2.22.2

📅 2021-10-20

• Fixed backward compatibility issues in Python interface: fixed reading of
  strings and characters; RecordType.__getattr__() now raises AttributeError
  if attribute does not exist; fixed potential ingestion problems linked to
  empty arrays and/or unavailable record fields.

• Improved CMake build system.

2.22.1

📅 2021-06-07

• Fixed memory issue in HDF4 backend when data would have a non-default
  scale_factor, add_offset, or _FillValue attribute.

• Fixed backward compatibility issue in the new Python interface
  (Record.__dict__ is now available again).

2.22

📅 2021-04-22

WARNING: Make sure to remove existing coda python installations before
upgrading (e.g. remove /usr/local/lib/python<x.y>/site-packages/coda).

• Major refactoring of the CODA Python interface:

  • changed implementation from swig to cffi (allows use of coda with pypy)
  • improved performance for coda.fetch() (greatest speedup when using pypy)
  • added new Object Oriented interface
    This new Python interface of CODA is fully backward compatible.

• Greatly improved performance of codacheck on HDF5/netCDF4 products.

• Allow xml attributes to be interpreted as any ascii datatype via a
  codadef definition (instead of only allowing it to be text).

2.21.2

📅 2020-10-27

• Fixed potential memory issues and improved overall code stability.

2.21.1

📅 2020-03-31

• Fixed issue in name mangling of embedded expat/zlib libraries.

2.21

📅 2019-04-17

• Default CODA definition path for CODA command line tools on Windows is
  now also ../share/coda/definitions instead of ../definitions.

• CMake build now also creates share/coda/definitions directory when
  installing CODA.

• Fixed problem in matlab mex extension determination on linux when using
  the configure script.

• Fixed wrong installation location of coda.jar on Windows.

• Fixed issue where CODA could end up in an infinite loop on Windows
  when trying to traverse the coda definition path.

• Fixed setting of default CODA definition path for Python on Windows.

• Fixed issue where tests on attributes were not evaluated before tests on
  sub-paths in codadef detection rules (important for e.g. xmlns tests).

• Allow building the MATLAB interface using MATLAB R2018a or higher
  (similar to using mex with the -R2017b option).

• Updated embedded PCRE library to version 8.43.

• Updated embedded expat library to version 2.2.6.

• Updated embedded zlib library to version 1.2.11.

2.20

📅 2018-12-17

• The CODA IDL DLM files have been renamed from coda-idl.* to coda_idl.*
  (to better support CMake builds on Windows). Remove any old coda-idl.*
  files from the installation directory to prevent errors.

• The IDL, MATLAB, Python, and Java interface modules now link against the
  shared CODA library instead of the static version. The modules use an
  rpath setting based on the installation prefix to find the library.

• Removed the libcoda_internal.la target from the autotools build.
  Builds that used SUBPACKAGE_MODE should now use libcoda.la.

• The build system for Windows is now fully based on CMake and the binary
  installer is now created using WIX.

• Default CODA definition path for Python on Windows is now determined
  relative to coda.dll library location (../share/coda/definitions).

• The CODA C library now uses thread-local storage (TLS) variables
  internally to allow CODA to be used in multi-threaded applications.
  You will still have to call coda_init() separately for each thread and
  can't reuse products/cursors/etc. from one thread in another.

• Added -p/--path option to codacmp.

• Fixed issue in generated documentation of detection expressions.

• Fixed memory issue when trying to access (non-existing) attributes of an
  HDF5 dataset element.

• Fixed build system to work with automake 1.16.

• Updated SWIG generated files (for Java/Python interfaces) to SWIG 3.0.12.

• Fixed round() coda expression function for negative values
  (round(-0.6) was returning -2 instead of -1).

• Updated build system to link against HDF4 installations that use libhdf
  instead of libdf as library name.

2.19

📅 2018-03-28

• Added support for netCDF CDF-5 (NC_64BIT_DATA).

• Fixed issue where filename() expression would provide full filename path
  on Windows (instead of just the filename). This impacted product type
  detection on Windows when the detection expression was using the filename.

• Fixed issue reading variable length strings from a scalar dataset in HDF5.

• Added support for empty string attributes in HDF5 (i.e. using H5S_NULL).

• Allow min() and max() coda expression functions to work on arrays and on
  (arrays of) string values.

2.18.3

📅 2017-09-22

• The CODA Python interface now automatically sets a default CODA_DEFINITION
  path relative to the CODA Python package.

• Fixed memory leak in HDF5 backend.

• Allow fixed length array dimension definitions for XML formatted products.

2.18.2

📅 2017-06-20

• Added automatic scale_factor/add_offset/missing_value/_FillValue conversion
  for HDF5 (similar to what was already done for netcdf and HDF4).

• Leading and trailing whitespace that is allowed when parsing numerical
  values from ascii data can now consist of either space or a tab characters
  (instead of just space characters).

• Improved reading performance of GRIB messages that use 'simple packing'.