Version History¶

This page outlines changes from each release.

v1.0.0 (April 20, 2020)¶

Refactored the entire package; sasoptpy now has core, abstract, interface, session, and util directories
Experimental RESTful API was dropped
get_obj_by_name function was removed
lso solver was renamed blackbox
Because of the use of literal strings (PEP 498), only Python 3.6 or later versions are supported

Support for evaluating nonlinear expressions is added, see Expression.get_value() and utils._evaluate()
Support for multiple objectives is added for LSO solver, see Model.set_objective() and Multiobjective example
Support for spaces inside variable indices is added
Experimental RESTful API is added

Dictionaries inside components are replaced with ordered dictionaries to preserve deterministic behavior
Math operators are added into the keys of linear coefficient dictionaries
Some iterators are rewritten by using the yield keyword for performance
key_name and col_names parameters are added into read_table()

Fixed: Using a single variable as an objective is producing incorrect input
Fixed: Expression.get_value() fails to evaluate expressions with operators
Fixed: Expression.add() overrides operators in some instances
Fixed: Expressions with same components but different operators get summed incorrectly
Fixed: New version of Viya complains about pandas.DataFrame column types
Syntax fixes for PEP 8 compliance

Support for the new runOptmodel CAS action is added
Nonlinear optimization model building support is added for both SAS 9.4 and SAS Viya solvers
Abstract model building support is added when using SAS Viya solvers
New object types, Set, SetIterator, Parameter, ParameterValue, ImplicitVar, ExpressionDict, and Statement are added for abstract model building
Model.to_optmodel() method is added for exporting model objects into PROC OPTMODEL code as a string
Wrapper functions read_table() and read_data() are added to read CASTable and DataFrame objects into the models
Math function wrappers are added
_expr and _defn methods are added to all object types for producing OPTMODEL expression and definitions
Multiple solutions are now returned when using solveMilp action and can be retrieved by using Model.get_solution() method
Model.get_variable_value() is added to get solution values of abstract variables

Variable and constraint naming schemes are replaced with OPTMODEL equivalent versions
Variables and constraints now preserve the order in which they are inserted to the problem
Model.to_frame() method is updated to reflect changes to VariableGroup and ConstraintGroup orderings
Two solve methods, solve_on_cas and solve_on_viya are merged into Model.solve()
Model.solve() method checks the available CAS actions and uses the runOptmodel action whenever possible
As part of the merging process, lp and milp arguments are replaced with options argument in Model.solve() and Model.to_optmodel()
An optional argument frame is added to Model.solve() for forcing use of MPS mode and solveLp-solveMilp actions
Minor changes are applied to __str__ and __repr__ methods
Creation indices for objects are being kept by using the return of the register_name() function
Objective constant values are now being passed by using new CAS action arguments when posssible
A linearity check is added for models
Test folder is added to the repository

Abstract and nonlinear models can be solved on Viya only if the runOptmodel action is available on the CAS server.
Three new examples are added which demonstrate abstract model building.
Some minor changes are applied to the existing examples.

As an experimental feature, sasoptpy now supports SASPy connections
Model.solve_local() method is added for solving optimization problems by using SAS 9.4 installations
Model.drop_variable(), Model.drop_variables(), Model.drop_constraint(), Model.drop_constraints() methods are added
Model.get_constraint() and Model.get_constraints() methods are added to retrieve Constraint objects in a model
Model.get_variables() method is added
_dual attribute is added to the Expression objects
Variable.get_dual() and Constraint.get_dual() methods are added
Expression.set_name() method is added

Session argument accepts saspy.SASsession objects
VariableGroup.mult() method now supports pandas.DataFrame
Type check for the Model.set_session() is removed to support new session types
Problem and solution summaries are not being printed by default anymore, see Model.get_problem_summary() and Model.get_solution_summary()
The default behavior of dropping the table after each solve is changed, but can be controlled with the drop argument of the Model.solve() method

A .gitignore file is added to the repository.
A new example is added: Decentralization. Both CAS and SAS versions of the new example are added.
There is a known issue with the nondeterministic behavior when creating MPS tables. This will be fixed with a hotfix after the release.
A new option (no-ex) is added to makedocs script for skipping examples when building docs.

Initial value argument ‘init’ is added for Variable objects
Variable.set_init() method is added for variables
Initial value option ‘primalin’ is added to Model.solve() method
Table name argument ‘name’, table drop option ‘drop’, and replace option ‘replace’ are added to Model.solve() method
Decomposition block implementation is rewritten; block numbers does not need to be consecutive and ordered Model.upload_user_blocks()
VariableGroup.get_name() and ConstraintGroup.get_name() methods are added
Model.test_session() method is added for checking if session is defined for models
expr_sum() function is added for faster summation of Expression objects