Automating Leapfrog Workflows with Pollywog – An Independent Open-Source Tool

Pollywog is a Python library for building, manipulating, and automating Leapfrog calculation sets programmatically.

Table of Contents

• Why Pollywog?
• Key Features
• Quick Start
• Installation
• Usage Examples
• Documentation
• Contributing
• License

Why Pollywog?

If you work with Leapfrog for geological modeling and resource estimation, you know that building calculation sets (.lfcalc files) manually can be:

• Time-consuming – Repetitive point-and-click operations
• Error-prone – Easy to make mistakes in complex formulas
• Hard to maintain – Difficult to update across multiple projects
• Not version-controlled – Changes are hard to track and review
• Not automatable – Can't script or integrate with other tools

Pollywog solves these problems by letting you define calculations in Python code that is:

• Programmatic and automatable
• Version-controlled (Git-friendly)
• Testable and reproducible
• Easy to refactor and maintain
• Integrated with ML pipelines (scikit-learn)

Key Features

Core Functionality

• Read and write .lfcalc files programmatically
• Create calculations with Python classes (Number, Category, If, etc.)
• Query and filter calculation sets like pandas DataFrames
• Topological sorting for automatic dependency resolution
• Rich display in Jupyter notebooks with interactive trees

Helper Functions

• Mathematical operations: Sum, Product, Average, WeightedAverage
• Transformations: Scale, Normalize
• Classification: CategoryFromThresholds
• Dual mode: Return complete calculations (with name) or expressions (without name) for composition

Machine Learning Integration

• Convert scikit-learn decision trees to Leapfrog calculations
• Convert random forests to ensemble calculations
• Convert linear models to equations
• Support for both regression and classification

Domain-Based Calculations

• Multi-domain resource modeling
• Weighted averages by domain proportions
• Conditional logic for different geological units

Quick Start

from pollywog.core import CalcSet, Number
from pollywog.helpers import WeightedAverage

# Create a calculation set
calcset = CalcSet([
    # Clean data
    Number("Au_clean", "clamp([Au], 0)",
           comment_equation="Remove negative values"),

    # Domain-weighted grade
    WeightedAverage(
        variables=["Au_oxide", "Au_sulfide", "Au_transition"],
        weights=["prop_oxide", "prop_sulfide", "prop_transition"],
        name="Au_composite",
        comment="Domain-weighted gold grade"
    ),

    # Apply recovery
    Number("Au_recovered", "[Au_composite] * 0.88",
           comment_equation="88% metallurgical recovery"),
])

# Export to Leapfrog
calcset.to_lfcalc("my_calculations.lfcalc")

Then import my_calculations.lfcalc into Leapfrog and you're done! ✨

⚠️ Note: Pollywog is in active development. Always backup your Leapfrog projects before testing. Report issues on GitHub.

Legal Disclaimer

Pollywog is an independent open-source tool developed to support the automation of workflows involving .lfcalc files used in Leapfrog software by Seequent. This tool does not perform reverse engineering, does not modify Leapfrog, and does not access its source code or proprietary libraries. Pollywog operates exclusively on user-generated files and is designed to complement Leapfrog through external automation.

Important:

• Pollywog is not affiliated with, endorsed by, or sponsored by Seequent or any company associated with Leapfrog
• Use of this tool does not violate Leapfrog’s license terms or Seequent’s policies
• Users are encouraged to review Leapfrog’s terms of use before integrating Pollywog into commercial or corporate environments
• The author is not responsible for any misuse of the tool that may breach Seequent’s licensing terms

Installation

From PyPI (Recommended)

pip install lf_pollywog

From GitHub (Latest Development Version)

pip install git+https://github.com/endarthur/pollywog.git

Try in Your Browser (No Installation)

Try Pollywog without installing anything using JupyterLite: https://endarthur.github.io/pollyweb

Note: JupyterLite runs in your browser and has limitations (no file system access, limited libraries). Files are stored in browser memory and won't persist if you clear your cache. Download your work regularly! For production use, preferably install locally.

Usage Examples

1. Reading and Writing .lfcalc Files

from pollywog.core import CalcSet, Number

# Read existing file
calcset = CalcSet.read_lfcalc("path/to/file.lfcalc")

# Modify calculations
calcset.items.append(Number("new_calc", "[Au] * 2"))

# Export modified version
calcset.to_lfcalc("output.lfcalc")

2. Creating Calculations from Scratch

from pollywog.core import Number, CalcSet

calcset = CalcSet([
    Number("Au_clean", "clamp([Au], 0)",
           comment_equation="Remove negative values"),
    Number("Au_log", "log([Au_clean] + 1e-6)",
           comment_equation="Log transform for kriging"),
])

calcset.to_lfcalc("drillhole_preprocessing.lfcalc")

3. Using Helper Functions

Helpers can return either complete calculations or just expressions for composition:

from pollywog.helpers import WeightedAverage, Product, CategoryFromThresholds
from pollywog.core import CalcSet, Number

calcset = CalcSet([
    # With name: Returns complete Number object
    WeightedAverage(
        variables=["Au_oxide", "Au_sulfide", "Au_transition"],
        weights=["prop_oxide", "prop_sulfide", "prop_transition"],
        name="Au_composite",
        comment="Domain-weighted gold grade"
    ),

    # Calculate gold equivalent (Ag and Cu converted to Au)
    Number("AuEq",
        "[Au_composite] + ([Ag_composite] * 0.011) + ([Cu_composite] * 1.5)",
        comment_equation="Gold equivalent grade (Ag/91, Cu*1.5 for price ratio)"
    ),

    # Without name: Returns expression for composition
    # Calculate net smelter return (NSR) per tonne
    Number("NSR_per_tonne",
        f"{Product(['Au_composite', '1800', '0.88'])} + "  # Au: price $1800/oz, 88% recovery
        f"{Product(['Ag_composite', '22', '0.75'])} + "    # Ag: price $22/oz, 75% recovery
        f"{Product(['Cu_composite', '3.5', '0.85'])}"      # Cu: price $3.5/lb, 85% recovery
    ),

    # Classify by gold equivalent grade
    CategoryFromThresholds(
        variable="AuEq",
        thresholds=[0.5, 2.0],
        categories=["waste", "low_grade", "high_grade"],
        name="ore_class"
    ),
])

calcset.to_lfcalc("resource_model.lfcalc")

4. Machine Learning Model Conversion

Deploy machine learning models directly in Leapfrog:

from pollywog.conversion.sklearn import convert_tree, convert_forest
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import RandomForestRegressor
from pollywog.core import CalcSet
import numpy as np

# Training data: Au grade, Cu grade, grind size (P80)
X = np.array([[1.2, 0.3, 75], [0.8, 0.5, 100], [2.0, 0.2, 75]])
y = np.array([0.88, 0.82, 0.91])  # Recovery values

# Train and convert decision tree
model = DecisionTreeRegressor(max_depth=3, random_state=42)
model.fit(X, y)

recovery_calc = convert_tree(
    model,
    ["Au_composite", "Cu_composite", "P80"],
    "Au_recovery_predicted"
)

# Export to Leapfrog
CalcSet([recovery_calc]).to_lfcalc("ml_recovery_model.lfcalc")

5. Domain-Based Calculations

from pollywog.core import CalcSet, Number, If
from pollywog.helpers import WeightedAverage

domains = ["oxide", "transition", "sulfide"]
metals = ["Au", "Ag", "Cu"]

# Domain-weighted grades for all metals
calcset = CalcSet([
    WeightedAverage(
        variables=[f"{metal}_{domain}" for domain in domains],
        weights=[f"prop_{domain}" for domain in domains],
        name=f"{metal}_composite",
        comment=f"Domain-weighted {metal} grade"
    )
    for metal in metals
])

# Apply domain-specific recovery
# Note: If objects require a list since they are separate structures
calcset.items.append(
    Number("Au_recovered", [
        If([
            ("[domain] = 'oxide'", "[Au_composite] * 0.92"),
            ("[domain] = 'transition'", "[Au_composite] * 0.85"),
            ("[domain] = 'sulfide'", "[Au_composite] * 0.78"),
        ], otherwise=["[Au_composite] * 0.75"])
    ])
)

calcset.to_lfcalc("multi_domain_workflow.lfcalc")

6. Querying CalcSets

Filter calculations like pandas DataFrames:

# Select items by name pattern
au_calcs = calcset.query('name.startswith("Au")')

# Use external variables
metals_of_interest = ["Au", "Ag"]
selected = calcset.query('any(name.startswith(m) for m in @metals_of_interest)')

# Complex queries
filtered = calcset.query('len(expression) > 1 and "log" in name')

• Use item attributes (e.g., name, item_type) in expressions.
• Reference external variables using @var syntax (e.g., name.startswith(@prefix)).
• Supported helpers: len, any, all, min, max, sorted, re, str.

Documentation

📚 Full documentation: https://pollywog.readthedocs.io

• Getting Started – Installation and first steps
• Tutorial for Resource Geologists – For non-programmers familiar with Leapfrog
• Tutorials – Complete workflow examples
• Expression Syntax – Leapfrog expression reference
• Workflow Patterns – Common patterns and recipes
• Helper Functions – Helper function guide
• Best Practices – Production recommendations
• API Reference – Complete API documentation

License

MIT License – See LICENSE file for details.

Contributions

Contributions are very welcome! Please see CONTRIBUTING.md for detailed guidelines.

Quick start:

• Fork the repository
• Create a feature branch (git checkout -b feature-name)
• Make your changes and commit (git commit -m 'Add new feature')
• Submit a pull request with a clear explanation of your changes

Before contributing:

• Read CONTRIBUTING.md for full guidelines
• Follow the Code of Conduct
• Run tests and formatters (pytest, black, ruff)
• It's ok to use LLMs to help write code, but review everything carefully

Feel free to open an issue if you have questions or suggestions.

Acknowledgements

Thanks to Debora Roldão for helping with organization of the project, documentation and design, Eduardo Takafuji for the initial discussion of the feasability of this all those years ago and Jessica da Matta for support and sanity checks along the way.

Links

• 📘 Documentation: https://pollywog.readthedocs.io
• 💻 GitHub: https://github.com/endarthur/pollywog
• 🚀 Try Online: https://endarthur.github.io/pollyweb
• 📦 PyPI: https://pypi.org/project/lf-pollywog/
• 🐛 Issues: https://github.com/endarthur/pollywog/issues
• 📖 Examples: examples/ folder