What’s New in Iris 1.9

Release:1.9.2 Date:28th January 2016

This document explains the new/changed features of Iris in version 1.9 (View all changes.)

Iris 1.9 Features

• Support for running on Python 3.4 has been added to the whole code base. Some features which depend on external libraries will not be available until they also support Python 3, namely:

  • gribapi does not yet provide a Python 3 interface

• Added the UM pseudo level type to the information made available in the STASH_TRANS table in iris.fileformats.um._ff_cross_references

• When reading “cell_methods” attributes from NetCDF files, allow optional whitespace before the colon. This is not strictly in the CF spec, but is a common occurrence.

• Basic cube arithemetic (plus, minus, times, divide) now supports lazy evaluation.

• iris.analysis.cartography.rotate_winds() can now operate much faster on multi-layer (i.e. > 2-dimensional) cubes, as it calculates rotation coefficients only once and reuses them for additional layers.

• Linear regridding of a multi-layer (i.e. > 2-dimensional) cube is now much faster, as it calculates transform coefficients just once and reuses them for additional layers.

• Ensemble statistics can now be saved to GRIB2, using Product Definition Template 4.11.

• Loading of NetCDF data with ocean vertical coordinates now returns a ‘depth’ in addition to an ‘eta’ cube. This operates on specific defined dimensionless coordinates : see CF spec version 1.6, Appendix D.

• iris.analysis.stats.pearsonr() updates:

  • Cubes can now be different shapes, provided one is broadcastable to the other.
  • Accepts weights keyword for weighted correlations.
  • Accepts mdtol keyword for missing data tolerance level.
  • Accepts common_mask keyword for restricting calculation to unmasked pairs of cells.

• Added a new point-in-cell regridding scheme, iris.experimental.regrid.PointInCell.

• Added iris.analysis.WPERCENTILE() - a new weighted aggregator for calculating percentiles.

• Added cell-method translations for LBPROC=64 and 192 in UM files, encoding ‘zonal mean’ and ‘zonal+time mean’.

• Support for loading GRIB2 messages defined on a Lambert conformal grid has been added to the GRIB2 loader.

• Data on potential-temperature (theta) levels can now be saved to GRIB2, with a fixed surface type of 107.

• Added several new helper functions for file-save customisation, (see also : Saving Iris Cubes):

  • iris.fileformats.grib.as_pairs()
  • iris.fileformats.grib.as_messages()
  • iris.fileformats.grib.save_messages()
  • iris.fileformats.pp.as_pairs()
  • iris.fileformats.pp.as_fields()
  • iris.fileformats.pp.save_fields()

• Loading data from GRIB2 now supports most of the currently defined ‘data representation templates’ : code numbers 0, 1, 2, 3, 4, 40, 41, 50, 51 and 61.

• When a Fieldsfile is opened for update as a iris.experimental.um.FieldsFileVariant, unmodified packed data in the file can now be retained in the original form. Previously it could only be stored in an unpacked form.

• When reading and writing NetCDF data, the CF ‘flag’ attributes, “flag_masks”, “flag_meanings” and “flag_values” are now preserved through Iris load and save.

• mo_pack was added as an optional dependency. It is used to encode and decode data in WGDOS packed form.

• The iris.experimental.um.Field.get_data() method can now be used to read Fieldsfile data after the original iris.experimental.um.FieldsFileVariant has been closed.

Bugs Fixed

• Fixed a bug in iris.unit.Unit.convert() (and the equivalent in cf_units) so that it now converts data to the native endianness, without which udunits could not read it correctly.
• Fixed a bug with loading WGDOS packed data in iris.experimental.um, which could occasionally crash, with some data.
• Ignore non-numeric suffices in the numpy version string, which would otherwise crash some regridding routines.
• fixed a bug in iris.fileformats.um_cf_map where the standard name for the stash code m01s12i187 was incorrectly set, such that it is inconsistent with the stated unit of measure, ‘m s-1’. The different name, a long_name of ‘change_over_time_in_upward_air_velocity_due_to_advection’ with units of ‘m s-1’ is now used instead.
• Fixed a bug in iris.cube.Cube.intersection(). When edge points were at (base + period), intersection would unnecessarily wrap the data.
• Fixed a bug in iris.fileformats.pp. A previous release removed the ability to pass a partial constraint on STASH attribute.
• iris.plot.default_projection_extent() now correctly raises an exception if a cube has X bounds but no Y bounds, or vice versa. Previously it never failed this, as the test was wrong.
• When loading NetCDF data, a “units” attribute containing unicode characters is now transformed by backslash-replacement. Previously this caused a crash. Note: unicode units are not supported in the CF conventions.
• When saving to NetCDF, factory-derived auxiliary coordinates are now correctly saved with different names when they are not identical. Previously, such coordinates could be saved with the same name, leading to errors.
• Fixed a bug in iris.experimental.um.FieldsFileVariant.close(), which now correctly allocates extra blocks for larger lookups when saving. Previously, when larger files open for update were closed, they could be written out with data overlapping the lookup table.
• Fixed a bug in iris.aux_factory.OceanSigmaZFactory which sometimes caused crashes when fetching the points of an “ocean sigma z” coordinate.

Version 1.9.1

• Fixed a unicode bug preventing standard names from being built cleanly when installing in Python3

Version 1.9.2

• New warning regarding data loss if writing to an open file which is also open to read, with lazy data.
• Removal of a warning about data payload loading from concatenate.
• Updates to concatenate documentation.
• Fixed a bug with a name change in the netcdf4-python package.
• Fixed a bug building the documentation examples.
• Fixed a bug avoiding sorting classes directly when iris.cube.Cube.coord_system() is used in Python3.
• Fixed a bug regarding unsuccessful dot import.

Incompatible Changes

• GRIB message/file reading and writing may not be available for Python 3 due to GRIB API limitations.

Deprecations

• Deprecated iris.unit, with unit functionality provided by cf_units instead.
• When loading from NetCDF, a deprecation warning is emitted if there is vertical coordinate information that would produce extra result cubes if iris.FUTURE.netcdf_promote were set, but it is not set.
• Deprecated iris.aux_factory.LazyArray

Documentation Changes

• A chapter on saving iris cubes has been added to the user guide.
• Added script and documentation for building a what’s new page from developer-submitted contributions. See Contributing a “What’s New” entry.