Install¶

From PyPI (recommended)¶

pip install calcfidata

Requires Python 3.9+. Installs pandas>=2.0 and requests>=2.28 as dependencies.

From Anaconda.org¶

conda install -c jeresalmisto calcfidata

Mirrors the same wheel as PyPI.

From source¶

The source lives in two synchronized mirrors. Either works:

GitLab (canonical)GitHub mirror

git clone https://gitlab.com/jere.salmisto/calcfi-open-data.git
cd calcfi-open-data/python
pip install -e .

git clone https://github.com/jerehere/calcfi-open-data.git
cd calcfi-open-data/python
pip install -e .

Verify¶

import calcfidata as cf
print(cf.__version__)
print(cf.list_series().shape[0])  # 34

Optional: dev tooling¶

For contributing:

pip install calcfidata[dev]

This pulls in pytest, build, and twine.

Data location¶

The package does not ship the data. It lazy-fetches each series from the Hugging Face dataset mirror on first use, then caches in-process for the session. So you only pay the network cost for series you actually touch.

Cache key is the slug — same slug twice in the same process = one HTTP call.

If you need the raw CSVs offline, download from any of the permanent DOI mirrors.