v1.3.11

This release make improvements to the nc_plot method. This now covers a larger number of grids, and has been tested to plot all CMIP6 native grids, and likely covers >90% of netCF grids.

v1.3.10

This release makes some improvements to nctoolkit internals to reduce the likelihood of processes not being fully terminated when nctoolkit crashes. No changes to methods etc. have been made.

v1.3.9

This release makes some improvements to NCToolkit internals to make the package compatible with recent conda forge releases of Climate Data Operators.

v1.3.8

This release makes nctoolkit compatible with macOS once again.

No other changes apply.

v1.3.7

This release makes some improvements to nctooolkit internals. An issue was causing nctoolkit to stall occassionaly when it was called within a parallel processing chain. This release should deal with that problem.

v1.3.6

This is a minor release that fixes issues with the setup files for pip. No changes have been made to the nctoolkit code base.

v1.3.5

This is a minor release that makes some improvements to nctoolkit internals. Some improvements have been made that will reduce the occurrence of temporary files hanging around longer than is needed.

v1.3.4

v1.3.4

This releases makes some internal improvements for nctoolkit to ensure users can switch cleanly between parallel and non-parallel model.

Fixed

• Fixed safe-list handling when switching parallel mode: items were being
  removed from a list while looping over it, which skipped every other file.
  Whole lists are now moved instead. In annual_anomaly, this also removed a
  workaround loop that is no longer needed, stops the pool before resetting
  on failure, keeps results in input-file order rather than completion
  order, and re-raises all worker errors instead of swallowing some of them.
• get_protected() returned the live, shared parallel-mode file list rather
  than a copy, so callers could accidentally mutate nctoolkit's internal
  bookkeeping. It now returns a snapshot, matching get_safe() and
  get_tempdirs().
• Iterating the shared warnings/safe-file lists used in parallel mode relied
  on a manager-proxy IndexError to signal the end of the list. On Python
  3.12, that exception could form a reference cycle that kept the calling
  frame — and so the DataSet it referenced — alive longer than
  expected, delaying cleanup. A new internal get_warnings() helper, and
  slicing rather than iterating elsewhere, avoid this.
• generate_grid(): the regular-grid check mixed is False with |, which
  parsed as a chained comparison and only triggered when both longitude and
  latitude were irregular. Grids irregular in a single dimension were
  wrongly written out as lon/lat grids; both dimensions are now checked
  independently.
• plot(): a single-file DataSet is now passed to ncplot as a single
  path rather than a one-item list, for compatibility with ncplot
  versions that cannot handle a list there.

Dependencies

• Requires ncplot>=0.4.0 (previously >=0.2.8).

v1.3.3

v1.3.3

New

• DataSet.to_transect() — extracts a straight-line transect between
  two points. It places nsteps evenly spaced points (including both ends)
  along the line between start and end, given as [lon, lat], in lon/lat
  space, and regrids the dataset onto them:

  ds.to_transect(start=[-30, 0], end=[-30, 65], nsteps=66)

  The remapping method defaults to bilinear ("bil") and accepts any of the
  regrid methods ("nn", "bic", "dis", "con", "con2", "laf").
  Invalid start, end or nsteps values raise a clear TypeError or
  ValueError.

Documentation

• New Gallery section on the website, with worked examples:
  • Working with transects — extract a transect through the Atlantic
    with to_transect, then map it and chart it against latitude and time.
  • Nitrate from the World Ocean Atlas — a transect down the Atlantic
    and a single profile.
• to_transect is documented in the API reference, and the cheat sheet has a
  new how-to entry for extracting a straight-line transect.
• Each pub_plot option in the gallery now shows its accepted values on click.

v1.3.2

v1.3.2

New

• nc.panel_plot() — draws a grid of pub_plot-style maps from a
  dictionary of {title: dataset}. Panels share one colour scale and one
  colour bar by default, worked out from the data of every panel together, so
  they can be compared directly. Set shared_colourbar=False to give each
  panel its own colour scale and colour bar instead. The grid layout
  (nrow/ncol) is chosen automatically if not given.
• Any panel_plot argument — including the ones it shares with
  pub_plot, such as var, extent, land, colours and limits —
  can be a single value applied to every panel, or a list with one value per
  panel. Colour-bar arguments (colours, norm, limits, robust,
  legend, legend_position, breaks, font) must stay single values when
  shared_colourbar=True.
• pub_plot(grid_labels=False) hides the lon/lat labels around the map,
  independently of grid, which now only controls the grid lines.

Fixed

pub_plot:

• dpi was silently ignored; saved figures always used matplotlib's default
  resolution rather than the requested one.
• coast=None raised a ValueError instead of turning coastlines off.
• coast="coarse"/"low"/"intermediate"/"high"/"full" all produced the
  same coastline resolution.
• A user-supplied projection was silently overridden with Robinson whenever
  the data covered the whole globe.
• A custom legend was placed beside the colour bar instead of underneath it
  when legend_position="bottom".

panel_plot:

• out="~/some/path.png" failed with FileNotFoundError; ~ is now
  expanded to the home directory, matching to_nc.

Documentation

• Every pub_plot parameter is now documented in detail, both in the
  function's docstring and on the website: accepted formats, valid options
  and default behaviour, including previously undocumented arguments
  (grid_colour, font, fig/gs, relief).
• panel_plot is documented on the API reference and on the visualization
  guide, with worked examples (a shared seasonal climatology, per-panel
  colour bars, and a fine-tuning example varying land, grid, colours
  and limits per panel) and a gallery of pub_plot options.