df.axes if len(df.axes) > len(axes) else axes
return axes, shape
def align_with_static_component(n, c, attr):
"""
Alignment of time-dependent variables with static components.
If c is a pypsa.component name, it will sort the columns of the
variable according to the static component.
"""
if c in n.all_components and (c, attr) in n.variables.index:
if not n.variables.pnl[c, attr]:
return
if len(n.vars[c].pnl[attr].columns) != len(n.df(c).index):
return
n.vars[c].pnl[attr] = n.vars[c].pnl[attr].reindex(columns=n.df(c).index)
def linexpr(*tuples, as_pandas=True, return_axes=False):
"""
Elementwise concatenation of tuples in the form (coefficient, variables).
Coefficient and variables can be arrays, series or frames. Per default
returns a pandas.Series or pandas.DataFrame of strings. If return_axes is
set to True the return value is split into values and axes, where values
are the numpy.array and axes a tuple containing index and column if
present.
Parameters
----------
tuples: tuple of tuples
Each tuple must of the form (coeff, var), where
* coeff is a numerical value, or a numerical array, series, frame
* var is a str or a array, series, frame of variable strings
as_pandas : bool, default True
Whether to return to resulting array as a series, if 1-dimensional, or
a frame, if 2-dimensional. Supersedes return_axes argument.
return_axes: Boolean, default False
Whether to return index and column (if existent)
Example
-------
Initialize coefficients and variables
>>> coeff1 = 1
>>> var1 = pd.Series(['a1', 'a2', 'a3'])
>>> coeff2 = pd.Series([-0.5, -0.3, -1])
>>> var2 = pd.Series(['b1', 'b2', 'b3'])
Create the linear expression strings
>>> linexpr((coeff1, var1), (coeff2, var2))
0 +1.0 a1 -0.5 b1
1 +1.0 a2 -0.3 b2
2 +1.0 a3 -1.0 b3
dtype: object
For a further step the resulting frame can be used as the lhs of
:func:`pypsa.linopt.define_constraints`
For retrieving only the values:
>>> linexpr((coeff1, var1), (coeff2, var2), as_pandas=False)
array(['+1.0 a1 -0.5 b1', '+1.0 a2 -0.3 b2', '+1.0 a3 -1.0 b3'], dtype=object)
"""
axes, shape = broadcasted_axes(*tuples)
expr = np.repeat("", np.prod(shape)).reshape(shape).astype(object)
if np.prod(shape):
for coeff, var in tuples:
newexpr = _str_array(coeff) + " x" + _str_array(var, True) + "\n"
if isinstance(expr, np.ndarray):
isna = np.isnan(coeff) | np.isnan(var) | (var == -1)
newexpr = np.where(isna, "", newexpr)
expr = expr + newexpr
if return_axes:
return (expr, *axes)
if as_pandas:
return to_pandas(expr, *axes)
return expr
def to_pandas(array, *axes):
"""
Convert a numpy array to pandas.Series if 1-dimensional or to a
pandas.DataFrame if 2-dimensional.
Provide index and columns if needed
"""
return pd.Series(array, *axes) if array.ndim == 1 else pd.DataFrame(array, *axes)
def _to_float_str(f):
return "%+f" % f
_v_to_float_str = np.vectorize(_to_float_str, otypes=[object])
def _to_int_str(d):
return "%d" % d
_v_to_int_str = np.vectorize(_to_int_str, otypes=[object])
def _str_array(array, integer_string=False):
if isinstance(array, (float, int)):
if integer_string:
return _to_int_str(array)
return _to_float_str(array)
array = np.asarray(array)
if array.dtype.type == np.str_:
array = np.asarray(array, dtype=object)
if array.dtype < str and array.size:
if integer_string:
array = np.nan_to_num(array, False, -1)
return _v_to_int_str(array)
return _v_to_float_str(array)
return array
def join_exprs(df):
"""
Helper function to join arrays, series or frames of strings together.
"""
return "".join(np.asarray(df).flatten())
# =============================================================================
# references to vars and cons, rewrite this part to not store every reference
# =============================================================================
def _add_reference(ref_dict, df, attr, pnl=True):
if pnl:
if attr in ref_dict.pnl:
ref_dict.pnl[attr].loc[df.index, df.columns] = df
else:
ref_dict.pnl[attr] = df
else:
if attr in ref_dict.df:
ref_dict.df = pd.concat([ref_dict.df, df.to_frame(attr)])
else:
ref_dict.df[attr] = df
def set_varref(n, variables, c, attr, spec=""):
"""
Sets variable references to the network. One-dimensional variable
references will be collected at `n.vars[c].df`, two-dimensional varaibles
in `n.vars[c].pnl`. For example:
* nominal capacity variables for generators are stored in
`n.vars.Generator.df.p_nom`
* operational variables for generators are stored in
`n.vars.Generator.pnl.p`
"""
if not variables.empty:
pnl = variables.ndim == 2
if c not in n.variables.index:
n.vars[c] = Dict(df=pd.DataFrame(), pnl=Dict())
if ((c, attr) in n.variables.index) and (spec != ""):
n.variables.at[idx[c, attr], "specification"] += ", " + spec
else:
n.variables.loc[idx[c, attr], :] = [pnl, spec]
_add_reference(n.vars[c], variables, attr, pnl=pnl)
def set_conref(n, constraints, c, attr, spec=""):
"""
Sets constraint references to the network. One-dimensional constraint
references will be collected at `n.cons[c].df`, two-dimensional in
`n.cons[c].pnl` For example:
* constraints for nominal capacity variables for generators are stored in
`n.cons.Generator.df.mu_upper`
* operational capacity limits for generators are stored in
`n.cons.Generator.pnl.mu_upper`
"""
if not constraints.empty:
pnl = constraints.ndim == 2
if c not in n.constraints.index:
n.cons[c] = Dict(df=pd.DataFrame(), pnl=Dict())
if ((c, attr) in n.constraints.index) and (spec != ""):
n.constraints.at[idx[c, attr], "specification"] += ", " + spec
else:
n.constraints.loc[idx[c, attr], :] = [pnl, spec]
_add_reference(n.cons[c], constraints, attr, pnl=pnl)
def get_var(n, c, attr, pop=False):
"""
Retrieves variable references for a given static or time-depending
attribute of a given component. The function looks into n.variables to
detect whether the variable is a time-dependent or static.
Parameters
----------
n : pypsa.Network
c : str
component name to which the constraint belongs
attr: str
attribute name of the constraints
Example
-------
>>> get_var(n, 'Generator', 'p')
"""
vvars = n.vars[c].pnl if n.variables.pnl[c, attr] else n.vars[c].df
return vvars.pop(attr) if pop else vvars[attr]
def get_con(n, c, attr, pop=False):
"""
Retrieves constraint references for a given static or time-depending
attribute of a give component.
Parameters
----------
n : pypsa.Network
c : str
component name to which the constraint belongs
attr: str
attribute name of the constraints
Example
-------
get_con(n, 'Generator', 'mu_upper')
"""
cons = n.cons[c].pnl if n.constraints.pnl[c, attr] else n.cons[c].df
return cons.pop(attr) if pop else cons[attr]
def get_sol(n, name, attr=""):
"""
Retrieves solution for a given variable. Note that a lookup of all stored
solutions is given in n.solutions.
Parameters
----------
n : pypsa.Network
c : str
general variable name (or component name if variable is attached to a
component)
attr: str
attribute name of the variable
Example
-------
get_dual(n, 'Generator', 'mu_upper')
"""
pnl = n.solutions.at[(name, attr), "pnl"]
if n.solutions.at[(name, attr), "in_comp"]:
return n.pnl(name)[attr] if pnl else n.df(name)[attr + "_opt"]
return n.sols[name].pnl[attr] if pnl else n.sols[name].df[attr]
def get_dual(n, name, attr=""):
"""
Retrieves shadow price for a given constraint. Note that for retrieving
shadow prices of a custom constraint, its name has to be passed to
`keep_references` in the lopf, or `keep_references` has to be set to True.
Note that a lookup of all stored shadow prices is given in n.dualvalues.
Parameters
----------
n : pypsa.Network
c : str
constraint name to which the constraint belongs
attr: str
attribute name of the constraints
Example
-------
get_dual(n, 'Generator', 'mu_upper')
"""
pnl = n.dualvalues.at[(name, attr), "pnl"]
if n.dualvalues.at[(name, attr), "in_comp"]:
return n.pnl(name)[attr] if pnl else n.df(name)[attr]
return n.duals[name].pnl[attr] if pnl else n.duals[name].df[attr]
# =============================================================================
# solvers
# =============================================================================
def set_int_index(ser):
ser.index = ser.index.str[1:].astype(int)
return ser
def run_and_read_highs(
n,
problem_fn,
solution_fn,
solver_logfile,
solver_options={},
warmstart=None,
store_basis=True,
):
"""
Highs solver function. Reads a linear problem file and passes it to the highs
solver. If the solution is feasible the function returns the objective,
solution and dual constraint variables. Highs must be installed for usage.
Documentation: https://www.maths.ed.ac.uk/hall/HiGHS/
Notes
-----
The script might only work for version HiGHS 1.1.1. Installation steps::
sudo apt-get install cmake # if not installed
git clone git@github.com:ERGO-Code/HiGHS.git
cd HiGHS
git checkout 95342daa73543cc21e5b27db3e0fbf7330007541 # moves to HiGHS 1.1.1
mkdir build
cd build
cmake ..
make
ctest
Then in .bashrc add paths of executables and library ::
export PATH="${PATH}:/foo/HiGHS/build/bin"
export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/foo/HiGHS/build/lib"
source .bashrc
Now when typing ``highs`` in the terminal you should see something like ::
Running HiGHS 1.1.1 [date: 2021-11-14, git hash: 95342daa]
The function reads and execute (i.e. subprocess.Popen, ...) terminal
commands of the solver. Meaning the command can be also executed at your
command window/terminal if HiGHs is installed. Executing the commands on
your local terminal helps to identify the raw outputs that are useful for
developing the interface further.
All functions below the "process = ..." do only read and save the outputs
generated from the HiGHS solver. These parts are solver specific and
depends on the solver output.
Solver options are read by the 1) command window and the 2) option_file.txt
1) An example list of solver options executable by the command window is given here:
Examples:
--model_file arg File of model to solve.
--presolve arg Presolve: "choose" by default - "on"/"off" are alternatives.
--solver arg Solver: "choose" by default - "simplex"/"ipm" are alternatives.
--parallel arg Parallel solve: "choose" by default - "on"/"off" are alternatives.
--time_limit arg Run time limit (double).
--options_file arg File containing HiGHS options.
-h, --help Print help.
2) The options_file.txt gives some more options, see a full list here:
https://www.maths.ed.ac.uk/hall/HiGHS/HighsOptions.set
By default, we insert a couple of options for the ipm solver. The dictionary
can be overwritten by simply giving the new values. For instance, you could
write a dictionary replacing some of the default values or adding new options:
```
solver_options = {
name: highs,
method: ipm,
parallel: "on",
<option_name>: <value>,
}
```
Note, the <option_name> and <value> must be equivalent to the name convention
of HiGHS. Some function exist that are not documented, check their GitHub file:
https://github.com/ERGO-Code/HiGHS/blob/master/src/lp_data/HighsOptions.h
Returns
-------
status : string,
"ok" or "warning"
termination_condition : string,
Contains "optimal", "infeasible",
variables_sol : series
constraints_dual : series
objective : float
"""
logger.warning(
"The HiGHS solver can potentially solve towards variables that slightly deviate from Gurobi, cbc, glpk"
)
options_fn = "highs_options.txt"
default_dict = {
"method": "ipm",
"primal_feasibility_tolerance": 1e-04,
"dual_feasibility_tolerance": 1e-05,
"ipm_optimality_tolerance": 1e-6,
"presolve": "on",
"run_crossover": True,
"parallel": "off",
"threads": 4,
"solution_file": solution_fn,
"write_solution_to_file": True,
"write_solution_style": 1,
"log_to_console": True,
}
# update default_dict through solver_options and write to file
default_dict.update(solver_options)
method = default_dict.pop("method", "ipm")
logger.info(
f'Options: "{default_dict}". List of options: https://www.maths.ed.ac.uk/hall/HiGHS/HighsOptions.set'
)
f1 = open(options_fn, "w")
f1.write("\n".join([f"{k} = {v}" for k, v in default_dict.items()]))
f1.close()
# write (terminal) commands
command = f"highs --model_file {problem_fn} "
if warmstart:
logger.warning("Warmstart, not available in HiGHS. Will be ignored.")
command += f"--solver {method} --options_file {options_fn}"
logger.info(f'Solver command: "{command}"')
# execute command and store command window output
process = subprocess.Popen(
command.split(" "), stdout=subprocess.PIPE, universal_newlines=True
)
def read_until_break():
# Function that reads line by line the command window
while True:
out = process.stdout.readline(1)
if out == "" and process.poll() is not None:
break
if out != "":
yield out
# converts stdout (standard terminal output) to pandas dataframe
log = io.StringIO("".join(read_until_break())[:])
log = pd.read_csv(log, sep=":", index_col=0, header=None)[1].squeeze()
if solver_logfile is not None:
log.to_csv(solver_logfile, sep="\t")
log.index = log.index.str.strip()
os.remove(options_fn)
# read out termination_condition from `info`
model_status = log["Model status"].strip().lower()
if "optimal" in model_status:
status = "ok"
termination_condition = model_status
elif "infeasible" in model_status:
status = "warning"
termination_condition = model_status
else:
status = "warning"
termination_condition = model_status
objective = float(log["Objective value"])
# read out solution file (.sol)
f = open(solution_fn, "rb")
trimed_sol_fn = re.sub(rb"\*\*\s+", b"", f.read())
f.close()
sol = pd.read_fwf(io.BytesIO(trimed_sol_fn), header=[1])
sol = sol.iloc[:-2, :] # last two rows are model info: status and objective value
row_no = sol[sol["Index"] == "Rows"].index[0]
sol = sol.drop(row_no + 1) # Removes header line after "Rows"
sol_rows = sol[(sol.index > row_no)]
sol_cols = sol[(sol.index < row_no)].set_index("Name").pipe(set_int_index)
variables_sol = pd.to_numeric(sol_cols["Primal"], errors="raise")
constraints_dual = pd.to_numeric(sol_rows["Dual"], errors="raise").reset_index(
drop=True
)
constraints_dual.index += 1
return (status, termination_condition, variables_sol, constraints_dual, objective)
def run_and_read_cbc(
n,
problem_fn,
solution_fn,
solver_logfile,
solver_options,
warmstart=None,
store_basis=True,
):
"""
Solving function. Reads the linear problem file and passes it to the cbc
solver. If the solution is successful it returns variable solutions and
constraint dual values.
For more information on the solver options, run 'cbc' in your shell
"""
with open(problem_fn, "rb") as f:
for str in f.readlines():
assert ("> " in str.decode("utf-8")) is False, ">, must be" "changed to >="
assert ("< " in str.decode("utf-8")) is False, "<, must be" "changed to <="
# printingOptions is about what goes in solution file
command = f"cbc -printingOptions all -import {problem_fn} "
if warmstart:
command += f"-basisI {warmstart} "
if (solver_options is not None) and (solver_options != {}):
command += solver_options
command += f"-solve -solu {solution_fn} "
if store_basis:
n.basis_fn = solution_fn.replace(".sol", ".bas")
command += f"-basisO {n.basis_fn} "
if not os.path.exists(solution_fn):
os.mknod(solution_fn)
log = open(solver_logfile, "w") if solver_logfile is not None else subprocess.PIPE
result = subprocess.Popen(command.split(" "), stdout=log)
result.wait()
with open(solution_fn, "r") as f:
data = f.readline()
if data.startswith("Optimal - objective value"):
status = "ok"
termination_condition = "optimal"
objective = float(data[len("Optimal - objective value ") :])
elif "Infeasible" in data:
status = "warning"
termination_condition = "infeasible"
else:
status = "warning"
termination_condition = "other"
if termination_condition != "optimal":
return status, termination_condition, None, None, None
f = open(solution_fn, "rb")
trimed_sol_fn = re.sub(rb"\*\*\s+", b"", f.read())
f.close()
sol = pd.read_csv(
io.BytesIO(trimed_sol_fn),
header=None,
skiprows=[0],
sep=r"\s+",
usecols=[1, 2, 3],
index_col=0,
)
variables_b = sol.index.str[0] == "x"
variables_sol = sol[variables_b][2].pipe(set_int_index)
constraints_dual = sol[~variables_b][3].pipe(set_int_index)
return (status, termination_condition, variables_sol, constraints_dual, objective)
def run_and_read_glpk(
n,
problem_fn,
solution_fn,
solver_logfile,
solver_options,
warmstart=None,
store_basis=True,
):
"""
Solving function. Reads the linear problem file and passes it to the glpk
solver. If the solution is successful it returns variable solutions and
constraint dual values.
For more information on the glpk solver options:
https://kam.mff.cuni.cz/~elias/glpk.pdf
"""
# TODO use --nopresol argument for non-optimal solution output
command = f"glpsol --lp {problem_fn} --output {solution_fn}"
if solver_logfile is not None:
command += f" --log {solver_logfile}"
if warmstart:
command += f" --ini {warmstart}"
if store_basis:
n.basis_fn = solution_fn.replace(".sol", ".bas")
command += f" -w {n.basis_fn}"
if (solver_options is not None) and (solver_options != {}):
command += solver_options
result = subprocess.Popen(command.split(" "), stdout=subprocess.PIPE)
result.wait()
f = open(solution_fn)
def read_until_break(f):
linebreak = False
while not linebreak:
line = f.readline()
linebreak = line == "\n"
yield line
info = io.StringIO("".join(read_until_break(f))[:-2])
info = pd.read_csv(info, sep=":", index_col=0, header=None)[1]
termination_condition = info.Status.lower().strip()
objective = float(re.sub(r"[^0-9\.\+\-e]+", "", info.Objective))
if termination_condition in ["optimal", "integer optimal"]:
status = "ok"
termination_condition = "optimal"
elif termination_condition == "undefined":
status = "warning"
termination_condition = "infeasible"
else:
status = "warning"
if termination_condition != "optimal":
return status, termination_condition, None, None, None
duals = io.StringIO("".join(read_until_break(f))[:-2])
duals = pd.read_fwf(duals)[1:].set_index("Row name")
if "Marginal" in duals:
constraints_dual = (
pd.to_numeric(duals["Marginal"], "coerce").fillna(0).pipe(set_int_index)
)
else:
logger.warning("Shadow prices of MILP couldn't be parsed")
constraints_dual = pd.Series(index=duals.index, dtype=float)
sol = io.StringIO("".join(read_until_break(f))[:-2])
variables_sol = (
pd.read_fwf(sol)[1:]
.set_index("Column name")["Activity"]
.astype(float)
.pipe(set_int_index)
)
f.close()
return (status, termination_condition, variables_sol, constraints_dual, objective)
def run_and_read_cplex(
n,
problem_fn,
solution_fn,
solver_logfile,
solver_options,
warmstart=None,
store_basis=True,
):
"""
Solving function.
Reads the linear problem file and passes it to the cplex solver. If
the solution is successful it returns variable solutions and
constraint dual values. Cplex must be installed for using this
function
"""
if find_spec("cplex") is None:
raise ModuleNotFoundError(
"Optional dependency 'cplex' not found."
"Install via 'conda install -c ibmdecisionoptimization cplex' "
"or 'pip install cplex'"
)
import cplex
_version = parse(cplex.__version__)
m = cplex.Cplex()
if solver_logfile is not None:
if _version >= Version("12.10"):
log_file_or_path = open(solver_logfile, "w")
else:
log_file_or_path = solver_logfile
m.set_log_stream(log_file_or_path)
if solver_options is not None:
for key, value in solver_options.items():
param = m.parameters
for key_layer in key.split("."):
param = getattr(param, key_layer)
param.set(value)
m.read(problem_fn)
if warmstart:
m.start.read_basis(warmstart)
m.solve()
is_lp = m.problem_type[m.get_problem_type()] == "LP"
if solver_logfile is not None:
if isinstance(log_file_or_path, io.IOBase):
log_file_or_path.close()
termination_condition = m.solution.get_status_string()
if "optimal" in termination_condition:
status = "ok"
termination_condition = "optimal"
else:
status = "warning"
if (status == "ok") and store_basis and is_lp:
n.basis_fn = solution_fn.replace(".sol", ".bas")
try:
m.solution.basis.write(n.basis_fn)
except cplex.exceptions.errors.CplexSolverError:
logger.info("No model basis stored")
del n.basis_fn
objective = m.solution.get_objective_value()
variables_sol = pd.Series(m.solution.get_values(), m.variables.get_names()).pipe(
set_int_index
)
if is_lp:
constraints_dual = pd.Series(
m.solution.get_dual_values(), m.linear_constraints.get_names()
).pipe(set_int_index)
else:
logger.warning("Shadow prices of MILP couldn't be parsed")
constraints_dual = pd.Series(index=m.linear_constraints.get_names()).pipe(
set_int_index
)
del m
return (status, termination_condition, variables_sol, constraints_dual, objective)
def run_and_read_gurobi(
n,
problem_fn,
solution_fn,
solver_logfile,
solver_options,
warmstart=None,
store_basis=True,
):
"""
Solving function. Reads the linear problem file and passes it to the gurobi
solver. If the solution is successful it returns variable solutions and
constraint dual values. Gurobipy must be installed for using this function.
For more information on solver options, see
<https://www.gurobi.com/documentation/{gurobi_verion}/refman/parameter_descriptions.html>_
"""
if find_spec("gurobipy") is None:
raise ModuleNotFoundError(
"Optional dependency 'gurobipy' not found. "
"Install via 'conda install -c gurobi gurobi' or follow the "
"instructions on the documentation page "
"https://www.gurobi.com/documentation/"
)
import gurobipy
# disable logging for this part, as gurobi output is doubled otherwise
logging.disable(50)
m = gurobipy.read(problem_fn)
if solver_options is not None:
for key, value in solver_options.items():
m.setParam(key, value)
if solver_logfile is not None:
m.setParam("logfile", solver_logfile)
if warmstart:
m.read(warmstart)
m.optimize()
logging.disable(1)
if store_basis:
n.basis_fn = solution_fn.replace(".sol", ".bas")
try:
m.write(n.basis_fn)
except gurobipy.GurobiError:
logger.info("No model basis stored")
del n.basis_fn
Status = gurobipy.GRB.Status
statusmap = {
getattr(Status, s): s.lower() for s in Status.__dir__() if not s.startswith("_")
}
termination_condition = statusmap[m.status]
if termination_condition == "optimal":
status = "ok"
elif termination_condition == "suboptimal":
status = "warning"
elif termination_condition == "infeasible":
status = "warning"
elif termination_condition == "inf_or_unbd":
status = "warning"
termination_condition = "infeasible or unbounded"
else:
status = "warning"
if termination_condition not in ["optimal", "suboptimal"]:
return status, termination_condition, None, None, None
variables_sol = pd.Series({v.VarName: v.x for v in m.getVars()}).pipe(set_int_index)
try:
constraints_dual = pd.Series({c.ConstrName: c.Pi for c in m.getConstrs()}).pipe(
set_int_index
)
except AttributeError:
logger.warning("Shadow prices of MILP couldn't be parsed")
constraints_dual = pd.Series(index=[c.ConstrName for c in m.getConstrs()])
objective = m.ObjVal
del m
return (status, termination_condition, variables_sol, constraints_dual, objective)
def run_and_read_xpress(
n,
problem_fn,
solution_fn,
solver_logfile,
solver_options,
keep_files,
warmstart=None,
store_basis=True,
):
"""
Solving function. Reads the linear problem file and passes it to the Xpress
solver. If the solution is successful it returns variable solutions and
constraint dual values. The xpress module must be installed for using this
function.
For more information on solver options, see
<https://www.fico.com/fico-xpress-optimization/docs/latest/solver/GUID-ACD7E60C-7852-36B7-A78A-CED0EA291CDD.html>_
"""
import xpress
m = xpress.problem()
m.read(problem_fn)
m.setControl(solver_options)
if solver_logfile is not None:
m.setlogfile(solver_logfile)
if warmstart:
try:
m.readbasis(n.basis_fn)
logger.info("Model basis loaded")
except:
logger.info("No model basis loaded")
pass
m.solve()
if store_basis:
n.basis_fn = solution_fn.replace(".sol", ".bss")
try:
m.writebasis(n.basis_fn)
except:
logger.info("No model basis stored")
del n.basis_fn
termination_condition = m.getProbStatusString()
if termination_condition == "mip_optimal" or termination_condition == "lp_optimal":
status = "ok"
termination_condition = "optimal"
elif (
termination_condition == "mip_unbounded"
or termination_condition == "mip_infeasible"
or termination_condition == "lp_unbounded"
or termination_condition == "lp_infeasible"
):
status = "infeasible or unbounded"
else:
status = "warning"
if termination_condition not in ["optimal"]:
return status, termination_condition, None, None, None
var = [str(v) for v in m.getVariable()]
variables_sol = pd.Series(m.getSolution(var), index=var).pipe(set_int_index)
try:
dual = [str(d) for d in m.getConstraint()]
constraints_dual = pd.Series(m.getDual(dual), index=dual).pipe(set_int_index)
except xpress.SolverError:
logger.warning("Shadow prices of MILP couldn't be parsed")
constraints_dual = pd.Series(index=dual).pipe(set_int_index)
objective = m.getObjVal()
del m
return (status, termination_condition, variables_sol, constraints_dual, objective)
