JijModeling 1.10.0 Release Notes

JijModeling 1.10.0 Release Notes#

JijModeling 1.10.0 provides critical bug fixes and feature enhancements to improve the stability and functionality of the library.

Feature Enhancements#

Interpreter 1-hot Constraint Detection#

The Interpreter now identifies 1-hot constraints during constraint evaluation in the following form:

\[ \sum_{i=1}^{n} x_i = 1, \quad x_i \in \{0, 1\} \]

This enhancement provides hints to improve solver performance for compatible solvers. Detected constraints can be accessed via the constraint_hints property in ommx.v1.Instance.

import jijmodeling as jm

n = jm.Placeholder("n")
x = jm.BinaryVar("x", shape=(n,))
i = jm.Element("i", belong_to=(0,n))

problem = jm.Problem("problem")
# Add 1-hot constraint
problem += jm.Constraint("onehot", jm.sum(i, x[i]) == 1)

interpreter = jm.Interpreter({"n": 3})
instance = interpreter.eval_problem(problem)

# Display detected constraint information
print(instance.raw.constraint_hints)

one_hot_constraints {
  decision_variables: 0
  decision_variables: 1
  decision_variables: 2
}

Future updates will enable the detection of various constraints, including SOS1/SOS2 and n-hot constraints.

Further Improvements in Type Hints#

Type hinting across submodules has been significantly improved to resolve Pyright errors that previously occurred. This improvement leverages advancements in pyo3-stub-gen (see) to ensure seamless type checking in submodules.

Prior to this release, type errors were common in submodules:

image.png

These issues are now resolved, ensuring a more robust developer experience:

image.png

Bug Fixes#

Fixed ndim Calculation for Jagged Arrays#

This release resolves a logic error in calculating the ndim (number of dimensions) for jagged arrays. The bug caused errors during Interpreter initialization and prevented evaluation of instances loaded through the MIPLIB loader.

import jijmodeling as jm

# Load MIPLIB instance "air05"
dataset = jm.dataset.Miplib()
problem, instance_data = dataset.load("air05")

try:
    # In jijmodeling <= 1.9.0, either of these two lines would raise an InterpreterError
    interpreter = jm.Interpreter(instance_data)
    interpreter.eval_problem(problem)
except jm.InterpreterError as e:
    print(f"InterpreterError: {e}")

Corrected Arithmetic Assignment Operator Behavior (+=) for Objective Functions#

A long-standing issue with the += operator for objective functions has been fixed. Previously, adding to an existing objective would overwrite the original instead of appending to it.

import jijmodeling as jm

x = jm.BinaryVar("x")
y = jm.BinaryVar("y")

problem = jm.Problem("problem")
# Add x to objective function
problem += x
# Add y to objective function
problem += y

# In jijmodeling <= 1.9.0, the objective function would be y instead of x + y
assert jm.is_same(problem.objective, x + y)

Changelog#

Bug Fixes#

  • Fixed incorrect ndim calculation for jagged arrays, ensuring seamless Interpreter initialization and MIPLIB loader evaluation.

  • Resolved behavior of the += operator to properly append terms to existing objective functions.

Feature Enhancements#

  • 1-Hot Constraint Detection: Supported constraints must adhere to the following form: \( \displaystyle \sum_{i=1}^{n} x_i = 1, \quad x_i \in \{0, 1\} \)

    • Left-hand side: exactly one summation symbol.

    • Summation operand: a single binary variable with an index.

    • Constraint type: equality.

    • Right-hand side: a NumberLit with value of 1.

  • Resolved Pyright errors in submodules by improving type hint support through pyo3-stub-gen enhancements.