v0.1.3

Twenty-two commits of hardening since 0.1.2, driven by reading the project
the way a stranger would. Every change below was found by running the
documented commands rather than by a failing test — CI was green throughout.

Fixed

• pypic validate crashed on 2D data. It raised TypeError: max_div_b() missing 1 required positional argument: 'd3' on any 2D dataset — the shape
  of most BATSRUS z=0 output, and of three committed fixtures. It now
  reports max |div B|: skipped (needs a 3D grid) and completes.
• The BATSRUS .out read path was unreachable. available_fields()
  dispatched to the wrong parser and a binary .out raised
  UnicodeDecodeError; that path also skipped SI conversion and computed the
  wrong grid spacing. OpenGGCM's stagger metadata was corrected alongside it.
• The OpenGGCM quick-start did not run, and library code printed to
  stdout.
• Documented examples that raised when executed: plane="xy" passed
  where a PlaneSelection is required, a three-value unpack of a two-tuple
  return, Simulation.read(steps=...), and a "|u|" lookup for a quantity
  that is not registered.
• Wrong headline claims in the docs: PI step control was advertised for
  the adaptive tracer (the shipped controller is error-norm), register_field()
  was named as the compute() extension point (it is register_recipe()), and
  the unknown_simulation error kind is spelled unknown_sim on the wire.

Added

• Eleven names reachable straight from pypic — power_spectrum_1d/2d/3d,
  find_saddle_points, reconnection_rate, schindler_xi and friends — plus
  twenty-five more from pypic.derived and pypic.diagnostics that were
  public in their own modules but only reachable as pypic.derived.x.
• API reference pages for pypic.io, pypic.plotting, pypic.compute,
  pypic.comparison, pypic.regrid, pypic.exceptions and pypic.schema —
  seven public modules had none — plus a command-line reference covering
  every subcommand.
• A numbered example series in examples/, from arrays to a full
  simulation.toml, run by the test suite.
• SECURITY.md, a code of conduct, a pull-request template, and this
  changelog.

Changed

• Name-resolution failures raise UnknownFieldError instead of a bare
  KeyError in reduce, compute and field_info. It subclasses
  KeyError, so existing handlers are unaffected.
• FieldDataset.from_arrays accepts any Mapping, not only dict.
• Missing-extra errors come from one guard, so the message names the missing
  modules, the feature and the install command identically every time. Install
  hints now quote the extra — pip install "pypic-plasma[zarr]" — since
  unquoted brackets are a glob in zsh.

Removed

• The lazy extra. It installed dask, but nothing under src/ imported it;
  the chunked-loading path it was reserved for is still unwritten. It will
  come back when there is code behind it.

Full detail in CHANGELOG.md.

Install

uv add pypic-plasma                # library
uv add "pypic-plasma[plot,cli]"    # what most installs want

v0.1.2

Fixed

• pypic --version failed on a bare install. The pypic console
  script is registered unconditionally, but Typer and Rich ship in the
  cli extra, so pip install pypic-plasma followed by pypic --version raised ModuleNotFoundError: No module named 'typer'.
  The command now reports what to install and exits cleanly:

  $ pypic --version
  The 'pypic' command requires typer, which ships in the cli extra.
  Install with: pip install 'pypic-plasma[cli]'
  

  Only cli-extra imports are translated — any other import failure
  still propagates, so genuine breakage is not disguised as a missing
  extra. pypic.cli:app is unchanged and still importable directly.

Using pypic as a library was never affected; import pypic works on a
bare install and always did.

Install

uv add pypic-plasma          # library
uv add "pypic-plasma[cli]"   # library + the pypic command

Full test suite: 2764 passing. Documentation:
https://rusaitis.github.io/pypic/

v0.1.1

First release published to PyPI. Install with:

uv add pypic-plasma        # or: pip install pypic-plasma

The distribution is named pypic-plasma; the import name is pypic.

Fixed

• pypic --version crashed on a clean install. The CLI looked its
  version up as importlib.metadata.version("pypic"), but the
  distribution is pypic-plasma, so a fresh environment raised
  PackageNotFoundError. The server's /health probe had the same
  lookup behind a fallback and reported "unknown". Both now read
  pypic.__version__, which does not depend on the distribution name.

Internal

• New tests/test_packaging.py pins pypic.__version__ against
  pyproject.toml and rejects any distribution-metadata lookup that
  names something other than the declared distribution.
• CLI error assertions no longer depend on terminal colour. Typer
  force-enables rich colour under GITHUB_ACTIONS, and rich highlights
  option names, so Invalid --plane reached the assertion split across
  ANSI escapes. Thirteen tests had been failing in CI only.

Full test suite: 2761 passing. Documentation:
https://rusaitis.github.io/pypic/

Initial Release

Core is near complete, active development is in the applications.