[cfbastarz]

Motivation

• Remove multiple robustness and usability issues found while reading .gcv matrices and when producing plots, and make packaging/CI sane for modern Python tooling.
• Ensure plotting helpers behave predictably with unit conversions and that human-readable matrix names can be changed without corrupting the API.

Description

• Improve the GCV reader in gsiberror/__init__.py to validate header/record sizes, reject truncated or invalid files, detect Git LFS pointer files with an actionable message, and produce a non-duplicated cyclic longitude axis (endpoint=False for longitudes).
• Fix Berror.my_name so it no longer overwrites itself and provide get_name to return the configured display name, defaulting to the filename.
• Harden plotting utilities in gsiberror/plot_functions.py by validating inputs, raising descriptive ValueError for unknown variables, applying global min/max after unit conversions (e.g. pressure and humidity scalings), and returning the created matplotlib figure objects.
• Clean up setup.py and environment.yml metadata and dependency constraints (project_urls, extras_require, relax Cartopy pin, align numpy/python requirements).
• Add regression tests tests/test_berror.py and tests/test_plot_functions.py covering LFS pointer detection, truncated files, name-setting, var_name validation, and global_minmax scaling behaviour.
• Add a CI job in .github/workflows/ci.yml to run lint/format checks and pytest on pull requests and limit docs deployment to pushes on main.

Testing

• Ran PYTHONWARNINGS=error python -m compileall -q gsiberror tests setup.py which completed without warnings.
• Ran ruff check . --exclude notebooks and ruff format --check gsiberror tests setup.py which passed after automatic fixes.
• Ran git diff --check and static checks to ensure formatting and no leftover issues, which passed.
• Ran pytest -q in the local image but collection failed due to missing runtime dependencies (ModuleNotFoundError: No module named 'numpy' and inability to import the package because the test runner environment did not have the package installed), so unit tests could not be executed locally; the CI job added to the workflow installs pip install -e ".[dev]" before running tests to ensure dependencies are available.
• Attempts to install missing packages (pip install) and to create a remote issue via gh failed in this environment because network/proxy and GitHub authentication are not available, and python setup.py --name could not run here because setuptools is not present in the image.

---

Codex Task

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: de8f26577b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Point the pip cache at an existing dependency file

In the added test job, actions/setup-python@v5 enables the pip cache without cache-dependency-path, so it searches its default dependency manifests (requirements.txt or pyproject.toml). This commit contains neither file—the dependencies are declared in setup.py—causing setup-python to stop before the install, lint, and test steps. Set cache-dependency-path to setup.py/environment.yml or disable caching.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Create an LFS pointer fixture instead of using tracked data

When tests run in a normal clone where Git LFS smudged the sample—or after following the reader's own git lfs pull advice—SAMPLE is real GCV data, so read_records() does not raise the expected pointer error and this test fails. The current CI hides the problem with lfs: false; write a pointer into tmp_path so the test remains deterministic regardless of checkout state.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: de8f26577b

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Point the pip cache at an existing dependency file

In the added test job, actions/setup-python@v5 enables the pip cache without cache-dependency-path, so it searches its default dependency manifests (requirements.txt or pyproject.toml). This commit contains neither file—the dependencies are declared in setup.py—causing setup-python to stop before the install, lint, and test steps. Set cache-dependency-path to setup.py/environment.yml or disable caching.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Create an LFS pointer fixture instead of using tracked data

When tests run in a normal clone where Git LFS smudged the sample—or after following the reader's own git lfs pull advice—SAMPLE is real GCV data, so read_records() does not raise the expected pointer error and this test fails. The current CI hides the problem with lfs: false; write a pointer into tmp_path so the test remains deterministic regardless of checkout state.

Useful? React with 👍 / 👎.