This module is an example implementation of an ordered dict for the collections module. It’s not written for performance (it actually performs pretty bad) but to show how the API works.
Why would anyone need ordered dicts?
Dicts in python are unordered which means that the order of items when iterating over dicts is undefined. As a matter of fact it is most of the time useless and differs from implementation to implementation.
Many developers stumble upon that problem sooner or later when comparing the output of doctests which often does not match the order the developer thought it would.
Also XML systems such as Genshi have their problems with unordered dicts as the input and output ordering of tag attributes is often mixed up because the ordering is lost when converting the data into a dict. Switching to lists is often not possible because the complexity of a lookup is too high.
Another very common case is metaprogramming. The default namespace of a class in python is a dict. With Python 3 it becomes possible to replace it with a different object which could be an ordered dict. Django is already doing something similar with a hack that assigns numbers to some descriptors initialized in the class body of a specific subclass to restore the ordering after class creation.
When porting code from programming languages such as PHP and Ruby where the item-order in a dict is guaranteed it’s also a great help to have an equivalent data structure in Python to ease the transition.
Where are new keys added?
At the end. This behavior is consistent with Ruby 1.9 Hashmaps and PHP Arrays. It also matches what common ordered dict implementations do currently.
What happens if an existing key is reassigned?
The key is not moved. This is consitent with existing implementations and can be changed by a subclass very easily:
class movingodict(odict): def __setitem__(self, key, value): self.pop(key, None) odict.__setitem__(self, key, value)Moving keys to the end of a ordered dict on reassignment is not very useful for most applications.
Does it mean the dict keys are sorted by a sort expression?
That’s not the case. The odict only guarantees that there is an order and that newly inserted keys are inserted at the end of the dict. If you want to sort it you can do so, but newly added keys are again added at the end of the dict.
I initializes the odict with a dict literal but the keys are not ordered like they should!
Dict literals in Python generate dict objects and as such the order of their items is not guaranteed. Before they are passed to the odict constructor they are already unordered.
What happens if keys appear multiple times in the list passed to the constructor?
The same as for the dict. The latter item overrides the former. This has the side-effect that the position of the first key is used because the key is actually overwritten:
>>> odict([('a', 1), ('b', 2), ('a', 3)]) odict.odict([('a', 3), ('b', 2)])This behavor is consistent with existing implementation in Python and the PHP array and the hashmap in Ruby 1.9.
This odict doesn’t scale!
Yes it doesn’t. The delitem operation is O(n). This is file is a mockup of a real odict that could be implemented for collections based on an linked list.
Why is there no .insert()?
There are few situations where you really want to insert a key at an specified index. To now make the API too complex the proposed solution for this situation is creating a list of items, manipulating that and converting it back into an odict:
>>> d = odict([('a', 42), ('b', 23), ('c', 19)]) >>> l = d.items() >>> l.insert(1, ('x', 0)) >>> odict(l) odict.odict([('a', 42), ('x', 0), ('b', 23), ('c', 19)])
Ordered dict example implementation.
This is the proposed interface for a an ordered dict as proposed on the Python mailinglist (proposal).
It’s a dict subclass and provides some list functions. The implementation of this class is inspired by the implementation of Babel but incorporates some ideas from the ordereddict and Django’s ordered dict.
The constructor and update() both accept iterables of tuples as well as mappings:
>>> d = odict([('a', 'b'), ('c', 'd')])
>>> d.update({'foo': 'bar'})
>>> d
odict.odict([('a', 'b'), ('c', 'd'), ('foo', 'bar')])
Keep in mind that when updating from dict-literals the order is not preserved as these dicts are unsorted!
You can copy an odict like a dict by using the constructor, copy.copy or the copy method and make deep copies with copy.deepcopy:
>>> from copy import copy, deepcopy
>>> copy(d)
odict.odict([('a', 'b'), ('c', 'd'), ('foo', 'bar')])
>>> d.copy()
odict.odict([('a', 'b'), ('c', 'd'), ('foo', 'bar')])
>>> odict(d)
odict.odict([('a', 'b'), ('c', 'd'), ('foo', 'bar')])
>>> d['spam'] = []
>>> d2 = deepcopy(d)
>>> d2['spam'].append('eggs')
>>> d
odict.odict([('a', 'b'), ('c', 'd'), ('foo', 'bar'), ('spam', [])])
>>> d2
odict.odict([('a', 'b'), ('c', 'd'), ('foo', 'bar'), ('spam', ['eggs'])])
All iteration methods as well as keys, values and items return the values ordered by the the time the key-value pair is inserted:
>>> d.keys()
['a', 'c', 'foo', 'spam']
>>> d.values()
['b', 'd', 'bar', []]
>>> d.items()
[('a', 'b'), ('c', 'd'), ('foo', 'bar'), ('spam', [])]
>>> list(d.iterkeys())
['a', 'c', 'foo', 'spam']
>>> list(d.itervalues())
['b', 'd', 'bar', []]
>>> list(d.iteritems())
[('a', 'b'), ('c', 'd'), ('foo', 'bar'), ('spam', [])]
Index based lookup is supported too by byindex which returns the key/value pair for an index:
>>> d.byindex(2)
('foo', 'bar')
You can reverse the odict as well:
>>> d.reverse()
>>> d
odict.odict([('spam', []), ('foo', 'bar'), ('c', 'd'), ('a', 'b')])
And sort it like a list:
>>> d.sort(key=lambda x: x[0].lower())
>>> d
odict.odict([('a', 'b'), ('c', 'd'), ('foo', 'bar'), ('spam', [])])