Installation

ScientificFitting supports Julia 1.10 and later. CI targets Julia 1.10 and the latest stable Julia; the optional NativeMinuit adapter requires Julia 1.11+. Documentation rendering is pinned to Julia 1.12.

Install The Julia Package

Install from Julia's General registry:

using Pkg
Pkg.add("ScientificFitting")

Then load the numerical core:

using ScientificFitting

This loads fitting, likelihoods, diagnostics, profiles, contours, and text reports. It does not load Makie.

For static PNG, PDF, and SVG plots, add CairoMakie to the environment where you added ScientificFitting, then load both:

using Pkg
Pkg.add("CairoMakie")

using ScientificFitting
using CairoMakie

Work From A Checkout

Instantiate the numerical core in the repository root:

cd /path/to/ScientificFitting
julia --project=. --startup-file=no -e 'using Pkg; Pkg.instantiate()'

Start a Julia session in that environment:

julia --project=.

Plotting and documentation dependencies live in a separate environment. Instantiate it when running the gallery or building the site:

julia --project=docs --startup-file=no -e 'using Pkg; Pkg.instantiate()'
julia --project=docs examples/gallery/01_quickstart_linear.jl

The example writes its figure to the ignored examples/output/ directory. The docs environment already provides CairoMakie; do not Pkg.add packages into the checkout's root project — it is the package itself.

First Use And Compilation

The first using ScientificFitting in a new environment compiles the numerical core. The first using CairoMakie and first rendered figure take longer because Julia also compiles Makie's layout, text, and rendering methods. Later sessions reuse the precompile cache unless Julia, package versions, preferences, or the target environment change.

Do not use the full package test suite to check an installation; it is a slow release gate. A core-only check is enough:

julia --project=. --startup-file=no -e 'using ScientificFitting; println("ScientificFitting core ready")'

Run this in the environment where you installed the package: from the checkout root as shown, or with --project pointing at your own project for a registry install.

For plotting, run the tracked quickstart example shown above; it confirms CairoMakie export.

Python Interface

In a Python 3.10+ virtual environment:

python -m pip install 'scientificfitting[plot]'

The Python interface uses NumPy models and optional native Matplotlib plots, without Makie. JuliaCall provisions the registered Julia core automatically on first use, which needs network access and compilation. Omit [plot] when no plots are needed. Supported features, examples, and the conda-forge submission status are on the Python Interface page.

Troubleshooting

symptom	first check
`using ScientificFitting` is slow once	Let precompilation finish; this is not fit runtime.
Every fresh session recompiles	Reuse the same project and depot; check whether Julia or package versions keep changing.
`plot_fit` says the extension is unavailable	Add and load `CairoMakie` before calling plotting functions.
PDF or SVG export fails	Verify a minimal CairoMakie figure in the same environment; inspect backend and font errors first.
A fit is unexpectedly slow	Check for dense covariance, bounds, constraints, priors, parameter-dependent covariance, or x uncertainties (`sigma_x`, `cov_x`), which re-evaluate the model slope at every data point in each iteration. These select more general numerical paths.
Package versions will not resolve	Confirm Julia is at least 1.10 and instantiate a clean environment rather than mixing incompatible manifests.

For bug reports, include a minimal example, package versions, and the full report_text(result) and diagnose(result) output.

Continue with the Quickstart. For package internals and scaling limits, see Backend Design, in particular its Performance Checks section.