Datary 0.2.1 — checked-in evidence integrity

Datary 0.2.1 — checked-in evidence integrity

Oh! The final post-publication check caught a platform detail worth fixing immediately. Python's CSV writer emitted CRLF on Windows, then Git normalized tracked text to LF. The example still parsed, but its data.csv bytes no longer matched the SHA-256 recorded before the commit.

I did not want to explain that away as “only an example.” An example session is evidence too. Datary 0.2.1 makes every CSV-producing path choose LF explicitly, regenerates the example, and adds a byte-level regression assertion. A fresh merged checkout now passes datary doctor and verifies the example session exactly.

What changed

• recorder, converter, and generator CSV output use an explicit LF line terminator;
• the checked-in example session and reports were regenerated with Datary 0.2.1;
• the recorder test now asserts that session data.csv contains no CRLF bytes;
• README, changelog, package version, and release narrative point to 0.2.1.

Verification

• 87 tests pass with 79.75% measured coverage and a 75% enforced floor;
• strict Mypy and Ruff pass;
• all 11 GitHub jobs passed across Python 3.9–3.14, Linux, Windows, and macOS;
• the clean wheel records a session whose CSV uses LF and whose integrity verifies;
• the merged repository example passes datary doctor;
• the merged tree exactly matches the tree that passed CI.

See PR #14, the changelog, and the focused release note.

Install

python -m pip install https://github.com/devkyato/datary-lab/releases/download/v0.2.1/datary_lab-0.2.1-py3-none-any.whl

Release artefacts

• datary_lab-0.2.1-py3-none-any.whl — SHA-256 c32ca01df83b9ef1356ae7ac9057107876e7a5bdca463db5c8c42a5d1f523a75
• datary_lab-0.2.1.tar.gz — SHA-256 226710bfbaa565c4ffe7fff06a9975a55f9abe0c12c178e3f869a056b586c8c7
• SHA256SUMS — machine-readable checksum list

This remains an alpha prerelease. All correctness, streaming, parser, security, and documentation work from 0.2.0 is unchanged; this patch closes the final checked-in evidence mismatch found by the release audit.