leapfrog-goals 0.0.4

Python interface for Leapfrog Goals model

goals

Installation

pip install leapfrog-goals

Usage

Get parameters

In the following example you will need parameters - inputs to the model. We have provided utilities to easily get test parameters (we use hdf5 file format to store these). If you already have different parameters in the correct shape you can skip this section.

from leapfrog_goals import read_h5_file

parameters_child = read_h5_file("../leapfrogr/tests/testthat/testdata/child_parms_full.h5")

We need to add a good way to import the goals parameters, for now add manually.

import numpy as np
parameters_child["ex_input"] = np.full((81, 2), 1)

Run model

Here are a couple of ways to run the model

from leapfrog_goals import run_goals

# Run the model
run_goals(parameters_child)

# Can specify which years to output from the model
run_goals(parameters_child, [1970, 1971, 2000])

Run model from initial state (TODO)

You can also run the model from an initial state. For example, to run the model in two parts

from leapfrog_goals import run_goals, run_goals_from_state, get_time_slice

# Run model until year 2000
ret = run_goals(parameters_child, range(1970, 2001))

run_goals_from_state(
    # parameters
    parameters_child,

    # get the last time slice of the state returned by run model so we
    # can continue from where the model left off
    get_time_slice(ret, 30),

    # simulation start year, this should be the year the previous
    # argument represents
    2000,

    # years you want the model to output
    range(2001, 2031)
)

Run model for a single year

The previous two functions will return a dictionary with values that are have one more dimension (time dimension) than the state specified in the model config. This means if you want to carry on from that initial single year state you would have to use the get_time_slice utility at index 1. Example:

from leapfrog_goals import run_goals, run_goals_from_state, get_time_slice

ret_single_year = run_goals(parameters_child, [1970])

for year in range(1970, 2030):
    ret_single_year = run_goals_from_state(
        parameters_child,
        get_time_slice(ret_single_year, 0),
        year,
        [year + 1]
    )

This is tedious for two reasons:

1. Repeatedly getting the time slice for one time dimension (initial one is unavoidable since we use run model and don't have any initial state)
2. Specifying output year as year + 1

So to make projecting for a single year more ergonomic we decided to create a separate function that returns the time slice for a single year:

from leapfrog_goals import run_goals, run_goals_single_year, get_time_slice

ret_single_year = get_time_slice(run_goals(parameters_child, [1970]), 0)

for year in range(1970, 2030):
    ret_single_year = run_goals_single_year(
        parameters_child,
        ret_single_year,
        year
    )

Development

This project uses scikit-build-core with nanobind to build the C++ project. You'll need a recent version of CMake (>3.15) and Python (>3.7).

We use uv to manage the project:

• Sync the virtual env uv sync
• Run tests uv run pytest
• Lint uvx ruff check .
• Format uvx ruff format
• Format imports uvx ruff check --select I --fix

uv will rebuild automatically if you make a change to the C++ code however if you need to force recompilation use --reinstall-package leapfrog-goals.

For local installation run pip install .

License

leapfrog-goals is distributed under the terms of MIT license.