GitLab

Added a performance optimization when syncing datasets with ADI.  Since Xarray always recreates coordinate variable data arrays when constructing the dataset, we now check if any values have changed in addition to whether the structure pointer has changed.

Fixed a bug caused by misunstanding - ADI calls post transform hook even if no transform is performed (I thought this wouldn't happen), so I had to move my debug dump code into the decorator class such that no dumps will occur if the hook function is empty.

Added a flag so users can build in dev mode with pip install -e option set.  Added some new methods for qc and units conversion.  Improving documentation.

Fixed the get dataset methods so that they automatically capture embedded metadata for the coordinate system name, datastream name, dataset type, and obs index that can be used when de-serializing in the sync_datasets method.  Changed sync_datasets to a single method that takes one or more xarray datasets. No
other parameters are now required as this information is contained in embedded metadata.

Added improvements such as accessor for debug_level and shortcut method for checking if multiple variables exist.

Added some more props to the process for site and facility.  Converted get_location into a prop.  Fixed xarray serialization/deserialization to convert ADI times to datetime64 objects.

Cleaned up the files so that the python classes in the new api are better organized and correspondingly changed the imports.

Stubbed out a bunch of adi wrapper functions in the Process class that will be useful for init_process_hook.  Added logic to initialize data array when adi variable has no data.  Fixed the dsproc.cds_type_to_dtype_obj method to be cpdef instead of cdef so we can call it from python.

Fixed some bugs in Process class.  Fixed build to use an editable pip install so we don't have to rebuild every time we edit a file.

Finished implemented to xarray.  Started from xarray.  Made more improvements to the process base class.