In order to dump a Quantity to disk, store it in a database or transmit it over the wire you need to be able to serialize and then deserialize the object.
The easiest way to do this is by converting the quantity to a string:
>>> import pint >>> ureg = pint.UnitRegistry() >>> duration = 24.2 * ureg.years >>> duration <Quantity(24.2, 'year')> >>> serialized = str(duration) >>> print(serialized) 24.2 year
Remember that you can easily control the number of digits in the representation as shown in String formatting.
You dump/store/transmit the content of serialized (‘24.2 year’). When you want to recover it in another process/machine, you just:
>>> import pint >>> ureg = pint.UnitRegistry() >>> duration = ureg('24.2 year') >>> print(duration) 24.2 year
Notice that the serialized quantity is likely to be parsed in another registry as shown in this example. Pint Quantities do not exist on their own but they are always related to a UnitRegistry. Everything will work as expected if both registries, are compatible (e.g. they were created using the same definition file). However, things could go wrong if the registries are incompatible. For example, year could not be defined in the target registry. Or what is even worse, it could be defined in a different way. Always have to keep in mind that the interpretation and conversion of Quantities are UnitRegistry dependent.
In certain cases, you want a binary representation of the data. Python’s standard algorithm for serialization is called Pickle. Pint quantities implement the magic __reduce__ method and therefore can be Pickled and Unpickled. However, you have to bear in mind, that the application registry is used for unpickling and this might be different from the one that was used during pickling.
By default, the application registry is one initialized with
other words, the same as what you get when creating a
arguments and without adding any definitions afterwards.
If your application is fine just using
defaults_en.txt, you don’t need to worry
If your application needs a single, global registry with custom definitions, you must
make sure that it is registered using
unpickling anything. You may use
pint.get_application_registry() to get the
current instance of the application registry.
>>> to_serialize = duration.to_tuple() >>> print(to_serialize) (24.2, (('year', 1),))
And then you can just pickle that:
>>> import pickle >>> serialized = pickle.dumps(to_serialize, -1)
To unpickle, just
>>> loaded = pickle.loads(serialized) >>> ureg.Quantity.from_tuple(loaded) <Quantity(24.2, 'year')>
(To pickle to and from a file just use the dump and load method as described in _Pickle)
You can use the same mechanism with any serialization protocol, not only with binary ones. (In fact, version 0 of the Pickle protocol is ASCII). Other common serialization protocols/packages are json, yaml, shelve, hdf5 (or via PyTables) and dill. Notice that not all of these packages will serialize properly the magnitude (which can be any numerical type such as numpy.ndarray).
Using the serialize package you can load and read from multiple formats:
>>> from serialize import dump, load, register_class >>> register_class(ureg.Quantity, ureg.Quantity.to_tuple, ureg.Quantity.from_tuple) >>> dump(duration, 'output.yaml') >>> r = load('output.yaml')
(Check out the serialize docs for more information)