What’s new in Iris 1.8

Release:1.8.1 Date:3rd June 2015

This document explains the new/changed features of Iris in version 1.8. (View all changes.)

Iris 1.8 features

Showcase: Rotate winds

Iris can now rotate and unrotate wind vector data by transforming the wind vector data to another coordinate system.

For example:

>>> from iris.analysis.cartography import rotate_winds
>>> u_cube = iris.load_cube('my_rotated_u_wind_cube.pp')
>>> v_cube = iris.load_cube('my_rotated_v_wind_cube.pp')
>>> target_cs = iris.coord_systems.GeogCS(6371229.0)
>>> u_prime, v_prime = rotate_winds(u_cube, v_cube, target_cs)

Showcase: Nearest-neighbour scheme

A nearest-neighbour scheme for interpolation and regridding has been added to Iris. This joins the existing Linear and AreaWeighted interpolation and regridding schemes.

For example:

>>> result = cube.interpolate(sample_points, iris.analysis.Nearest())
>>> regridded_cube = cube.regrid(target_grid, iris.analysis.Nearest())

Showcase: Slices over a coordinate

You can slice over one or more dimensions of a cube using iris.cube.Cube.slices_over(). This provides similar functionality to slices() but with almost the opposite outcome.

Using slices() to slice a cube on a selected dimension returns all possible slices of the cube with the selected dimension retaining its dimensionality. Using slices_over() to slice a cube on a selected dimension returns all possible slices of the cube over the selected dimension.

To demonstrate this:

>>> cube = iris.load(iris.sample_data_path('colpex.pp'))[0]
>>> print(cube.summary(shorten=True))
air_potential_temperature / (K)     (time: 6; model_level_number: 10; grid_latitude: 83; grid_longitude: 83)
>>> my_slice = next(cube.slices('time'))
>>> my_slice_over = next(cube.slices_over('time'))
>>> print(my_slice.summary(shorten=True))
air_potential_temperature / (K)     (time: 6)
>>> print(my_slice_over.summary(shorten=True))
air_potential_temperature / (K)     (model_level_number: 10; grid_latitude: 83; grid_longitude: 83)

• iris.cube.CubeList.concatenate() now works with biggus arrays and so now supports concatenation of cubes with deferred data.
• Improvements to NetCDF saving through using biggus:

• A cube’s lazy data payload will still be lazy after saving; the data will not be loaded into memory by the save operation.
• Cubes with data payloads larger than system memory can now be saved to NetCDF through biggus streaming the data to disk.

• iris.util.demote_dim_coord_to_aux_coord() and iris.util.promote_aux_coord_to_dim_coord() allow a coordinate to be easily demoted or promoted within a cube.
• iris.util.squeeze() removes all length 1 dimensions from a cube, and demotes any associated squeeze dimension DimCoord to be a scalar coordinate.
• iris.cube.Cube.slices_over(), which returns an iterator of all sub-cubes along a given coordinate or dimension index.
• iris.cube.Cube.interpolate() now accepts datetime.datetime and netcdftime.datetime instances for date or time coordinates.
• Many new and updated translations between CF spec and STASH codes or GRIB2 parameter codes.
• PP/FF loader creates a height coordinate at 1.5m or 10m for certain relevant stash codes.
• Lazy aggregator support for the standard deviation aggregator has been added.
• A speed improvement in calculation of iris.analysis.cartography.area_weights().
• Experimental support for unstructured grids has been added with iris.experimental.ugrid(). This has been implemented using UGRID.
• iris.cube.CubeList.extract_overlapping() supports extraction of cubes over regions where common coordinates overlap, over multiple coordinates.
• Warnings raised due to invalid units in loaded data have been suppressed.
• Experimental low-level read and write access for FieldsFile variants is now supported via iris.experimental.um.FieldsFileVariant.
• PP loader will return cubes for all fields prior to a field with a problematic header before raising an exception.
• NetCDF loader skips invalid global attributes, raising a warning rather than raising an exception.
• A warning is now raised rather than an exception when constructing an AuxCoordFactory fails.
• Supported aux coordinate factories have been extended to include:

• ocean sigma coordinate,
• ocean s coordinate,
• ocean s coordinate, generic form 1, and
• ocean s coordinate, generic form 2.

• iris.cube.Cube.intersection() now supports taking a points-only intersection. Any bounds on intersected coordinates are ignored but retained.
• The FF loader’s known handled grids now includes Grid 21.
• A nearest neighbour scheme is now provided for iris.cube.Cube.interpolate() and iris.cube.Cube.regrid().
• iris.analysis.cartography.rotate_winds() supports transformation of wind vectors to a different coordinate system.
• NumPy universal functions can now be applied to cubes using iris.analysis.maths.apply_ufunc().
• Generic functions can be applied to Cube instances using iris.analysis.maths.IFunc.
• The iris.analysis.Linear scheme now supports regridding as well as interpolation. This enables iris.cube.Cube.regrid() to perform bilinear regridding, which now replaces the experimental routine “iris.experimental.regrid.regrid_bilinear_rectilinear_src_and_grid”.

Bugs fixed

1.8.0

• Fix in netCDF loader to correctly determine whether the longitude coordinate (including scalar coordinates) is circular.
• iris.cube.Cube.intersection() now supports bounds that extend slightly beyond 360 degrees.
• Lateral Boundary Condition (LBC) type FieldFiles are now handled correctly by the FF loader.
• Making a copy of a scalar cube with no data now correctly copies the data array.
• Height coordinates in NAME trajectory output files have been changed to match other NAME output file formats.
• Fixed datatype when loading an integer_constants array from a FieldsFile.
• FF/PP loader adds appropriate cell methods for lbtim.ib = 3 intervals.
• An exception is raised if the units of the latitude and longitude coordinates of the cube passed into iris.analysis.cartography.area_weights() are not convertible to radians.
• GRIB1 loader now creates a time coordinate for a time range indicator of 2.
• NetCDF loader now loads units that are empty strings as dimensionless.

1.8.1

• The PP loader now carefully handles floating point errors in date time conversions to hours.
• The handling fill values for lazy data loaded from NetCDF files is altered, such that the _FillValue set in the file is preserved through lazy operations.
• The risk that cube intersections could return incorrect results due to floating point tolerances is reduced.
• The new GRIB2 loading code is altered to enable the loading of various data representation templates; the data value unpacking is handled by the GRIB API.
• Saving cube collections to NetCDF, where multiple similar aux-factories exist within the cubes, is now carefully handled such that extra file variables are created where required in some cases.

1.8.2

• A fix to prevent the error: AttributeError: ‘module’ object has no attribute ‘date2num’. This was caused by the function netcdftime.date2num() being removed from the netCDF4 package in recent versions.

Deprecations

• The original GRIB loader has been deprecated and replaced with a new template-based GRIB loader.
• Deprecated default NetCDF save behaviour of assigning the outermost dimension to be unlimited. Switch to the new behaviour with no auto assignment by setting iris.FUTURE.netcdf_no_unlimited to True.
• The former experimental method “iris.experimental.regrid.regrid_bilinear_rectilinear_src_and_grid” has been removed, as iris.analysis.Linear now includes this functionality.

Documentation Changes

• A chapter on merge and concatenate has been added to the user guide.
• A section on installing Iris using conda has been added to the install guide.
• Updates to the chapter on regridding and interpolation have been added to the user guide.