Installation¶

fluxatlas needs Python 3.12 or 3.13. The upper bound is a pin taken from diive; a release will be needed to lift it for 3.14.

From source¶

Until v0.1.0 reaches PyPI, which still holds only the 0.0.1 name placeholder, install from a checkout.

git clone https://github.com/holukas/fluxatlas.git
cd fluxatlas
uv sync

uv sync builds the environment from uv.lock and installs the package in editable form, together with the dev group. The command line is then:

uv run fluxatlas --help

With pip instead:

python -m pip install -e .

Dependencies¶

Package	What it is for
`pandas` >= 2.2, `numpy` >= 1.26	Reading, reindexing and aggregating the half-hourly record.
`scipy` >= 1.11	Theil-Sen slopes and Kendall’s tau, behind every trend the page publishes.
`pyarrow` >= 15.0	Reading `.parquet` input, and the faster CSV reader used for the projected read.

The built page needs nothing. calendar.js, calendar.css and the mark are inlined into the output, so the file opens in any browser without a network.

The documentation¶

The documentation builds from the same checkout. Its dependencies are the docs group:

uv sync --extra docs
uv run sphinx-build -b html docs docs/_build/html

The package has to be importable: the API reference comes from the docstrings, and the variable tables are generated from the registry. uv sync installs it, so one command covers both.

Checking the installation¶

uv run pytest

About 183 tests, roughly a minute. Most of them run on synthetic data built in tests/conftest.py, so a fresh checkout passes with no data file present. The tests that read the bundled CH-LAE extract skip when it is missing.