Small lib for representing python objects as a dicts
Why would you need library like this? One of obvious use cases is to convert complex objects with methods, lots of atributes and so on into dicts for serializing (into json/yaml/xml/pickle/whatever). You can control dehydration process by describing how to fetch values from object and how to present it in dehydrated structure using simple syntax.
In simplest of possible cases you just want get object, list wanted attributes and get mapping with keys based on attribute names and values from them. Use dehydrate shortcut for this case:
>>> from dehydrate import dehydrate >>> from pretend import stub as Person >>> iron_man = Person(first_name='Tony', login='iron_man') >>> dehydrated = dehydrate(obj=iron_man, specs=('first_name', 'login')) >>> sorted(dehydrated.items()) [('first_name', 'Tony'), ('login', 'iron_man')]
- I use list representation of dict in examples because it has predictable order of items in it. It’s important, because this pieces of code are tests.
If requested attribute name resolves to method of object, then result of calling it will be set in dehydrated dict. In Person class we have method full_name, so let’s try to get its return value:
>>> from dehydrate import dehydrate >>> from pretend import stub as Person >>> iron_man = Person(full_name=lambda: 'Tony Stark') >>> dehydrated = dehydrate(obj=iron_man, specs=('full_name',)) >>> sorted(dehydrated.items()) [('full_name', 'Tony Stark')]
But what if you want put first_name attribute in name key of resulted dict? Just specify both strings in specs (spec can be one object or two-tuple):
>>> from dehydrate import dehydrate >>> from pretend import stub as Person >>> iron_man = Person(first_name='Tony', login='iron_man') >>> dehydrated = dehydrate(obj=iron_man, specs=( ... ('first_name', 'name'), ... 'login', ... )) >>> sorted(dehydrated.items()) [('login', 'iron_man'), ('name', 'Tony')]
Second argument always be used as a key if exists in spec.
More complex cases
Sometimes you will want to add some value in dehydrated dict, which is not attribute of dehydrated object. Or you may want not use attribute and add some another handling for this element instead. In our example we creating special class for this called PersonDehydrator (inherited from dehydrate.Dehydrator) and set some methods on it:
>>> from pretend import stub as Person >>> from examples import PersonDehydrator >>> iron_man = Person(password='iRon42', login='iron_man') >>> dehydrated = PersonDehydrator(specs=( ... 'password', ... ('superhero_status', 'is_superhero'), ... )).dehydrate(obj=iron_man) >>> sorted(dehydrated.items()) [('is_superhero', True), ('password', '******')]
In example you can see, that object has password attribute, but PersonDehydrator’s get_password used for password spec. Also you can mention, that result of calling get_superhero_status was set in key is_superhero because of second item in spec was declared. You can declare specs using attribute of dehydrator class or by passing argument into its __init__ method.
- In docs I will refer to examples package, which you can find in repo.
The most valuable feature of lib is that you can describe how to recursively dehydrate complex fields on object:
>>> from dehydrate import dehydrate, S >>> from pretend import stub as Person >>> from examples import PersonDehydrator >>> octopus = Person(login='octopus') >>> spider_man = Person(login='spidey', archenemy=octopus) >>> dehydrated = dehydrate( ... specs=( ... S('login'), ... S(target='archenemy', type='nested', specs=( ... S('login'), ... )), ... ), ... obj=spider_man ... ) >>> dehydrated['login'] 'spidey' >>> list(dehydrated['archenemy'].items()) [('login', 'octopus')]
You can see, that specs for nested elements are described with use of dehydrate.S shortcut (And simple specs as well for the sake of sanity). Acceptable arguments are for type='nested':
- target — name, that describes how to get value from object (or use hook on dehydrator)
- dehydrator — class, which can be used for dehydrating of complex target (dehydrate.Dehydrator by default).
- specs — iterable of same structure as described above (it is optional in case if you describe specs on dehydrator class, but make good sense, if you ant use default Dehydrator class).
pip install dehydrate
must be fine.
- six (did I mentioned python 3 support? We have one.)
- Easy things should be done easily.
- Complex things must be possible.
Test written with use of pytest library and neat pytest pep8 plugin. You should run python setup.py test for running full test suite or coverage run --source=dehydrate setup.py test for tests with coverage. Tests automatically runs at Travis CI. Examples in documentation are also picked by test command.
Any contribution is welcome. Use fork/pull request mechanism on github.
If you add some code, you should add some tests, so coverage of master branch should always be 100%. Refer to Testing section for more instructions.
Let me speak from my heart :). I will be very glad, if you correct my clumsy english phrases in docs and docstings or even advise more appropriate names for variables in code.
- Think about giving opportunity to put results in Ordered dict instead of simple dict.
- Add comprehensive docs about everything.
- Move complex examples with classes into docs from readme.
- Write docstrings and auto-generate some additional docs.
- New syntax for nested dehydration
- fields parameter renamed to specs
- improved README