NLSL

The NLSL project provides a Python interface to the ACERT nonlinear least squares (NLSL) spectroscopic fitting toolkit, bundling the original Fortran core with a modern Python API for building, fitting, and analysing spin resonance models.【F:pyproject.toml†L11-L24】【F:nlsl/init.py†L1-L104】

Key features

• Pythonic parameter access – the :class:nlsl.nlsl class exposes the complete parameter set from the legacy NLSL engine through mapping-style accessors, making it easy to read or update magnetic, diffusion, and fitting options from Python code.【F:nlsl/init.py†L60-L170】
• Programmable fitting control – the :class:nlsl.fit_params helper keeps the low-level Levenberg–Marquardt settings in sync with the Fortran structures, so tolerances, iteration limits, and other solver options can be modified directly without crafting text run files.【F:nlsl/init.py†L15-L58】
• Runfile compatibility – existing NLSL .run scripts can be replayed via :meth:nlsl.nlsl.procline, preserving the familiar workflow while allowing incremental migration to the Python API.【F:nlsl/init.py†L82-L109】
• Spectrum capture – helper methods such as :meth:nlsl.nlsl.current_spectrum and :meth:nlsl.nlsl.write_spc let you evaluate the current model or persist generated spectra for downstream analysis and visualisation.【F:nlsl/init.py†L105-L158】

Installation

Clone the repository and install the package in editable mode so the compiled Fortran components stay in sync with your working tree:

pip install -e . --no-build-isolation

If you see errors about NumPy being “incompatible”

acert_nlsl requires NumPy ≥ 2.0.
This might lead to errors that have two solutions.

One option is to create a new python environment -- e.g. on Anaconda, you would do:

conda create -n nlsl_env python=3.13

and then install again inside this environment. Whenever you went to use the package, you would then want to make sure you're inside the right environment with: conda activate nlsl_env

The other option is to upgrade your existing environment:

If your environment already has older packages that pin NumPy < 2.0 (e.g., gensim, numba), you must upgrade those packages so they accept modern NumPy.

Run:

pip install --upgrade gensim numba

If additional packages report similar “requires numpy<2.x” errors, upgrade them the same way. After all incompatible packages are upgraded, install NLSL normally:

pip install acert_nlsl

Usage overview

Instantiate :class:nlsl.nlsl to work with parameters programmatically, or stream traditional runfiles back into the engine.

import nlsl

n = nlsl.nlsl()
n["nsite"] = 1
n.fit_params["maxitr"] = 40
n.procline("data sampl1 ascii nspline 200 bc 20 shift")
site_spectra = n.fit()
total_spectrum = n["weights"] @ site_spectra

The mapping interface mirrors the parameter names defined by the original Fortran code (see nlshlp.txt for a full command reference), while :meth:nlsl.nlsl.fit runs the nonlinear least-squares optimiser and returns the latest site spectra, while the companion weight matrix stays accessible through the mapping interface.【F:nlsl/init.py†L60-L204】

Examples

A curated set of runnable examples is included under examples/ to help you get started with real datasets and typical NLSL workflows:

Download the examples directory

After installing, you should be able to run nlsl exampledir on any command line, which will unpack the examples directory as NLSL_examples. (This is the examples directory that it gets from unpacking a zip of the current code on github.)

Description of examples

• Runfile suite (sampl1–sampl5) – canonical .run scripts paired with .dat inputs that demonstrate anisotropic rotations, multi-spectrum fits, and other standard analyses. Run them with the classic command-line interface and compare the generated .spc and .log outputs to the provided reference files to verify your setup.【F:examples/README.rst†L1-L22】
• runexample.py – command-line helper that executes one of the numbered runfile scenarios, cascades any nested call directives, reports relative RMS errors, and plots the resulting spectra with experimental overlays for quick diagnostics.【F:examples/runexample.py†L1-L93】
• runexample_first.py – simplified variant that focuses on the first sample run, computing a global relative RMS metric while rendering the fitted spectrum.【F:examples/runexample_first.py†L1-L74】
• break_down_example.py – step-by-step walkthrough of the first sample showcasing programmatic parameter updates, basis set configuration, and fit execution directly from Python without relying on runfiles.【F:examples/break_down_example.py†L1-L83】
• BL05CHSA.py – reproduces the BL05CHSA.run workflow, highlighting how to plot individual spectral components alongside the overall fit and cumulative integrals.【F:examples/BL05CHSA.py†L1-L40】
• c16pc371e.py – demonstrates interacting with a custom runfile while inspecting the solver configuration before and after execution and plotting the normalised spectra.【F:examples/c16pc371e.py†L1-L46】

Each script expects to be run from within the examples/ directory so that the associated data and reference outputs are discovered automatically.【F:examples/runexample.py†L22-L41】

Additional resources

• The examples/README.rst file documents the legacy testing workflow for the runfile suite and points to the reference outputs for verification.
• nlshlp.txt lists every command available to the original interpreter along with detailed descriptions of their arguments and expected effects.