signfetch 0.1.1a1

Download data resources discovered via FAIR Signposting item links.

signfetch v0.1

Python package focused on downloading data discovered via FAIR Signposting.

Scope

Version 0.1.0 intentionally focuses on one public API:

• download_data(target, output_dir=...)

The function discovers item links from the three main Signposting delivery mechanisms:

1. HTTP Link headers
2. HTML <link> elements
3. Link Sets discoverable with the linkset relation

It then deduplicates discovered data URLs and downloads all unique resources.

Install

pip install .

For development:

pip install -e .[dev]

Example

from pathlib import Path
from signfetch import download_data

result = download_data(
    "https://doi.org/10.5281/zenodo.1000000",
    output_dir=Path("downloads")
)

print(result.unique_item_count)
for item in result.items:
    print(item.url, item.saved_path)

Design notes

Main components:

• SignpostingHarvester discovers item and linkset links
• LinkHeaderParser, HtmlLinkParser, LinksetParser parse representations
• DataDownloader downloads discovered resources
• download_data() is the small public façade

Docker

Build:

docker build -t signfetch:0.1.0 .

Run tests in the image:

docker run --rm signfetch:0.1.0 pytest

Release flow

Use prerelease first (for example 0.1.1a1) and tag the commit:

git tag v0.1.1a1
git push origin v0.1.1a1

Run CI first. At the end of CI (package-artifact job) you will get the artifact signfetch-package. Use that CI run ID in the publish workflow input ci_run_id.

Validate package and tag without publishing:

bash scripts/publish_testpypi.sh check

Publish to TestPyPI:

bash scripts/publish_testpypi.sh test

Publish to PyPI:

bash scripts/publish_testpypi.sh prod

Recommendation: one artifact build on Linux is enough here because the package is pure Python (py3-none-any). Cross-platform safety is still covered by build checks and unit tests on Linux/Windows/macOS in CI.