Source code for ubelt.util_dict

"""
Functions for working with dictionaries.

The :class:`UDict` is a subclass of :class:`dict` with quality of life
improvements. It contains methods for n-ary key-wise set operations as well as
support for the binary operators in addition to other methods for mapping,
inversion, subdicts, and peeking. It can be accessed via the alias
``ubelt.udict``.

The :class:`SetDict` only contains the key-wise set extensions to dict. It can
be accessed via the alias ``ubelt.sdict``.

The :func:`dict_hist` function counts the number of discrete occurrences of hashable
items. Similarly :func:`find_duplicates` looks for indices of items that occur more
than `k=1` times.

The :func:`map_keys` and :func:`map_values` functions are useful for
transforming the keys and values of a dictionary with less syntax than a dict
comprehension.

The :func:`dict_union`, :func:`dict_isect`, and :func:`dict_diff` functions are
similar to the set equivalents.

The :func:`dzip` function zips two iterables and packs them into a dictionary
where the first iterable is used to generate keys and the second generates
values.

The :func:`group_items` function takes two lists and returns a dict mapping
values in the second list to all items in corresponding locations in the first
list.

The :func:`invert_dict` function swaps keys and values. See the function docs
for details on dealing with unique and non-unique values.

The :func:`ddict` and :func:`odict` functions are alias for the commonly used
:func:`collections.defaultdict` and :func:`collections.OrderedDict` classes.

Related Work:
    * Note that Python does support set operations on dictionary **views** [DictView]_ [Pep3106]_, but these methods can be inflexible and often leave you only with keys (and no dictionary subset operation), whereas the ubelt definition of these operations is more straightforward.
    * There are several recipes for dictionaries that support set operations [SetDictRecipe1]_ [SetDictRecipe2]_.
    * The :py:mod:`dictmap` package contains a function similar to :func:`map_values` [GHDictMap]_.
    * The :py:mod:`dictdiffer` package contains tools for nested difference operations [PypiDictDiffer]_.
    * There are lots of other python dictionary utility libraries [PyPIAddict]_.

References:
    .. [PyPIAddict] https://github.com/mewwts/addict
    .. [SetDictRecipe1] https://gist.github.com/rossmacarthur/38fa948b175abb512e12c516cc3b936d
    .. [SetDictRecipe2] https://code.activestate.com/recipes/577471-setdict/
    .. [PypiDictDiffer] https://pypi.org/project/dictdiffer/
    .. [DictView] https://docs.python.org/3.0/library/stdtypes.html#dictionary-view-objects
    .. [Pep3106] https://peps.python.org/pep-3106/
    .. [GHDictMap] https://github.com/ulisesojeda/dictionary_map
"""
import sys
import operator as op
import itertools as it
from collections import OrderedDict
from collections import defaultdict
from ubelt.util_const import NoParam

__all__ = [
    'AutoDict',
    'AutoOrderedDict',
    'dzip',
    'ddict',
    'dict_hist',
    'dict_subset',
    'dict_union',
    'dict_isect',
    'dict_diff',
    'find_duplicates',
    'group_items',
    'invert_dict',
    'map_keys',
    'map_vals',
    'map_values',
    'sorted_keys',
    'sorted_vals',
    'sorted_values',
    'odict',
    'named_product',
    'varied_values',
    'SetDict',
    'UDict',
    'sdict',
    'udict',
]


# Expose for convenience
odict = OrderedDict
ddict = defaultdict


# Use an ordered dictionary in < 3.7 as the base

if sys.version_info[0:2] <= (3, 6):  # nocover
    DictBase = OrderedDict
else:  # nocover
    DictBase = dict


[docs] def dzip(items1, items2, cls=dict): """ Zips elementwise pairs between items1 and items2 into a dictionary. Values from items2 can be broadcast onto items1. Args: items1 (Iterable[KT]): full sequence items2 (Iterable[VT]): can either be a sequence of one item or a sequence of equal length to ``items1`` cls (Type[dict]): dictionary type to use. Defaults to ``dict``. Returns: Dict[KT, VT]: similar to ``dict(zip(items1, items2))``. Example: >>> import ubelt as ub >>> assert ub.dzip([1, 2, 3], [4]) == {1: 4, 2: 4, 3: 4} >>> assert ub.dzip([1, 2, 3], [4, 4, 4]) == {1: 4, 2: 4, 3: 4} >>> assert ub.dzip([], [4]) == {} """ try: len(items1) except TypeError: items1 = list(items1) try: len(items2) except TypeError: items2 = list(items2) if len(items1) == 0 and len(items2) == 1: # Corner case: # allow the first list to be empty and the second list to broadcast a # value. This means that the equality check won't work for the case # where items1 and items2 are supposed to correspond, but the length of # items2 is 1. items2 = [] if len(items2) == 1 and len(items1) > 1: items2 = items2 * len(items1) if len(items1) != len(items2): raise ValueError('out of alignment len(items1)=%r, len(items2)=%r' % ( len(items1), len(items2))) return cls(zip(items1, items2))
[docs] def group_items(items, key): """ Groups a list of items by group id. Args: items (Iterable[VT]): a list of items to group key (Iterable[KT] | Callable[[VT], KT]): either a corresponding list of group-ids for each item or a function used to map each item to a group-id. Returns: dict[KT, List[VT]]: a mapping from each group id to the list of corresponding items Example: >>> import ubelt as ub >>> items = ['ham', 'jam', 'spam', 'eggs', 'cheese', 'banana'] >>> groupids = ['protein', 'fruit', 'protein', 'protein', 'dairy', 'fruit'] >>> id_to_items = ub.group_items(items, groupids) >>> print(ub.repr2(id_to_items, nl=0)) {'dairy': ['cheese'], 'fruit': ['jam', 'banana'], 'protein': ['ham', 'spam', 'eggs']} """ if callable(key): keyfunc = key pair_list = ((keyfunc(item), item) for item in items) else: pair_list = zip(key, items) # Initialize a dict of lists id_to_items = defaultdict(list) # Insert each item into the correct group for key, item in pair_list: id_to_items[key].append(item) return id_to_items
[docs] def dict_hist(items, weights=None, ordered=False, labels=None): """ Builds a histogram of items, counting the number of time each item appears in the input. Args: items (Iterable[T]): hashable items (usually containing duplicates) weights (Iterable[float] | None): Corresponding weights for each item, defaults to 1 if unspecified. Defaults to None. ordered (bool): If True the result is ordered by frequency. Defaults to False. labels (Iterable[T] | None): Expected labels. Allows this function to pre-initialize the histogram. If specified the frequency of each label is initialized to zero and ``items`` can only contain items specified in labels. Defaults to None. Returns: dict[T, int] : dictionary where the keys are unique elements from ``items``, and the values are the number of times the item appears in ``items``. SeeAlso: :class:`collections.Counter` Example: >>> import ubelt as ub >>> items = [1, 2, 39, 900, 1232, 900, 1232, 2, 2, 2, 900] >>> hist = ub.dict_hist(items) >>> print(ub.repr2(hist, nl=0)) {1: 1, 2: 4, 39: 1, 900: 3, 1232: 2} Example: >>> import ubelt as ub >>> import pytest >>> items = [1, 2, 39, 900, 1232, 900, 1232, 2, 2, 2, 900] >>> hist1 = ub.dict_hist(items) >>> hist2 = ub.dict_hist(items, ordered=True) >>> with pytest.raises(KeyError): >>> hist3 = ub.dict_hist(items, labels=[]) >>> weights = [1, 1, 1, 0, 0, 0, 0, 1, 1, 1, 1] >>> hist4 = ub.dict_hist(items, weights=weights) >>> print(ub.repr2(hist1, nl=0)) {1: 1, 2: 4, 39: 1, 900: 3, 1232: 2} >>> print(ub.repr2(hist4, nl=0)) {1: 1, 2: 4, 39: 1, 900: 1, 1232: 0} """ if weights is None and labels is None: # Accumulate discrete frequency. # In this special case we use an optimized stdlib routine from collections import Counter hist_ = Counter() hist_.update(items) else: if labels is None: hist_ = defaultdict(lambda: 0) else: hist_ = {k: 0 for k in labels} if weights is None: weights = it.repeat(1) # 2x slower than Counter # Accumulate weighted frequency for item, weight in zip(items, weights): hist_[item] += weight if ordered: # Order by value getval = op.itemgetter(1) hist = OrderedDict([ (key, value) for (key, value) in sorted(hist_.items(), key=getval) ]) else: # Cast to a normal dictionary hist = dict(hist_) return hist
[docs] def find_duplicates(items, k=2, key=None): """ Find all duplicate items in a list. Search for all items that appear more than ``k`` times and return a mapping from each (k)-duplicate item to the positions it appeared in. Args: items (Iterable[T]): Hashable items possibly containing duplicates k (int): Only return items that appear at least ``k`` times. Defaults to 2. key (Callable[[T], Any] | None): Returns indices where `key(items[i])` maps to a particular value at least k times. Default to None. Returns: dict[T, List[int]] : Maps each duplicate item to the indices at which it appears Notes: Similar to :func:`more_itertools.duplicates_everseen`, :func:`more_itertools.duplicates_justseen`. Example: >>> import ubelt as ub >>> items = [0, 0, 1, 2, 3, 3, 0, 12, 2, 9] >>> duplicates = ub.find_duplicates(items) >>> # Duplicates are a mapping from each item that occurs 2 or more >>> # times to the indices at which they occur. >>> assert duplicates == {0: [0, 1, 6], 2: [3, 8], 3: [4, 5]} >>> # You can set k=3 if you want to don't mind duplicates but you >>> # want to find triplicates or quadruplets etc. >>> assert ub.find_duplicates(items, k=3) == {0: [0, 1, 6]} Example: >>> import ubelt as ub >>> items = [0, 0, 1, 2, 3, 3, 0, 12, 2, 9] >>> # note: k can less then 2 >>> duplicates = ub.find_duplicates(items, k=0) >>> print(ub.repr2(duplicates, nl=0)) {0: [0, 1, 6], 1: [2], 2: [3, 8], 3: [4, 5], 9: [9], 12: [7]} Example: >>> import ubelt as ub >>> items = [10, 11, 12, 13, 14, 15, 16] >>> duplicates = ub.find_duplicates(items, key=lambda x: x // 2) >>> print(ub.repr2(duplicates, nl=0)) {5: [0, 1], 6: [2, 3], 7: [4, 5]} """ # Build mapping from items to the indices at which they appear duplicates = defaultdict(list) if key is None: for count, item in enumerate(items): duplicates[item].append(count) else: for count, item in enumerate(items): duplicates[key(item)].append(count) # remove items seen fewer than k times. for key in list(duplicates.keys()): if len(duplicates[key]) < k: del duplicates[key] duplicates = dict(duplicates) return duplicates
[docs] def dict_subset(dict_, keys, default=NoParam, cls=OrderedDict): """ Get a subset of a dictionary Args: dict_ (Dict[KT, VT]): superset dictionary keys (Iterable[KT]): keys to take from ``dict_`` default (Optional[object] | NoParamType): if specified uses default if keys are missing. cls (Type[Dict]): type of the returned dictionary. Defaults to ``OrderedDict``. Returns: Dict[KT, VT]: subset dictionary SeeAlso: :func:`dict_isect` - similar functionality, but ignores missing keys ::py:meth:`UDict.subdict` - object oriented version of this function Example: >>> import ubelt as ub >>> dict_ = {'K': 3, 'dcvs_clip_max': 0.2, 'p': 0.1} >>> keys = ['K', 'dcvs_clip_max'] >>> subdict_ = ub.dict_subset(dict_, keys) >>> print(ub.repr2(subdict_, nl=0)) {'K': 3, 'dcvs_clip_max': 0.2} """ from ubelt import util_list keys = list(keys) items = util_list.take(dict_, keys, default) subdict_ = cls(list(zip(keys, items))) return subdict_
[docs] def dict_union(*args): """ Dictionary set extension for ``set.union`` Combines items with from multiple dictionaries. For items with intersecting keys, dictionaries towards the end of the sequence are given precedence. Args: *args (List[Dict]) : A sequence of dictionaries. Values are taken from the last Returns: Dict | OrderedDict : OrderedDict if the first argument is an OrderedDict, otherwise dict Notes: In Python 3.8+, the bitwise or operator "|" operator performs a similar operation, but as of 2022-06-01 there is still no public method for dictionary union (or any other dictionary set operator). References: .. [SO38987] https://stackoverflow.com/questions/38987/merge-two-dict SeeAlso: :func:`collections.ChainMap` - a standard python builtin data structure that provides a view that treats multiple dicts as a single dict. `<https://docs.python.org/3/library/collections.html#chainmap-objects>`_ ::py:meth:`UDict.union` - object oriented version of this function Example: >>> import ubelt as ub >>> result = ub.dict_union({'a': 1, 'b': 1}, {'b': 2, 'c': 2}) >>> assert result == {'a': 1, 'b': 2, 'c': 2} >>> output = ub.dict_union( >>> ub.odict([('a', 1), ('b', 2)]), >>> ub.odict([('c', 3), ('d', 4)])) >>> print(ub.urepr(output, nl=0)) {'a': 1, 'b': 2, 'c': 3, 'd': 4} >>> ub.dict_union() {} """ if not args: return {} else: dictclass = OrderedDict if isinstance(args[0], OrderedDict) else dict return dictclass(it.chain.from_iterable(d.items() for d in args))
[docs] def dict_diff(*args): """ Dictionary set extension for :func:`set.difference` Constructs a dictionary that contains any of the keys in the first arg, which are not in any of the following args. Args: *args (List[Dict[KT, VT] | Iterable[KT]]) : A sequence of dictionaries (or sets of keys). The first argument should always be a dictionary, but the subsequent arguments can just be sets of keys. Returns: Dict[KT, VT] | OrderedDict[KT, VT] : OrderedDict if the first argument is an OrderedDict, otherwise dict SeeAlso: ::py:meth:`UDict.difference` - object oriented version of this function Example: >>> import ubelt as ub >>> ub.dict_diff({'a': 1, 'b': 1}, {'a'}, {'c'}) {'b': 1} >>> result = ub.dict_diff(ub.odict([('a', 1), ('b', 2)]), ub.odict([('c', 3)])) >>> print(ub.urepr(result, nl=0)) {'a': 1, 'b': 2} >>> ub.dict_diff() {} >>> ub.dict_diff({'a': 1, 'b': 2}, {'c'}) """ if not args: return {} else: first_dict = args[0] dictclass = OrderedDict if isinstance(first_dict, OrderedDict) else dict # remove_keys = set.union(*map(set, args[1:])) # new = dictclass((k, v) for k, v in first_dict.items() if k not in remove_keys) remove_keys = set.union(*map(set, args[1:])) new = dictclass((k, first_dict[k]) for k in first_dict.keys() if k not in remove_keys) return new
[docs] def dict_isect(*args): """ Dictionary set extension for :func:`set.intersection` Constructs a dictionary that contains keys common between all inputs. The returned values will only belong to the first dictionary. Args: *args (List[Dict[KT, VT] | Iterable[KT]]) : A sequence of dictionaries (or sets of keys). The first argument should always be a dictionary, but the subsequent arguments can just be sets of keys. Returns: Dict[KT, VT] | OrderedDict[KT, VT] : OrderedDict if the first argument is an OrderedDict, otherwise dict SeeAlso: ::py:meth:`UDict.intersection` - object oriented version of this function Note: This function can be used as an alternative to :func:`dict_subset` where any key not in the dictionary is ignored. See the following example: >>> import ubelt as ub >>> # xdoctest: +IGNORE_WANT >>> ub.dict_isect({'a': 1, 'b': 2, 'c': 3}, ['a', 'c', 'd']) {'a': 1, 'c': 3} Example: >>> import ubelt as ub >>> ub.dict_isect({'a': 1, 'b': 1}, {'b': 2, 'c': 2}) {'b': 1} >>> ub.dict_isect(odict([('a', 1), ('b', 2)]), odict([('c', 3)])) OrderedDict() >>> ub.dict_isect() {} """ if not args: return {} else: dictclass = OrderedDict if isinstance(args[0], OrderedDict) else dict common_keys = set.intersection(*map(set, args)) first_dict = args[0] return dictclass((k, first_dict[k]) for k in first_dict if k in common_keys)
[docs] def map_values(func, dict_, cls=None): """ Apply a function to every value in a dictionary. Creates a new dictionary with the same keys and modified values. Args: func (Callable[[VT], T] | Mapping[VT, T]): a function or indexable object dict_ (Dict[KT, VT]): a dictionary cls (type | None): specifies the dict subclassof the result. if unspecified will be dict or OrderedDict. This behavior may change. SeeAlso: ::py:meth:`UDict.map_values` - object oriented version of this function Returns: Dict[KT, T]: transformed dictionary Notes: Similar to :py:mod:`dictmap.dict_map` Example: >>> import ubelt as ub >>> dict_ = {'a': [1, 2, 3], 'b': []} >>> newdict = ub.map_values(len, dict_) >>> assert newdict == {'a': 3, 'b': 0} Example: >>> # Can also use an indexable as ``func`` >>> import ubelt as ub >>> dict_ = {'a': 0, 'b': 1} >>> func = [42, 21] >>> newdict = ub.map_values(func, dict_) >>> assert newdict == {'a': 42, 'b': 21} >>> print(newdict) """ if not hasattr(func, '__call__'): func = func.__getitem__ keyval_list = [(key, func(val)) for key, val in dict_.items()] if cls is None: cls = OrderedDict if isinstance(dict_, OrderedDict) else dict newdict = cls(keyval_list) return newdict
map_vals = map_values # backwards compatibility
[docs] def map_keys(func, dict_, cls=None): """ Apply a function to every key in a dictionary. Creates a new dictionary with the same values and modified keys. An error is raised if the new keys are not unique. Args: func (Callable[[KT], T] | Mapping[KT, T]): a function or indexable object dict_ (Dict[KT, VT]): a dictionary cls (type | None): specifies the dict subclassof the result. if unspecified will be dict or OrderedDict. This behavior may change. SeeAlso: ::py:meth:`UDict.map_keys` - object oriented version of this function Returns: Dict[T, VT]: transformed dictionary Raises: Exception : if multiple keys map to the same value Example: >>> import ubelt as ub >>> dict_ = {'a': [1, 2, 3], 'b': []} >>> func = ord >>> newdict = ub.map_keys(func, dict_) >>> print(newdict) >>> assert newdict == {97: [1, 2, 3], 98: []} >>> dict_ = {0: [1, 2, 3], 1: []} >>> func = ['a', 'b'] >>> newdict = ub.map_keys(func, dict_) >>> print(newdict) >>> assert newdict == {'a': [1, 2, 3], 'b': []} """ if not hasattr(func, '__call__'): func = func.__getitem__ keyval_list = [(func(key), val) for key, val in dict_.items()] if cls is None: cls = OrderedDict if isinstance(dict_, OrderedDict) else dict newdict = cls(keyval_list) if len(newdict) != len(dict_): raise Exception('multiple input keys mapped to the same output key') return newdict
[docs] def sorted_values(dict_, key=None, reverse=False, cls=OrderedDict): """ Return an ordered dictionary sorted by its values Args: dict_ (Dict[KT, VT]): dictionary to sort. The values must be of comparable types. key (Callable[[VT], Any] | None): If given as a callable, customizes the sorting by ordering using transformed values. reverse (bool, default): If True returns in descending order. Defaults to False. cls (type): Specifies the dict return type. Default to OrderedDict. SeeAlso: ::py:meth:`UDict.sorted_values` - object oriented version of this function Returns: OrderedDict[KT, VT]: new dictionary where the values are ordered Example: >>> import ubelt as ub >>> dict_ = {'spam': 2.62, 'eggs': 1.20, 'jam': 2.92} >>> newdict = sorted_values(dict_) >>> print(ub.repr2(newdict, nl=0)) {'eggs': 1.2, 'spam': 2.62, 'jam': 2.92} >>> newdict = sorted_values(dict_, reverse=True) >>> print(ub.repr2(newdict, nl=0)) {'jam': 2.92, 'spam': 2.62, 'eggs': 1.2} >>> newdict = sorted_values(dict_, key=lambda x: x % 1.6) >>> print(ub.repr2(newdict, nl=0)) {'spam': 2.62, 'eggs': 1.2, 'jam': 2.92} """ if key is None: newdict = OrderedDict(sorted(dict_.items(), key=lambda kv: kv[1], reverse=reverse)) else: newdict = OrderedDict(sorted(dict_.items(), key=lambda kv: key(kv[1]), reverse=reverse)) return newdict
sorted_vals = sorted_values # backwards compatibility
[docs] def sorted_keys(dict_, key=None, reverse=False, cls=OrderedDict): """ Return an ordered dictionary sorted by its keys Args: dict_ (Dict[KT, VT]): Dictionary to sort. The keys must be of comparable types. key (Callable[[KT], Any] | None): If given as a callable, customizes the sorting by ordering using transformed keys. reverse (bool): If True returns in descending order. Default to False. cls (type): specifies the dict return type SeeAlso: ::py:meth:`UDict.sorted_keys` - object oriented version of this function Returns: OrderedDict[KT, VT]: new dictionary where the keys are ordered Example: >>> import ubelt as ub >>> dict_ = {'spam': 2.62, 'eggs': 1.20, 'jam': 2.92} >>> newdict = sorted_keys(dict_) >>> print(ub.repr2(newdict, nl=0)) {'eggs': 1.2, 'jam': 2.92, 'spam': 2.62} >>> newdict = sorted_keys(dict_, reverse=True) >>> print(ub.repr2(newdict, nl=0)) {'spam': 2.62, 'jam': 2.92, 'eggs': 1.2} >>> newdict = sorted_keys(dict_, key=lambda x: sum(map(ord, x))) >>> print(ub.repr2(newdict, nl=0)) {'jam': 2.92, 'eggs': 1.2, 'spam': 2.62} """ if key is None: newdict = OrderedDict(sorted(dict_.items(), key=lambda kv: kv[0], reverse=reverse)) else: newdict = OrderedDict(sorted(dict_.items(), key=lambda kv: key(kv[0]), reverse=reverse)) return newdict
[docs] def invert_dict(dict_, unique_vals=True, cls=None): """ Swaps the keys and values in a dictionary. Args: dict_ (Dict[KT, VT]): dictionary to invert unique_vals (bool): if False, the values of the new dictionary are sets of the original keys. Defaults to True. cls (type | None): specifies the dict subclassof the result. if unspecified will be dict or OrderedDict. This behavior may change. SeeAlso: ::py:meth:`UDict.invert` - object oriented version of this function Returns: Dict[VT, KT] | Dict[VT, Set[KT]]: the inverted dictionary Note: The must values be hashable. If the original dictionary contains duplicate values, then only one of the corresponding keys will be returned and the others will be discarded. This can be prevented by setting ``unique_vals=False``, causing the inverted keys to be returned in a set. Example: >>> import ubelt as ub >>> dict_ = {'a': 1, 'b': 2} >>> inverted = ub.invert_dict(dict_) >>> assert inverted == {1: 'a', 2: 'b'} Example: >>> import ubelt as ub >>> dict_ = ub.odict([(2, 'a'), (1, 'b'), (0, 'c'), (None, 'd')]) >>> inverted = ub.invert_dict(dict_) >>> assert list(inverted.keys())[0] == 'a' Example: >>> import ubelt as ub >>> dict_ = {'a': 1, 'b': 0, 'c': 0, 'd': 0, 'f': 2} >>> inverted = ub.invert_dict(dict_, unique_vals=False) >>> assert inverted == {0: {'b', 'c', 'd'}, 1: {'a'}, 2: {'f'}} """ if cls is None: cls = OrderedDict if isinstance(dict_, OrderedDict) else dict if unique_vals: # Wonder what byte code is better here? if cls is dict: inverted = {val: key for key, val in dict_.items()} else: inverted = cls((val, key) for key, val in dict_.items()) else: # Handle non-unique keys using groups inverted = defaultdict(set) for key, value in dict_.items(): inverted[value].add(key) inverted = cls(inverted) return inverted
[docs] def named_product(_=None, **basis): """ Generates the Cartesian product of the ``basis.values()``, where each generated item labeled by ``basis.keys()``. In other words, given a dictionary that maps each "axes" (i.e. some variable) to its "basis" (i.e. the possible values that it can take), generate all possible points in that grid (i.e. unique assignments of variables to values). Args: _ (Dict[str, List[VT]] | None): Use of this positional argument is not recommend. Instead specify all arguments as keyword args. Defaults to None. If specified, this should be a dictionary is unioned with the keyword args. This exists to support ordered dictionaries before Python 3.6, and may eventually be removed. basis (Dict[str, List[VT]]): A dictionary where the keys correspond to "columns" and the values are a list of possible values that "column" can take. I.E. each key corresponds to an "axes", the values are the list of possible values for that "axes". Yields: Dict[str, VT] : a "row" in the "longform" data containing a point in the Cartesian product. Note: This function is similar to :func:`itertools.product`, the only difference is that the generated items are a dictionary that retains the input keys instead of an tuple. This function used to be called "basis_product", but "named_product" might be more appropriate. This function exists in other places ([minstrel271_namedproduct]_, [pytb_namedproduct]_, and [Hettinger_namedproduct]_). References: .. [minstrel271_namedproduct] https://gist.github.com/minstrel271/d51654af3fa4e6411267 .. [pytb_namedproduct] https://py-toolbox.readthedocs.io/en/latest/modules/itertools.html# .. [Hettinger_namedproduct] https://twitter.com/raymondh/status/970380630822305792 Example: >>> # An example use case is looping over all possible settings in a >>> # configuration dictionary for a grid search over parameters. >>> import ubelt as ub >>> basis = { >>> 'arg1': [1, 2, 3], >>> 'arg2': ['A1', 'B1'], >>> 'arg3': [9999, 'Z2'], >>> 'arg4': ['always'], >>> } >>> import ubelt as ub >>> # sort input data for older python versions >>> basis = ub.odict(sorted(basis.items())) >>> got = list(ub.named_product(basis)) >>> print(ub.repr2(got, nl=-1)) [ {'arg1': 1, 'arg2': 'A1', 'arg3': 9999, 'arg4': 'always'}, {'arg1': 1, 'arg2': 'A1', 'arg3': 'Z2', 'arg4': 'always'}, {'arg1': 1, 'arg2': 'B1', 'arg3': 9999, 'arg4': 'always'}, {'arg1': 1, 'arg2': 'B1', 'arg3': 'Z2', 'arg4': 'always'}, {'arg1': 2, 'arg2': 'A1', 'arg3': 9999, 'arg4': 'always'}, {'arg1': 2, 'arg2': 'A1', 'arg3': 'Z2', 'arg4': 'always'}, {'arg1': 2, 'arg2': 'B1', 'arg3': 9999, 'arg4': 'always'}, {'arg1': 2, 'arg2': 'B1', 'arg3': 'Z2', 'arg4': 'always'}, {'arg1': 3, 'arg2': 'A1', 'arg3': 9999, 'arg4': 'always'}, {'arg1': 3, 'arg2': 'A1', 'arg3': 'Z2', 'arg4': 'always'}, {'arg1': 3, 'arg2': 'B1', 'arg3': 9999, 'arg4': 'always'}, {'arg1': 3, 'arg2': 'B1', 'arg3': 'Z2', 'arg4': 'always'} ] Example: >>> import ubelt as ub >>> list(ub.named_product(a=[1, 2, 3])) [{'a': 1}, {'a': 2}, {'a': 3}] >>> # xdoctest: +IGNORE_WANT >>> list(ub.named_product(a=[1, 2, 3], b=[4, 5])) [{'a': 1, 'b': 4}, {'a': 1, 'b': 5}, {'a': 2, 'b': 4}, {'a': 2, 'b': 5}, {'a': 3, 'b': 4}, {'a': 3, 'b': 5}] """ # Handle one positional argument. if _ is not None: _basis = _ _basis.update(basis) basis = _basis keys = list(basis.keys()) for vals in it.product(*basis.values()): kw = dict(zip(keys, vals)) yield kw
[docs] def varied_values(longform, min_variations=0, default=NoParam): """ Given a list of dictionaries, find the values that differ between them. Args: longform (List[Dict[KT, VT]]): This is longform data, as described in [SeabornLongform]_. It is a list of dictionaries. Each item in the list - or row - is a dictionary and can be thought of as an observation. The keys in each dictionary are the columns. The values of the dictionary must be hashable. Lists will be converted into tuples. min_variations (int): "columns" with fewer than ``min_variations`` unique values are removed from the result. Defaults to 0. default (VT | NoParamType): if specified, unspecified columns are given this value. Defaults to NoParam. Returns: Dict[KT, List[VT]] : a mapping from each "column" to the set of unique values it took over each "row". If a column is not specified for each row, it is assumed to take a `default` value, if it is specified. Raises: KeyError: If ``default`` is unspecified and all the rows do not contain the same columns. References: .. [SeabornLongform] https://seaborn.pydata.org/tutorial/data_structure.html#long-form-data Example: >>> # An example use case is to determine what values of a >>> # configuration dictionary were tried in a random search >>> # over a parameter grid. >>> import ubelt as ub >>> longform = [ >>> {'col1': 1, 'col2': 'foo', 'col3': None}, >>> {'col1': 1, 'col2': 'foo', 'col3': None}, >>> {'col1': 2, 'col2': 'bar', 'col3': None}, >>> {'col1': 3, 'col2': 'bar', 'col3': None}, >>> {'col1': 9, 'col2': 'bar', 'col3': None}, >>> {'col1': 1, 'col2': 'bar', 'col3': None}, >>> ] >>> varied = ub.varied_values(longform) >>> print('varied = {}'.format(ub.repr2(varied, nl=1))) varied = { 'col1': {1, 2, 3, 9}, 'col2': {'bar', 'foo'}, 'col3': {None}, } Example: >>> import ubelt as ub >>> import random >>> longform = [ >>> {'col1': 1, 'col2': 'foo', 'col3': None}, >>> {'col1': 1, 'col2': [1, 2], 'col3': None}, >>> {'col1': 2, 'col2': 'bar', 'col3': None}, >>> {'col1': 3, 'col2': 'bar', 'col3': None}, >>> {'col1': 9, 'col2': 'bar', 'col3': None}, >>> {'col1': 1, 'col2': 'bar', 'col3': None, 'extra_col': 3}, >>> ] >>> # Operation fails without a default >>> import pytest >>> with pytest.raises(KeyError): >>> varied = ub.varied_values(longform) >>> # >>> # Operation works with a default >>> varied = ub.varied_values(longform, default='<unset>') >>> expected = { >>> 'col1': {1, 2, 3, 9}, >>> 'col2': {'bar', 'foo', (1, 2)}, >>> 'col3': set([None]), >>> 'extra_col': {'<unset>', 3}, >>> } >>> print('varied = {!r}'.format(varied)) >>> assert varied == expected Example: >>> # xdoctest: +REQUIRES(PY3) >>> # Random numbers are different in Python2, so skip in that case >>> import ubelt as ub >>> import random >>> num_cols = 11 >>> num_rows = 17 >>> rng = random.Random(0) >>> # Generate a set of columns >>> columns = sorted(ub.hash_data(i)[0:8] for i in range(num_cols)) >>> # Generate rows for each column >>> longform = [ >>> {key: ub.hash_data(key)[0:8] for key in columns} >>> for _ in range(num_rows) >>> ] >>> # Add in some varied values in random positions >>> for row in longform: >>> if rng.random() > 0.5: >>> for key in sorted(row.keys()): >>> if rng.random() > 0.95: >>> row[key] = 'special-' + str(rng.randint(1, 32)) >>> varied = ub.varied_values(longform, min_variations=1) >>> print('varied = {}'.format(ub.repr2(varied, nl=1, sort=True))) varied = { '095f3e44': {'8fb4d4c9', 'special-23'}, '365d11a1': {'daa409da', 'special-31', 'special-32'}, '5815087d': {'1b823610', 'special-3'}, '7b54b668': {'349a782c', 'special-10'}, 'b8244d02': {'d57bca90', 'special-8'}, 'f27b5bf8': {'fa0f90d1', 'special-19'}, } """ # Enumerate all defined columns columns = set() for row in longform: if default is NoParam and len(row) != len(columns) and len(columns): missing = set(columns).symmetric_difference(set(row)) raise KeyError(( 'No default specified and not every ' 'row contains columns {}').format(missing)) columns.update(row.keys()) # Build up the set of unique values for each column varied = ddict(set) for row in longform: for key in columns: value = row.get(key, default) if isinstance(value, list): value = tuple(value) varied[key].add(value) # Remove any column that does not have enough variation if min_variations > 0: for key, values in list(varied.items()): if len(values) <= min_variations: varied.pop(key) return varied
[docs] class SetDict(dict): """ A dictionary subclass where all set operations are defined. All of the set operations are defined in a key-wise fashion, that is it is like performing the operation on sets of keys. Value conflicts are handled with left-most priority (default for ``intersection`` and ``difference``), right-most priority (default for ``union`` and ``symmetric_difference``), or via a custom ``merge`` callable similar to [RubyMerge]_. The set operations are: * union (or the ``|`` operator) combines multiple dictionaries into one. This is nearly identical to the update operation. Rightmost values take priority. * intersection (or the ``&`` operator). Takes the items from the first dictionary that share keys with the following dictionaries (or lists or sets of keys). Leftmost values take priority. * difference (or the ``-`` operator). Takes only items from the first dictionary that do not share keys with following dictionaries. Leftmost values take priority. * symmetric_difference (or the ``^`` operator). Takes the items from all dictionaries where the key appears an odd number of times. Rightmost values take priority. The full set of set operations was originally determined to be beyond the scope of [Pep584]_, but there was discussion of these additional operations. Some choices were ambiguous, but we believe this design could be considered "natural". Note: By default the right-most values take priority in union / symmetric_difference and left-most values take priority in intersection / difference. In summary this is because we consider intersection / difference to be "subtractive" operations, and union / symmetric_difference to be "additive" operations. We expand on this in the following points: 1. intersection / difference is for removing keys --- i.e. is used to find values in the first (main) dictionary that are also in some other dictionary (or set or list of keys), whereas 2. union is for adding keys --- i.e. it is basically just an alias for dict.update, so the new (rightmost) keys clobber the old. 3. symmetric_difference is somewhat strange if you aren't familiar with it. At a pure-set level it's not really a difference, its a pairty operation (think of it more like xor or addition modulo 2). You only keep items where the key appears an odd number of times. Unlike intersection and difference, the results may not be a subset of either input. The union has the same property. This symmetry motivates having the newest (rightmost) keys cobber the old. Also, union / symmetric_difference does not make sense if arguments on the rights are lists/sets, whereas difference / intersection does. Note: The SetDict class only defines key-wise set operations. Value-wise or item-wise operations are in general not hashable and therefore not supported. A heavier extension would be needed for that. TODO: - [ ] implement merge callables so the user can specify how to resolve value conflicts / combine values. References: .. [RubyMerge] https://ruby-doc.org/core-2.7.0/Hash.html#method-i-merge .. [Pep584] https://peps.python.org/pep-0584/#what-about-the-full-set-api CommandLine: xdoctest -m ubelt.util_dict SetDict Example: >>> import ubelt as ub >>> a = ub.SetDict({'A': 'Aa', 'B': 'Ba', 'D': 'Da'}) >>> b = ub.SetDict({'A': 'Ab', 'B': 'Bb', 'C': 'Cb', }) >>> print(a.union(b)) >>> print(a.intersection(b)) >>> print(a.difference(b)) >>> print(a.symmetric_difference(b)) {'A': 'Ab', 'B': 'Bb', 'D': 'Da', 'C': 'Cb'} {'A': 'Aa', 'B': 'Ba'} {'D': 'Da'} {'D': 'Da', 'C': 'Cb'} >>> print(a | b) # union >>> print(a & b) # intersection >>> print(a - b) # difference >>> print(a ^ b) # symmetric_difference {'A': 'Ab', 'B': 'Bb', 'D': 'Da', 'C': 'Cb'} {'A': 'Aa', 'B': 'Ba'} {'D': 'Da'} {'D': 'Da', 'C': 'Cb'} Example: >>> import ubelt as ub >>> a = ub.SetDict({'A': 'Aa', 'B': 'Ba', 'D': 'Da'}) >>> b = ub.SetDict({'A': 'Ab', 'B': 'Bb', 'C': 'Cb', }) >>> c = ub.SetDict({'A': 'Ac', 'B': 'Bc', 'E': 'Ec'}) >>> d = ub.SetDict({'A': 'Ad', 'C': 'Cd', 'D': 'Dd'}) >>> # 3-ary operations >>> print(a.union(b, c)) >>> print(a.intersection(b, c)) >>> print(a.difference(b, c)) >>> print(a.symmetric_difference(b, c)) {'A': 'Ac', 'B': 'Bc', 'D': 'Da', 'C': 'Cb', 'E': 'Ec'} {'A': 'Aa', 'B': 'Ba'} {'D': 'Da'} {'D': 'Da', 'C': 'Cb', 'A': 'Ac', 'B': 'Bc', 'E': 'Ec'} >>> # 4-ary operations >>> print(ub.UDict.union(a, b, c, c)) >>> print(ub.UDict.intersection(a, b, c, c)) >>> print(ub.UDict.difference(a, b, c, d)) >>> print(ub.UDict.symmetric_difference(a, b, c, d)) {'A': 'Ac', 'B': 'Bc', 'D': 'Da', 'C': 'Cb', 'E': 'Ec'} {'A': 'Aa', 'B': 'Ba'} {} {'B': 'Bc', 'E': 'Ec'} Example: >>> import ubelt as ub >>> primes = ub.sdict({v: f'prime_{v}' for v in [2, 3, 5, 7, 11]}) >>> evens = ub.sdict({v: f'even_{v}' for v in [0, 2, 4, 6, 8, 10]}) >>> odds = ub.sdict({v: f'odd_{v}' for v in [1, 3, 5, 7, 9, 11]}) >>> squares = ub.sdict({v: f'square_{v}' for v in [0, 1, 4, 9]}) >>> div3 = ub.sdict({v: f'div3_{v}' for v in [0, 3, 6, 9]}) >>> # All of the set methods are defined >>> results1 = {} >>> results1['ints'] = ints = odds.union(evens) >>> results1['composites'] = ints.difference(primes) >>> results1['even_primes'] = evens.intersection(primes) >>> results1['odd_nonprimes_and_two'] = odds.symmetric_difference(primes) >>> print('results1 = {}'.format(ub.repr2(results1, nl=2, sort=True))) results1 = { 'composites': { 0: 'even_0', 1: 'odd_1', 4: 'even_4', 6: 'even_6', 8: 'even_8', 9: 'odd_9', 10: 'even_10', }, 'even_primes': { 2: 'even_2', }, 'ints': { 0: 'even_0', 1: 'odd_1', 2: 'even_2', 3: 'odd_3', 4: 'even_4', 5: 'odd_5', 6: 'even_6', 7: 'odd_7', 8: 'even_8', 9: 'odd_9', 10: 'even_10', 11: 'odd_11', }, 'odd_nonprimes_and_two': { 1: 'odd_1', 2: 'prime_2', 9: 'odd_9', }, } >>> # As well as their corresponding binary operators >>> assert results1['ints'] == odds | evens >>> assert results1['composites'] == ints - primes >>> assert results1['even_primes'] == evens & primes >>> assert results1['odd_nonprimes_and_two'] == odds ^ primes >>> # These can also be used as classmethods >>> assert results1['ints'] == ub.sdict.union(odds, evens) >>> assert results1['composites'] == ub.sdict.difference(ints, primes) >>> assert results1['even_primes'] == ub.sdict.intersection(evens, primes) >>> assert results1['odd_nonprimes_and_two'] == ub.sdict.symmetric_difference(odds, primes) >>> # The narry variants are also implemented >>> results2 = {} >>> results2['nary_union'] = ub.sdict.union(primes, div3, odds) >>> results2['nary_difference'] = ub.sdict.difference(primes, div3, odds) >>> results2['nary_intersection'] = ub.sdict.intersection(primes, div3, odds) >>> # Note that the definition of symmetric difference might not be what you think in the nary case. >>> results2['nary_symmetric_difference'] = ub.sdict.symmetric_difference(primes, div3, odds) >>> print('results2 = {}'.format(ub.repr2(results2, nl=2, sort=True))) results2 = { 'nary_difference': { 2: 'prime_2', }, 'nary_intersection': { 3: 'prime_3', }, 'nary_symmetric_difference': { 0: 'div3_0', 1: 'odd_1', 2: 'prime_2', 3: 'odd_3', 6: 'div3_6', }, 'nary_union': { 0: 'div3_0', 1: 'odd_1', 2: 'prime_2', 3: 'odd_3', 5: 'odd_5', 6: 'div3_6', 7: 'odd_7', 9: 'odd_9', 11: 'odd_11', }, } Example: >>> # A neat thing about our implementation is that often the right >>> # hand side is not required to be a dictionary, just something >>> # that can be cast to a set. >>> import ubelt as ub >>> primes = ub.sdict({2: 'a', 3: 'b', 5: 'c', 7: 'd', 11: 'e'}) >>> assert primes - {2, 3} == {5: 'c', 7: 'd', 11: 'e'} >>> assert primes & {2, 3} == {2: 'a', 3: 'b'} >>> # Union does need to have a second dictionary >>> import pytest >>> with pytest.raises(AttributeError): >>> primes | {2, 3} """
[docs] def copy(self): """ Example: >>> import ubelt as ub >>> a = ub.sdict({1: 1, 2: 2, 3: 3}) >>> b = ub.udict({1: 1, 2: 2, 3: 3}) >>> c = a.copy() >>> d = b.copy() >>> assert c is not a >>> assert d is not b >>> assert d == b >>> assert c == a >>> list(map(type, [a, b, c, d])) >>> assert isinstance(c, ub.sdict) >>> assert isinstance(d, ub.udict) """ return self.__class__(self)
# We could just use the builtin variant for this specific operation def __or__(self, other): """ The ``|`` union operator Args: other (SupportsKeysAndGetItem[Any, Any] | Iterable[Tuple[Any, Any]]): Returns: SetDict """ return self.union(other) def __and__(self, other): """ The ``&`` intersection operator Args: other (Mapping | Iterable): Returns: SetDict """ return self.intersection(other) def __sub__(self, other): """ The ``-`` difference operator Args: other (Mapping | Iterable): Returns: SetDict """ return self.difference(other) def __xor__(self, other): """ The ``^`` symmetric_difference operator Args: other (Mapping): Returns: SetDict """ return self.symmetric_difference(other) # - reverse versions def __ror__(self, other): """ Args: other (Mapping): Returns: dict Example: >>> import ubelt as ub >>> self = ub.sdict({1: 1, 2: 2, 3: 3}) >>> other = {1: 10, 2:20, 4: 40} >>> d1 = self | other >>> d2 = other | self >>> assert isinstance(d1, ub.SetDict), 'should use own type' >>> assert isinstance(d2, ub.SetDict), 'should promote type' >>> print(f'd1={d1}') >>> print(f'd2={d2}') d1={1: 10, 2: 20, 3: 3, 4: 40} d2={1: 1, 2: 2, 4: 40, 3: 3} """ return SetDict.union(other, self, cls=self.__class__) def __rand__(self, other): """ Args: other (Mapping): Returns: dict Example: >>> import ubelt as ub >>> self = ub.sdict({1: 1, 2: 2, 3: 3}) >>> other = {1: 10, 2:20, 4: 40} >>> d1 = self & other >>> d2 = other & self >>> assert isinstance(d1, ub.SetDict), 'should use own type' >>> assert isinstance(d2, ub.SetDict), 'should promote type' >>> print(f'd1={d1}') >>> print(f'd2={d2}') d1={1: 1, 2: 2} d2={1: 10, 2: 20} """ return SetDict.intersection(other, self, cls=self.__class__) def __rsub__(self, other): """ Args: other (Mapping): Returns: dict Example: >>> import ubelt as ub >>> self = ub.sdict({1: 1, 2: 2, 3: 3}) >>> other = {1: 10, 2:20, 4: 40} >>> d1 = self - other >>> d2 = other - self >>> assert isinstance(d1, ub.SetDict), 'should use own type' >>> assert isinstance(d2, ub.SetDict), 'should promote type' >>> print(f'd1={d1}') >>> print(f'd2={d2}') d1={3: 3} d2={4: 40} """ return SetDict.difference(other, self, cls=self.__class__) def __rxor__(self, other): """ Args: other (Mapping): Returns: dict Example: >>> import ubelt as ub >>> self = ub.sdict({1: 1, 2: 2, 3: 3}) >>> other = {1: 10, 2:20, 4: 40} >>> d1 = self ^ other >>> d2 = other ^ self >>> assert isinstance(d1, ub.SetDict), 'should use own type' >>> assert isinstance(d2, ub.SetDict), 'should promote type' >>> print(f'd1={d1}') >>> print(f'd2={d2}') d1={3: 3, 4: 40} d2={4: 40, 3: 3} """ return SetDict.symmetric_difference(other, self, cls=self.__class__) # - inplace versions def __ior__(self, other): """ The inplace union operator ``|=``. Args: other (SupportsKeysAndGetItem[Any, Any] | Iterable[Tuple[Any, Any]]): Returns: SetDict Example: >>> import ubelt as ub >>> self = orig_ref = ub.sdict({1: 1, 2: 2, 3: 3}) >>> orig_val = orig_ref.copy() >>> other = {1: 10, 2:20, 4: 40} >>> self |= other >>> print(f'self={self}') >>> assert self is orig_ref >>> assert self == (orig_val | other) self={1: 10, 2: 20, 3: 3, 4: 40} """ self.update(other) return self def __iand__(self, other): """ The inplace intersection operator ``&=``. Args: other (Mapping | Iterable): Example: >>> import ubelt as ub >>> self = orig_ref = ub.sdict({1: 1, 2: 2, 3: 3}) >>> orig_val = orig_ref.copy() >>> other = {1: 10, 2:20, 4: 40} >>> self &= other >>> print(f'self={self}') >>> assert self is orig_ref >>> assert self == (orig_val & other) self={1: 1, 2: 2} """ remove_keys = self.keys() - set(other) for k in remove_keys: del self[k] return self def __isub__(self, other): """ The inplace difference operator ``-=``. Args: other (Mapping | Iterable): Example: >>> import ubelt as ub >>> self = orig_ref = ub.sdict({1: 1, 2: 2, 3: 3}) >>> orig_val = orig_ref.copy() >>> other = {1: 10, 2:20} >>> self -= other >>> print(f'self={self}') >>> assert self is orig_ref >>> assert self == (orig_val - other) self={3: 3} >>> import ubelt as ub >>> self = orig_ref = ub.sdict({1: 1, 2: 2, 3: 3}) >>> orig_val = orig_ref.copy() >>> other = [1] >>> self -= other >>> print(f'self={self}') >>> assert self is orig_ref >>> assert self == (orig_val - other) self={2: 2, 3: 3} >>> import ubelt as ub >>> self = orig_ref = ub.sdict({1: 1, 2: 2, 3: 3}) >>> orig_val = orig_ref.copy() >>> other = {1: 10, 2:20, 4: 40} >>> self -= other >>> print(f'self={self}') >>> assert self is orig_ref >>> assert self == (orig_val - other) """ remove_keys = self.keys() & set(other) for k in remove_keys: del self[k] return self def __ixor__(self, other): """ The inplace symmetric difference operator ``^=``. Args: other (Mapping): Example: >>> import ubelt as ub >>> self = orig_ref = ub.sdict({1: 1, 2: 2, 3: 3}) >>> orig_val = orig_ref.copy() >>> other = {1: 10, 2:20, 4: 40} >>> self ^= other >>> print(f'self={self}') >>> assert self is orig_ref >>> assert self == (orig_val ^ other) """ other_keys = set(other.keys()) remove_keys = self.keys() & other_keys add_keys = other_keys - remove_keys for k in remove_keys: del self[k] for k in add_keys: self[k] = other[k] return self ### Main set operations
[docs] def union(self, *others, cls=None, merge=None): """ Return the key-wise union of two or more dictionaries. Values chosen with *right-most* priority. I.e. for items with intersecting keys, dictionaries towards the end of the sequence are given precedence. Args: self (SetDict | dict): if called as a static method this must be provided. *others : other dictionary like objects that have an ``items`` method. (i.e. it must return an iterable of 2-tuples where the first item is hashable.) cls (type | None): the desired return dictionary type. merge (None | Callable): if specified this function must accept an iterable of values and return a new value to use (which typically is derived from input values). NotImplemented, help wanted. Returns: dict : items from all input dictionaries. Conflicts are resolved with right-most priority unless ``merge`` is specified. Specific return type is specified by ``cls`` or defaults to the leftmost input. Example: >>> import ubelt as ub >>> a = ub.SetDict({k: 'A_' + chr(97 + k) for k in [2, 3, 5, 7]}) >>> b = ub.SetDict({k: 'B_' + chr(97 + k) for k in [2, 4, 0, 7]}) >>> c = ub.SetDict({k: 'C_' + chr(97 + k) for k in [2, 8, 3]}) >>> d = ub.SetDict({k: 'D_' + chr(97 + k) for k in [9, 10, 11]}) >>> e = ub.SetDict({k: 'E_' + chr(97 + k) for k in []}) >>> assert a | b == {2: 'B_c', 3: 'A_d', 5: 'A_f', 7: 'B_h', 4: 'B_e', 0: 'B_a'} >>> a.union(b) >>> a | b | c >>> res = ub.SetDict.union(a, b, c, d, e) >>> print(ub.repr2(res, sort=1, nl=0, si=1)) {0: B_a, 2: C_c, 3: C_d, 4: B_e, 5: A_f, 7: B_h, 8: C_i, 9: D_j, 10: D_k, 11: D_l} """ if cls is None: # Some subclass-constructors need special handling # Not sure if it is in-scope to do that here or not. # if isinstance(self.__class__, defaultdict): # ... cls = self.__class__ args = it.chain([self], others) if merge is None: new = cls(it.chain.from_iterable(d.items() for d in args)) else: raise NotImplementedError('merge function is not yet implemented') return new
[docs] def intersection(self, *others, cls=None, merge=None): """ Return the key-wise intersection of two or more dictionaries. Values returned with *left-most* priority. I.e. all items returned will be from the first dictionary for keys that exist in all other dictionaries / sets provided. Args: self (SetDict | dict): if called as a static method this must be provided. *others : other dictionary or set like objects that can be coerced into a set of keys. cls (type | None): the desired return dictionary type. merge (None | Callable): if specified this function must accept an iterable of values and return a new value to use (which typically is derived from input values). NotImplemented, help wanted. Returns: dict : items with keys shared by all the inputs. Values take left-most priority unless ``merge`` is specified. Specific return type is specified by ``cls`` or defaults to the leftmost input. Example: >>> import ubelt as ub >>> a = ub.SetDict({'a': 1, 'b': 2, 'd': 4}) >>> b = ub.SetDict({'a': 10, 'b': 20, 'c': 30}) >>> a.intersection(b) {'a': 1, 'b': 2} >>> a & b {'a': 1, 'b': 2} Example: >>> import ubelt as ub >>> a = ub.SetDict({k: 'A_' + chr(97 + k) for k in [2, 3, 5, 7]}) >>> b = ub.SetDict({k: 'B_' + chr(97 + k) for k in [2, 4, 0, 7]}) >>> c = ub.SetDict({k: 'C_' + chr(97 + k) for k in [2, 8, 3]}) >>> d = ub.SetDict({k: 'D_' + chr(97 + k) for k in [9, 10, 11]}) >>> e = ub.SetDict({k: 'E_' + chr(97 + k) for k in []}) >>> assert a & b == {2: 'A_c', 7: 'A_h'} >>> a.intersection(b) >>> a & b & c >>> res = ub.SetDict.intersection(a, b, c, d, e) >>> print(ub.repr2(res, sort=1, nl=0, si=1)) {} """ cls = cls or self.__class__ isect_keys = set(self.keys()) for v in others: isect_keys.intersection_update(v) if merge is None: new = cls((k, self[k]) for k in self if k in isect_keys) else: raise NotImplementedError('merge function is not yet implemented') return new
[docs] def difference(self, *others, cls=None, merge=None): """ Return the key-wise difference between this dictionary and one or more other dictionary / keys. Values returned with *left-most* priority. I.e. the returned items will be from the first dictionary, and will only contain keys that do not appear in any of the other dictionaries / sets. Args: self (SetDict | dict): if called as a static method this must be provided. *others : other dictionary or set like objects that can be coerced into a set of keys. cls (type | None): the desired return dictionary type. merge (None | Callable): if specified this function must accept an iterable of values and return a new value to use (which typically is derived from input values). NotImplemented, help wanted. Returns: dict : items from the first dictionary with keys not in any of the following inputs. Values take left-most priority unless ``merge`` is specified. Specific return type is specified by ``cls`` or defaults to the leftmost input. Example: >>> import ubelt as ub >>> a = ub.SetDict({k: 'A_' + chr(97 + k) for k in [2, 3, 5, 7]}) >>> b = ub.SetDict({k: 'B_' + chr(97 + k) for k in [2, 4, 0, 7]}) >>> c = ub.SetDict({k: 'C_' + chr(97 + k) for k in [2, 8, 3]}) >>> d = ub.SetDict({k: 'D_' + chr(97 + k) for k in [9, 10, 11]}) >>> e = ub.SetDict({k: 'E_' + chr(97 + k) for k in []}) >>> assert a - b == {3: 'A_d', 5: 'A_f'} >>> a.difference(b) >>> a - b - c >>> res = ub.SetDict.difference(a, b, c, d, e) >>> print(ub.repr2(res, sort=1, nl=0, si=1)) {5: A_f} """ cls = cls or self.__class__ other_keys = set() for v in others: other_keys.update(v) if merge is None: # Looping over original keys is important to maintain partial order. new = cls((k, self[k]) for k in self.keys() if k not in other_keys) else: raise NotImplementedError('merge function is not yet implemented') return new
[docs] def symmetric_difference(self, *others, cls=None, merge=None): """ Return the key-wise symmetric difference between this dictionary and one or more other dictionaries. Values chosen with *right-most* priority. Returns items that are (key-wise) in an odd number of the given dictionaries. This is consistent with the standard n-ary definition of symmetric difference [WikiSymDiff]_ and corresponds with the xor operation. Args: self (SetDict | dict): if called as a static method this must be provided. *others : other dictionary or set like objects that can be coerced into a set of keys. cls (type | None): the desired return dictionary type. merge (None | Callable): if specified this function must accept an iterable of values and return a new value to use (which typically is derived from input values). NotImplemented, help wanted. Returns: dict : items from input dictionaries where the key appears an odd number of times. Values take right-most priority unless ``merge`` is specified. Specific return type is specified by ``cls`` or defaults to the leftmost input. References: .. [WikiSymDiff] https://en.wikipedia.org/wiki/Symmetric_difference Example: >>> import ubelt as ub >>> a = ub.SetDict({k: 'A_' + chr(97 + k) for k in [2, 3, 5, 7]}) >>> b = ub.SetDict({k: 'B_' + chr(97 + k) for k in [2, 4, 0, 7]}) >>> c = ub.SetDict({k: 'C_' + chr(97 + k) for k in [2, 8, 3]}) >>> d = ub.SetDict({k: 'D_' + chr(97 + k) for k in [9, 10, 11]}) >>> e = ub.SetDict({k: 'E_' + chr(97 + k) for k in []}) >>> a ^ b {3: 'A_d', 5: 'A_f', 4: 'B_e', 0: 'B_a'} >>> a.symmetric_difference(b) >>> a - b - c >>> res = ub.SetDict.symmetric_difference(a, b, c, d, e) >>> print(ub.repr2(res, sort=1, nl=0, si=1)) {0: B_a, 2: C_c, 4: B_e, 5: A_f, 8: C_i, 9: D_j, 10: D_k, 11: D_l} """ cls = cls or self.__class__ new = cls(self) # shallow copy if merge is None: for d in others: for k, v in d.items(): if k in new: new.pop(k) else: new[k] = v else: raise NotImplementedError('merge function is not yet implemented') return new
sdict = SetDict # Might need to make these mixins for 3.6
[docs] class UDict(SetDict): """ A subclass of dict with ubelt enhancements This builds on top of :class:`SetDict` which itself is a simple extension that contains only that extra functionality. The extra invert, map, sorted, and peek functions are less fundamental and there are at least reasonable workarounds when they are not available. The UDict class is a simple subclass of dict that provides the following upgrades: * set operations - inherited from :class:`SetDict` + intersection - find items in common + union - merge dicts + difference - find items in one but not the other + symmetric_difference - find items that appear an odd number of times * subdict - take a subset with optional default values. (similar to intersection, but the later ignores non-common values) * inversion - + invert - swaps a dictionary keys and values (with options for dealing with duplicates). * mapping - + map_keys - applies a function over each key and keeps the values the same + map_values - applies a function over each key and keeps the values the same * sorting - + sorted_keys - returns a dictionary ordered by the keys + sorted_values - returns a dictionary ordered by the values IMO key-wise set operations on dictionaries are fundamentaly and sorely missing from the stdlib, mapping is super convinient, sorting and inversion are less common, but still useful to have. TODO: - [ ] UbeltDict, UltraDict, not sure what the name is. We may just rename this to Dict, Example: >>> import ubelt as ub >>> a = ub.udict({1: 20, 2: 20, 3: 30, 4: 40}) >>> b = ub.udict({0: 0, 2: 20, 4: 42}) >>> c = ub.udict({3: -1, 5: -1}) >>> # Demo key-wise set operations >>> assert a & b == {2: 20, 4: 40} >>> assert a - b == {1: 20, 3: 30} >>> assert a ^ b == {1: 20, 3: 30, 0: 0} >>> assert a | b == {1: 20, 2: 20, 3: 30, 4: 42, 0: 0} >>> # Demo new n-ary set methods >>> a.union(b, c) == {1: 20, 2: 20, 3: -1, 4: 42, 0: 0, 5: -1} >>> a.intersection(b, c) == {} >>> a.difference(b, c) == {1: 20} >>> a.symmetric_difference(b, c) == {1: 20, 0: 0, 5: -1} >>> # Demo new quality of life methods >>> assert a.subdict({2, 4, 6, 8}, default=None) == {8: None, 2: 20, 4: 40, 6: None} >>> assert a.invert() == {20: 2, 30: 3, 40: 4} >>> assert a.invert(unique_vals=0) == {20: {1, 2}, 30: {3}, 40: {4}} >>> assert a.peek_key() == ub.peek(a.keys()) >>> assert a.peek_value() == ub.peek(a.values()) >>> assert a.map_keys(lambda x: x * 10) == {10: 20, 20: 20, 30: 30, 40: 40} >>> assert a.map_values(lambda x: x * 10) == {1: 200, 2: 200, 3: 300, 4: 400} """
[docs] def subdict(self, keys, default=NoParam): """ Get a subset of a dictionary Args: self (Dict[KT, VT]): dictionary or the implicit instance keys (Iterable[KT]): keys to take from ``self`` default (Optional[object] | NoParamType): if specified uses default if keys are missing. Raises: KeyError : if a key does not exist and default is not specified SeeAlso: :func:`ubelt.util_dict.dict_subset` :func:`ubelt.UDict.take` Example: >>> import ubelt as ub >>> a = ub.udict({k: 'A_' + chr(97 + k) for k in [2, 3, 5, 7]}) >>> s = a.subdict({2, 5}) >>> print('s = {}'.format(ub.repr2(s, nl=0, sort=1))) s = {2: 'A_c', 5: 'A_f'} >>> import pytest >>> with pytest.raises(KeyError): >>> s = a.subdict({2, 5, 100}) >>> s = a.subdict({2, 5, 100}, default='DEF') >>> print('s = {}'.format(ub.repr2(s, nl=0, sort=1))) s = {2: 'A_c', 5: 'A_f', 100: 'DEF'} """ cls = self.__class__ if default is NoParam: new = cls([(k, self[k]) for k in keys]) else: new = cls([(k, self.get(k, default)) for k in keys]) return new
[docs] def take(self, keys, default=NoParam): """ Get values of an iterable of keys. Args: self (Dict[KT, VT]): dictionary or the implicit instance keys (Iterable[KT]): keys to take from ``self`` default (Optional[object] | NoParamType): if specified uses default if keys are missing. Yields: VT: a selected value within the dictionary Raises: KeyError : if a key does not exist and default is not specified SeeAlso: :func:`ubelt.util_list.take` :func:`ubelt.UDict.subdict` Example: >>> import ubelt as ub >>> a = ub.udict({k: 'A_' + chr(97 + k) for k in [2, 3, 5, 7]}) >>> s = list(a.take({2, 5})) >>> print('s = {}'.format(ub.repr2(s, nl=0, sort=1))) s = ['A_c', 'A_f'] >>> import pytest >>> with pytest.raises(KeyError): >>> s = a.subdict({2, 5, 100}) >>> s = list(a.take({2, 5, 100}, default='DEF')) >>> print('s = {}'.format(ub.repr2(s, nl=0, sort=1))) s = ['A_c', 'A_f', 'DEF'] """ if default is NoParam: for k in keys: yield self[k] else: for k in keys: yield self.get(k, default)
[docs] def invert(self, unique_vals=True): """ Swaps the keys and values in a dictionary. Args: self (Dict[KT, VT]): dictionary or the implicit instance to invert unique_vals (bool, default=True): if False, the values of the new dictionary are sets of the original keys. cls (type | None): specifies the dict subclassof the result. if unspecified will be dict or OrderedDict. This behavior may change. Returns: Dict[VT, KT] | Dict[VT, Set[KT]]: the inverted dictionary Note: The must values be hashable. If the original dictionary contains duplicate values, then only one of the corresponding keys will be returned and the others will be discarded. This can be prevented by setting ``unique_vals=False``, causing the inverted keys to be returned in a set. Example: >>> import ubelt as ub >>> inverted = ub.udict({'a': 1, 'b': 2}).invert() >>> assert inverted == {1: 'a', 2: 'b'} """ return invert_dict(self, unique_vals=unique_vals, cls=self.__class__)
[docs] def map_keys(self, func): """ Apply a function to every value in a dictionary. Creates a new dictionary with the same keys and modified values. Args: self (Dict[KT, VT]): a dictionary or the implicit instance. func (Callable[[VT], T] | Mapping[VT, T]): a function or indexable object Returns: Dict[KT, T]: transformed dictionary Example: >>> import ubelt as ub >>> new = ub.udict({'a': [1, 2, 3], 'b': []}).map_keys(ord) >>> assert new == {97: [1, 2, 3], 98: []} """ return map_keys(func, self, cls=self.__class__)
[docs] def map_values(self, func): """ Apply a function to every value in a dictionary. Creates a new dictionary with the same keys and modified values. Args: self (Dict[KT, VT]): a dictionary or the implicit instance. func (Callable[[VT], T] | Mapping[VT, T]): a function or indexable object Returns: Dict[KT, T]: transformed dictionary Example: >>> import ubelt as ub >>> newdict = ub.udict({'a': [1, 2, 3], 'b': []}).map_values(len) >>> assert newdict == {'a': 3, 'b': 0} """ return map_values(func, self, cls=self.__class__)
[docs] def sorted_keys(self, key=None, reverse=False): """ Return an ordered dictionary sorted by its keys Args: self (Dict[KT, VT]): dictionary to sort or the implicit instance. The keys must be of comparable types. key (Callable[[KT], Any] | None): If given as a callable, customizes the sorting by ordering using transformed keys. reverse (bool, default=False): if True returns in descending order Returns: OrderedDict[KT, VT]: new dictionary where the keys are ordered Example: >>> import ubelt as ub >>> new = ub.udict({'spam': 2.62, 'eggs': 1.20, 'jam': 2.92}).sorted_keys() >>> assert new == ub.odict([('eggs', 1.2), ('jam', 2.92), ('spam', 2.62)]) """ return sorted_keys(self, key=key, reverse=reverse, cls=self.__class__)
[docs] def sorted_values(self, key=None, reverse=False): """ Return an ordered dictionary sorted by its values Args: self (Dict[KT, VT]): dictionary to sort or the implicit instance. The values must be of comparable types. key (Callable[[VT], Any] | None): If given as a callable, customizes the sorting by ordering using transformed values. reverse (bool, default=False): if True returns in descending order Returns: OrderedDict[KT, VT]: new dictionary where the values are ordered Example: >>> import ubelt as ub >>> new = ub.udict({'spam': 2.62, 'eggs': 1.20, 'jam': 2.92}).sorted_values() >>> assert new == ub.odict([('eggs', 1.2), ('spam', 2.62), ('jam', 2.92)]) """ return sorted_values(self, key=key, reverse=reverse, cls=self.__class__)
[docs] def peek_key(self, default=NoParam): """ Get the first key in the dictionary Args: self (Dict): a dictionary or the implicit instance default (KT | NoParamType): default item to return if the iterable is empty, otherwise a StopIteration error is raised Returns: KT: the first value or the default Example: >>> import ubelt as ub >>> assert ub.udict({1: 2}).peek_key() == 1 """ from ubelt.util_list import peek return peek(self.keys(), default=default)
[docs] def peek_value(self, default=NoParam): """ Get the first value in the dictionary Args: self (Dict[KT, VT]): a dictionary or the implicit instance default (VT | NoParamType): default item to return if the iterable is empty, otherwise a StopIteration error is raised Returns: VT: the first value or the default Example: >>> import ubelt as ub >>> assert ub.udict({1: 2}).peek_value() == 2 """ from ubelt.util_list import peek return peek(self.values(), default=default)
[docs] class AutoDict(UDict): """ An infinitely nested default dict of dicts. Implementation of Perl's autovivification feature that follows [SO_651794]_. References: .. [SO_651794] http://stackoverflow.com/questions/651794/init-dict-of-dicts Example: >>> import ubelt as ub >>> auto = ub.AutoDict() >>> auto[0][10][100] = None >>> assert str(auto) == '{0: {10: {100: None}}}' """ _base = UDict def __getitem__(self, key): """ Args: key (KT): key to lookup Returns: VT | AutoDict: an existing value or a new AutoDict """ try: # value = super(AutoDict, self).__getitem__(key) value = self._base.__getitem__(self, key) except KeyError: value = self[key] = self.__class__() return value
[docs] def to_dict(self): """ Recursively casts a AutoDict into a regular dictionary. All directly nested AutoDict values are also converted. This effectively de-defaults the structure. Returns: dict: a copy of this dict without autovivification Example: >>> import ubelt as ub >>> auto = ub.AutoDict() >>> auto[1] = 1 >>> auto['n1'] = ub.AutoDict() >>> static = auto.to_dict() >>> assert not isinstance(static, ub.AutoDict) >>> assert not isinstance(static['n1'], ub.AutoDict) Example: >>> import ubelt as ub >>> auto = ub.AutoOrderedDict() >>> auto[0][3] = 3 >>> auto[0][2] = 2 >>> auto[0][1] = 1 >>> assert list(auto[0].values()) == [3, 2, 1] """ return self._base( (key, (value.to_dict() if isinstance(value, AutoDict) else value)) for key, value in self.items())
# DEPRECATED. This is no longer needed. AutoDict is always ordered AutoOrderedDict = AutoDict udict = UDict