mirror of
https://github.com/RetroDECK/RetroDECK.git
synced 2024-11-27 00:05:37 +00:00
220 lines
6.6 KiB
Python
220 lines
6.6 KiB
Python
|
# pylint: disable=no-member,unnecessary-dunder-call
|
||
|
from collections import Counter
|
||
|
|
||
|
_iter_values = 'values'
|
||
|
_range = range
|
||
|
_string_type = str
|
||
|
import collections as _c
|
||
|
|
||
|
|
||
|
class _kView(_c.KeysView):
|
||
|
def __iter__(self):
|
||
|
return self._mapping.iterkeys()
|
||
|
|
||
|
|
||
|
class _vView(_c.ValuesView):
|
||
|
def __iter__(self):
|
||
|
return self._mapping.itervalues()
|
||
|
|
||
|
|
||
|
class _iView(_c.ItemsView):
|
||
|
def __iter__(self):
|
||
|
return self._mapping.iteritems()
|
||
|
|
||
|
|
||
|
class VDFDict(dict):
|
||
|
def __init__(self, data=None):
|
||
|
"""
|
||
|
This is a dictionary that supports duplicate keys and preserves insert order
|
||
|
|
||
|
``data`` can be a ``dict``, or a sequence of key-value tuples. (e.g. ``[('key', 'value'),..]``)
|
||
|
The only supported type for key is str.
|
||
|
|
||
|
Get/set duplicates is done by tuples ``(index, key)``, where index is the duplicate index
|
||
|
for the specified key. (e.g. ``(0, 'key')``, ``(1, 'key')``...)
|
||
|
|
||
|
When the ``key`` is ``str``, instead of tuple, set will create a duplicate and get will look up ``(0, key)``
|
||
|
"""
|
||
|
super().__init__()
|
||
|
self.__omap = []
|
||
|
self.__kcount = Counter()
|
||
|
|
||
|
if data is not None:
|
||
|
if not isinstance(data, (list, dict)):
|
||
|
raise ValueError("Expected data to be list of pairs or dict, got %s" % type(data))
|
||
|
self.update(data)
|
||
|
|
||
|
def __repr__(self):
|
||
|
out = "%s(" % self.__class__.__name__
|
||
|
out += "%s)" % repr(list(self.iteritems()))
|
||
|
return out
|
||
|
|
||
|
def __len__(self):
|
||
|
return len(self.__omap)
|
||
|
|
||
|
def _verify_key_tuple(self, key):
|
||
|
if len(key) != 2:
|
||
|
raise ValueError("Expected key tuple length to be 2, got %d" % len(key))
|
||
|
if not isinstance(key[0], int):
|
||
|
raise TypeError("Key index should be an int")
|
||
|
if not isinstance(key[1], _string_type):
|
||
|
raise TypeError("Key value should be a str")
|
||
|
|
||
|
def _normalize_key(self, key):
|
||
|
if isinstance(key, _string_type):
|
||
|
key = (0, key)
|
||
|
elif isinstance(key, tuple):
|
||
|
self._verify_key_tuple(key)
|
||
|
else:
|
||
|
raise TypeError("Expected key to be a str or tuple, got %s" % type(key))
|
||
|
return key
|
||
|
|
||
|
def __setitem__(self, key, value):
|
||
|
if isinstance(key, _string_type):
|
||
|
key = (self.__kcount[key], key)
|
||
|
self.__omap.append(key)
|
||
|
elif isinstance(key, tuple):
|
||
|
self._verify_key_tuple(key)
|
||
|
if key not in self:
|
||
|
raise KeyError("%s doesn't exist" % repr(key))
|
||
|
else:
|
||
|
raise TypeError("Expected either a str or tuple for key")
|
||
|
super().__setitem__(key, value)
|
||
|
self.__kcount[key[1]] += 1
|
||
|
|
||
|
def __getitem__(self, key):
|
||
|
return super().__getitem__(self._normalize_key(key))
|
||
|
|
||
|
def __delitem__(self, key):
|
||
|
key = self._normalize_key(key)
|
||
|
result = super().__delitem__(key)
|
||
|
|
||
|
start_idx = self.__omap.index(key)
|
||
|
del self.__omap[start_idx]
|
||
|
|
||
|
dup_idx, skey = key
|
||
|
self.__kcount[skey] -= 1
|
||
|
tail_count = self.__kcount[skey] - dup_idx
|
||
|
|
||
|
if tail_count > 0:
|
||
|
for idx in _range(start_idx, len(self.__omap)):
|
||
|
if self.__omap[idx][1] == skey:
|
||
|
oldkey = self.__omap[idx]
|
||
|
newkey = (dup_idx, skey)
|
||
|
super().__setitem__(newkey, self[oldkey])
|
||
|
super().__delitem__(oldkey)
|
||
|
self.__omap[idx] = newkey
|
||
|
|
||
|
dup_idx += 1
|
||
|
tail_count -= 1
|
||
|
if tail_count == 0:
|
||
|
break
|
||
|
|
||
|
if self.__kcount[skey] == 0:
|
||
|
del self.__kcount[skey]
|
||
|
|
||
|
return result
|
||
|
|
||
|
def __iter__(self):
|
||
|
return iter(self.iterkeys())
|
||
|
|
||
|
def __contains__(self, key):
|
||
|
return super().__contains__(self._normalize_key(key))
|
||
|
|
||
|
def __eq__(self, other):
|
||
|
if isinstance(other, VDFDict):
|
||
|
return list(self.items()) == list(other.items())
|
||
|
return False
|
||
|
|
||
|
def __ne__(self, other):
|
||
|
return not self.__eq__(other)
|
||
|
|
||
|
def clear(self):
|
||
|
super().clear()
|
||
|
self.__kcount.clear()
|
||
|
self.__omap = []
|
||
|
|
||
|
def get(self, key, *args):
|
||
|
return super().get(self._normalize_key(key), *args)
|
||
|
|
||
|
def setdefault(self, key, default=None):
|
||
|
if key not in self:
|
||
|
self.__setitem__(key, default)
|
||
|
return self.__getitem__(key)
|
||
|
|
||
|
def pop(self, key):
|
||
|
key = self._normalize_key(key)
|
||
|
value = self.__getitem__(key)
|
||
|
self.__delitem__(key)
|
||
|
return value
|
||
|
|
||
|
def popitem(self):
|
||
|
if not self.__omap:
|
||
|
raise KeyError("VDFDict is empty")
|
||
|
key = self.__omap[-1]
|
||
|
return key[1], self.pop(key)
|
||
|
|
||
|
def update(self, data=None, **kwargs):
|
||
|
if isinstance(data, dict):
|
||
|
data = data.items()
|
||
|
elif not isinstance(data, list):
|
||
|
raise TypeError("Expected data to be a list or dict, got %s" % type(data))
|
||
|
|
||
|
for key, value in data:
|
||
|
self.__setitem__(key, value)
|
||
|
|
||
|
def iterkeys(self):
|
||
|
return (key[1] for key in self.__omap)
|
||
|
|
||
|
def keys(self):
|
||
|
return _kView(self)
|
||
|
|
||
|
def itervalues(self):
|
||
|
return (self[key] for key in self.__omap)
|
||
|
|
||
|
def values(self):
|
||
|
return _vView(self)
|
||
|
|
||
|
def iteritems(self):
|
||
|
return ((key[1], self[key]) for key in self.__omap)
|
||
|
|
||
|
def items(self):
|
||
|
return _iView(self)
|
||
|
|
||
|
def get_all_for(self, key):
|
||
|
""" Returns all values of the given key """
|
||
|
if not isinstance(key, _string_type):
|
||
|
raise TypeError("Key needs to be a string.")
|
||
|
return [self[(idx, key)] for idx in _range(self.__kcount[key])]
|
||
|
|
||
|
def remove_all_for(self, key):
|
||
|
""" Removes all items with the given key """
|
||
|
if not isinstance(key, _string_type):
|
||
|
raise TypeError("Key need to be a string.")
|
||
|
|
||
|
for idx in _range(self.__kcount[key]):
|
||
|
super().__delitem__((idx, key))
|
||
|
|
||
|
self.__omap = list(filter(lambda x: x[1] != key, self.__omap))
|
||
|
|
||
|
del self.__kcount[key]
|
||
|
|
||
|
def has_duplicates(self):
|
||
|
"""
|
||
|
Returns ``True`` if the dict contains keys with duplicates.
|
||
|
Recurses through any all keys with value that is ``VDFDict``.
|
||
|
"""
|
||
|
for n in getattr(self.__kcount, _iter_values)():
|
||
|
if n != 1:
|
||
|
return True
|
||
|
|
||
|
def dict_recurse(obj):
|
||
|
for v in getattr(obj, _iter_values)():
|
||
|
if isinstance(v, VDFDict) and v.has_duplicates():
|
||
|
return True
|
||
|
if isinstance(v, dict):
|
||
|
return dict_recurse(v)
|
||
|
return False
|
||
|
|
||
|
return dict_recurse(self)
|