Skip to content

JNevrly/cascadict

Repository files navigation

CascaDict - cascading dictionary for Python

This little class aims to solve just another almost nonexistent problem - handling of cascading properties. To describe it simply, CascaDict implements something like class inheritance, but on dictionary-key level. See the examples:

from cascadict import CascaDict

fruit_properties = CascaDict({'type':'fruit', 'taste':'sweet', 'color':"I don't have any color, I'm an abstract concept!"})
blueberry = fruit_properties.cascade({'name': 'blueberry', 'color':'blue'})

print(blueberry['color'])
print(blueberry['taste'])

blue
sweet

Internally, it's not just copy and append, one can also access all the cascaded values:

print(blueberry.get_cascaded('color'))

['blue', "I don't have any color, I'm an abstract concept!"]

Cascaded ancestors are referenced, not copied:

fruit_properties['taste'] = 'bittersweet'
blueberry['taste']

'bittersweet'

CascaDicts can be nested, and any dict element put into CascaDict is also nested as CascaDict:

fruit_properties['classification'] = {'kingdom': 'Plantae',}
blueberry['classification'] = {'Order': 'Ericales', 'Family': 'Ericaceae', 'Genus': 'Vaccinium', 'Section': 'Cyanococcus'}

blueberry['classification']['kingdom']

'Plantae'

CascaDicts are of course iterable...

for key, value in blueberry.items():
	print(key,value)

('color', 'blue')
('name', 'blueberry')
('classification', <{'Section': 'Cyanococcus', 'Genus': 'Vaccinium', 'Order': 'Ericales', 'Family': 'Ericaceae'}, Ancestor: <{}, Ancestor: <{'kingdom': 'Plantae'}, Ancestor: None>>>)
('taste', 'bittersweet')
('type', 'fruit')

... and picklable

import pickle
blueberry_jam = pickle.loads(pickle.dumps(blueberry))
for key, value in blueberry_jam.items():
	print(key,value)

('color', 'blue')
('name', 'blueberry')
('classification', <{'Section': 'Cyanococcus', 'Genus': 'Vaccinium', 'Order': 'Ericales', 'Family': 'Ericaceae'}, Ancestor: <{}, Ancestor: <{'kingdom': 'Plantae'}, Ancestor: None>>>)
('taste', 'bittersweet')
('type', 'fruit')

If needed, CascaDict can be "flattened" into normal (nested) dict:

blueberry.copy_flat()

{'classification': {'Family': 'Ericaceae',
  'Genus': 'Vaccinium',
  'Order': 'Ericales',
  'Section': 'Cyanococcus',
  'kingdom': 'Plantae'},
 'color': 'blue',
 'name': 'blueberry',
 'taste': 'bittersweet',
 'type': 'fruit'}

Or only the top (final) level of CascaDict, without any ancestor properties, can be copied:

blueberry.copy_flat(level='skim')

{'classification': {'Family': 'Ericaceae',
  'Genus': 'Vaccinium',
  'Order': 'Ericales',
  'Section': 'Cyanococcus'},
 'color': 'blue',
 'name': 'blueberry'}

Combined with (e.g.) yaml, it makes any configuration processing a breeze:

import yaml

config = '''
defaults:
	port: 5556
	login_required: True
	logging: 
		level: DEBUG
		handler: stream
		
process_1:
	max_runtime: 100
	login_required: False
	logging:
		handler: file
	
process_2:
	port: 6005
	halt_on_error: True
	logging:
		level: ERROR

'''

raw_config = yaml.load(config)
defaults = CascaDict(raw_config.pop('defaults'))
properties = {} #no dict comprehension, remember Python 2.7 folk
for k,v in raw_config.items():
	properties[k] = CascaDict(v, ancestor=defaults)
	
for k,v in properties.items():
	print("{0}: {1}".format(k, v.copy_flat()))

process_2: {'login_required': True, 'logging': {'handler': 'stream', 'level': 'ERROR'}, 'port': 6005, 'halt_on_error': True}
process_1: {'logging': {'handler': 'file', 'level': 'DEBUG'}, 'login_required': False, 'max_runtime': 100, 'port': 5556}

See full documentation on ReadTheDocs.

That's it

This whole thing is just one small file, works in both Python 2.7 and 3.x and is released under MIT License. Now, cascade!