Skip to main content
Join the official 2020 Python Developers SurveyStart the survey!

Utilities for parsing MPS and SMPS file formats.

Project description


This is a utility script for parsing MPS and SMPS file formats. It offers two main functions load_mps for loading mps files and load_smps for loading smps file directory.


The load_mps(path) method takes a path variable as input. It should be a .cor or .mps file. It opens the file with read-permissions and parses the described linear program into the following format:

  • name: The name given to the linear program (can't be blank)
  • objective_name: The name of the objective function value
  • row_names: list of row names
  • col_names: list of column names
  • types: list of constraint type indicators, i.e. either "E", "L" or "G" for equality, lower/equal or greater/equal constraint respectively.
  • c: the objective function coefficients
  • A: the constraint matrix
  • rhs_names: list of names of right hand sides (there can be multiple right hand side components be defined, seldom more than one though)
  • rhs: dictionary (rhs_name) => b, where b is the vector of constraint values for that given right hand side name.
  • bnd_names: list of names of box-bounds (seldom more than one)
  • bnd: dictionary (bnd_name) => {"LO": v_l, "UP": v_u} where v_l is the vector of lower bounds and v_u is the vector of upper bound values (defaults to v_l = 0 and v_u = +inf).

Finally this corresponds to the linear program

min 	c * x

s.t.	for each rhs_name with corresponding b:

			A[types == "E",:] * x  = b[types == "E"]
			A[types == "L",:] * x <= b[types == "L"]
			A[types == "G",:] * x >= b[types == "G"]

		for each bnd_name with corresponding v_l and v_u:

			v_l <= x < v_u


This function makes use of the load_mps function for parsing the .cor file. The SMPS file format consists of three files, a .cor, .tim and .sto file. The .cor file is in MPS format. Further the function expects a parameter path to be such that path + ".cor" is the core file, path + ".tim" the time file and path + ".sto" is the stochastic file. It does not support scenarios yet! It returns a stochastic multi-stage problem in the following format

  • name: name of the program (must be the same in all 3 files)

  • objective_name: name of the objective function value

  • constraints: list of tuples (name, period, type) for each constraint. It gives a name, a period in which the constraints appears and a type, i.e. "E", "L" or "G" as in MPS.

  • variables: list of tuples (name, period, type) for each variable. It defines a name and a period in which the variable joins the program. type denotes a string that is either integral or continuous depending on whether the INTORG MARKER was set when instantiating the variable.

  • c: vector of objective function coefficients (of all periods)

  • A: matrix of constraint coefficients (of all periods)

  • rhs_names: list of rhs names as in MPS

  • rhs: dictionary as in MPS

  • bounds: dictionary as in MPS

  • periods: list of all periods appearing. len(periods) is the number stages.

  • blocks: dictionary of Block,LinearTransform or SubRoutine objects. Dependent on what the .sto file defined. Blocks are independent random variables (every case of a Block must be combined with each case of another Block to get all possible appearences; the probabilities multiply), LinearTransform are linear transformations of continuous random variables. The user needs to write the sample script on his own. SubRoutine is a left-out in the file; it presupposes the user to know what to do with these values.

  • independent_variables: dictionary ((i,j)) => {position, period, distrib}, where (i,j) is the tuple of row/column indices. If one of them is -1 this means that it's either an objective value or a rhs-value respectively. position is a dictionary adapting to where the entry is (objective value, rhs value or matrix value), period defines the period in which this variable is stochastic, distrib is either a definition of a continuous random variables

    distrib: {type: "N(mu, sigma**2)"/"U(a, b)"/"B(p, q)"/"G(p, b)"/"LN(mu, sigma**2)", parameters}

    where parameters is a dictionary defining the required parameters. In the discrete case it is a list of tuples (v,p), where v is the value of this position and p is the probability of it appearing.

For an example on how to use this format i recommand looking at the code for load_2stage_problem.


Loads a SMPS directory and tries to bring it into a 2-staged stochastic linear program with fixed recourse. Output is a dictionary containing the values

  • c: first stage objective function value
  • A: first stage (equality) constraint coefficient matrix
  • b: first stage constraint values
  • q: list of second stage objective function coefficients (each case one entry)
  • h: list of second stage constraint values (each case one entry)
  • T: list of second stage constraint values for deterministic variables (each case one entry)
  • W: recourse matrix (since it's fixed recourse this is not a list)
  • p: list of probabilities for each case

The constellations in which (q,h,T,W) appear are the realizations given by (q[k], h[k], T[k], W). The problem then resembles one of the form

min		c * x + E_p[q * y]

s.t.	A * x         = b
    	T * x + W * y = h
    	x, y >= 0

which is a formal expression since T and h are also stochastic. In fact this notation means we assert the stochastic constraints inside of the expectation, making it a function of x only.

For casting the SMPS files into such a form we need to make certain assertments:

  • The upper right matrix needs to be zeroes only.
  • We only have one righthand side defined (len(rhs_names) == 1).
  • There are no boundaries or if we defined some they are the default values.
  • The first period parsed from the time file is the deterministic one, the other one is the stochastic one (especially there can only be two periods).
  • A and W are not stochastic.

This script however does

  • convert inequality constraints (deterministic and stochastic) into equality constraints by adding slack variables at the right places
  • calculate all combinations of independent accurances of stochastic components (BLOCKS and INDEP)
  • calculate the probabilities as products of independent elementary probabilities alongside.

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for pysmps, version 1.5.3
Filename, size File type Python version Upload date Hashes
Filename, size pysmps-1.5.3.tar.gz (8.3 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page