hdf5tools 0.5.0

Python hdf5 tools

This git repository contains a python package with an H5 class to load and combine one or more HDF5 data files (or xarray datasets) with optional filters. The class will then export the combined data to an HDF5 file, file object, or xr.Dataset. This class is designed to be fast and safe on memory. This means that files of any size can be combined and saved even on a PC with low memory (unlike xarray).

Installation

Using pip:

pip install hdf5tools

Or using conda/mamba from conda-forge:

conda install -c conda-forge hdf5tools

Usage

Currently, only the Combine class is recommended for other to use.

First, initiate the class with one or many: paths to netcdf3, netcdf4, or hdf5 files; xr.Dataset objects (opened via xr.open_dataset); or h5py.File objects.

from hdf5tools import Combine

###############################
### Parameters

path1 = '/path/to/file1.nc'
path2 = '/path/to/file2.nc'

##############################
### Combine files

c1 = Combine([path1, path2])

If you want to do some kind of selection via the coordinates or only select some of the data variables/coordinates then use the .sel method (like in xarray). Be sure to read the docstrings for additional info about the input parameters.

c2 = c1.sel({'time': slice('2001-01-01', '2020-01-01'), 'latitude': slice(0, 10)}, include_data_vars=['precipitation'])

And finally, save the combined data to a single hdf5/netcdf4 file using the .to_hdf5 method. The only additional parameters that are important include the output which should be a path or a io.Bytes object, and the compression parameter. If you plan on using this file outside of the python environment, use gzip for compression, otherwise use lzf. The docstrings have more details.

output = '/path/to/output.nc'

c2.to_hdf5(output, compression='gzip')

If you’ve passed xr.Dataset objects to Combine, it will be slower than passing the file as a path on disk. Only pass xr.Dataset objects to Combine if you don’t want to write the intermediate file to disk before reading it into Combine.

The package also comes with a bonus function called xr_to_hdf5. It is a convenience function to convert a single xr.Dataset to an hdf5/netcdf4 file.

from hdf5tools import xr_to_hdf5

xr_to_hdf5(xr_dataset, output, compression='gzip')