What Is Model JSON

Definition

A portable description of your optimization model

JSON (JavaScript Object Notation) is a widely used structured text format. In this context, model JSON is a JSON object that captures:

• Objective sense (maximize or minimize),
• Objective coefficients,
• Constraint matrix and senses (≤, =, ≥),
• Right-hand-side values,
• Variable types (continuous, integer, binary) and bounds,
• Key solver options (integrality tolerance, time/node limits, method choice).

When you click “Export model as JSON” in the ILP/MILP calculator, you get this structure as a text block you can save or share.

Structure

High-level schema (informal)

{
  "sense": "max" or "min",
  "n": <number of variables>,
  "m": <number of constraints>,
  "c": [c1, c2, ..., cn],            // objective coefficients
  "A": [[...], [...], ...],         // m × n constraint matrix
  "b": [b1, b2, ..., bm],           // right-hand sides
  "senseRows": ["<=", "=", ">=", ...],
  "varTypes": ["C","I","B", ...],   // C=continuous, I=integer, B=binary
  "lb": [l1, l2, ..., ln],          // lower bounds
  "ub": [u1, u2, ..., un],          // upper bounds
  "options": {
    "method": "branch_and_bound" or "branch_and_cut_small",
    "timeLimitMs": 5000,
    "nodeLimit": 5000,
    "integralityTol": 1e-9,
    "showSteps": true,
    "branchingRule": "most_fractional",
    "searchStrategy": "best_bound"
  }
}

The exact field names may evolve as the calculator grows, but the intent remains: one JSON object contains everything needed to re-create your ILP/MILP run.

Fields

Field-by-field meaning

Example

Project selection example (0–1 ILP)

Model:

maximize   5 x1 + 3 x2
subject to 2 x1 + 1 x2 ≤ 3
           x1, x2 ∈ {0,1}

One possible model JSON representation:

{
  "sense": "max",
  "n": 2,
  "m": 1,
  "c": [5, 3],
  "A": [[2, 1]],
  "b": [3],
  "senseRows": ["<="],
  "varTypes": ["B", "B"],
  "lb": [0, 0],
  "ub": [1, 1],
  "options": {
    "method": "branch_and_bound",
    "timeLimitMs": 5000,
    "nodeLimit": 5000,
    "integralityTol": 1e-9,
    "showSteps": true,
    "branchingRule": "most_fractional",
    "searchStrategy": "best_bound"
  }
}

If you paste this JSON into the “Import model from JSON” box in the ILP/MILP calculator and click “Import,” the tables will be filled with this model. Click “Solve” to re-run the optimization exactly.

People ask

Is model JSON specific to CalcTypes, or is it a standard?

Model JSON as described here is **specific to the CalcTypes ILP/MILP calculator**. It’s designed to be easy to read and manipulate, but it is not an official MPS/LP/MIP standard. Many solvers have their own formats (e.g., MPS, LP files, ProtoBufs).

People ask

Can I edit model JSON by hand?

Yes, for small models and careful users. You can:

• Change coefficients, bounds, or types,
• Adjust solver options (time limits, integrality tolerance),
• Add/remove constraints and variables (keeping dimensions consistent).

However, typos or mismatched lengths (e.g., n not matching the length of c) will cause errors or undefined behavior.

People ask

Is model JSON stable over time?

The core fields (objective, constraints, types, bounds) are expected to remain stable. Additional fields may be added to options as new solver features are exposed. The calculator will aim to be backward-compatible with reasonable older JSON files; when breaking changes are needed, they will be documented.

Workflow

Exporting your current model

1. Build your ILP/MILP model in the tables of the ILP/MILP calculator.
2. Open the “Advanced sharing (Import/Export JSON)” panel.
3. Click “Export model as JSON.”
4. Copy the JSON text and paste it into your notes, Git repo, or a shared document.

Workflow

Importing a model from JSON

1. Paste a valid model JSON object into the “Import model from JSON” textarea.
2. Click “Import model from JSON.” The tables should update to reflect the new model.
3. Click “Solve” to run the model with the current solver options.

Tip: if import fails, check your JSON with a validator (e.g., jsonlint.com) to ensure it’s syntactically correct.

Sharing

Model JSON vs share links

In the ILP/MILP calculator, both options are available. For long-term storage, model JSON is usually better; for quick reproduction, share links are convenient.

Mistakes

Mistake: mismatched dimensions (n, m, and arrays)

If n says “3 variables” but your vectors/rows have 2 or 4 elements, import will fail or behave incorrectly. Always ensure that:

• c.length == n,
• A.length == m and each row has length n,
• b.length == m, varTypes.length == n, lb.length == n, ub.length == n.

Mistakes

Mistake: editing numbers but forgetting to update bounds or types

If you change a variable from binary to continuous in varTypes but leave its bounds at [0,1], the model logic may change more than expected. Similarly, converting a continuous variable to integer without updating bounds can unexpectedly enlarge the search space.

Next steps

Make model JSON part of your modeling workflow

1. Build a small ILP/MILP in the ILP/MILP calculator and export its model JSON.
2. Check the JSON into version control (Git) alongside your data and reports.
3. When you change a constraint or objective, update the JSON and write a short change note.
4. Use share links for quick collaboration, and model JSON for long-term reproducibility.