""" Test extension array for storing nested data in a pandas container. The JSONArray stores lists of dictionaries. The storage mechanism is a list, not an ndarray. Note ---- We currently store lists of UserDicts. Pandas has a few places internally that specifically check for dicts, and does non-scalar things in that case. We *want* the dictionaries to be treated as scalars, so we hack around pandas by using UserDicts. """ from __future__ import annotations from collections import ( UserDict, abc, ) import itertools import numbers import random import string import sys from typing import ( Any, Mapping, ) import numpy as np from pandas._typing import type_t from pandas.core.dtypes.cast import construct_1d_object_array_from_listlike from pandas.core.dtypes.common import ( is_bool_dtype, is_list_like, pandas_dtype, ) import pandas as pd from pandas.api.extensions import ( ExtensionArray, ExtensionDtype, ) from pandas.core.indexers import unpack_tuple_and_ellipses class JSONDtype(ExtensionDtype): type = abc.Mapping name = "json" na_value: Mapping[str, Any] = UserDict() @classmethod def construct_array_type(cls) -> type_t[JSONArray]: """ Return the array type associated with this dtype. Returns ------- type """ return JSONArray class JSONArray(ExtensionArray): dtype = JSONDtype() __array_priority__ = 1000 def __init__(self, values, dtype=None, copy=False) -> None: for val in values: if not isinstance(val, self.dtype.type): raise TypeError("All values must be of type " + str(self.dtype.type)) self.data = values # Some aliases for common attribute names to ensure pandas supports # these self._items = self._data = self.data # those aliases are currently not working due to assumptions # in internal code (GH-20735) # self._values = self.values = self.data @classmethod def _from_sequence(cls, scalars, dtype=None, copy=False): return cls(scalars) @classmethod def _from_factorized(cls, values, original): return cls([UserDict(x) for x in values if x != ()]) def __getitem__(self, item): if isinstance(item, tuple): item = unpack_tuple_and_ellipses(item) if isinstance(item, numbers.Integral): return self.data[item] elif isinstance(item, slice) and item == slice(None): # Make sure we get a view return type(self)(self.data) elif isinstance(item, slice): # slice return type(self)(self.data[item]) elif not is_list_like(item): # e.g. "foo" or 2.5 # exception message copied from numpy raise IndexError( r"only integers, slices (`:`), ellipsis (`...`), numpy.newaxis " r"(`None`) and integer or boolean arrays are valid indices" ) else: item = pd.api.indexers.check_array_indexer(self, item) if is_bool_dtype(item.dtype): return self._from_sequence([x for x, m in zip(self, item) if m]) # integer return type(self)([self.data[i] for i in item]) def __setitem__(self, key, value): if isinstance(key, numbers.Integral): self.data[key] = value else: if not isinstance(value, (type(self), abc.Sequence)): # broadcast value value = itertools.cycle([value]) if isinstance(key, np.ndarray) and key.dtype == "bool": # masking for i, (k, v) in enumerate(zip(key, value)): if k: assert isinstance(v, self.dtype.type) self.data[i] = v else: for k, v in zip(key, value): assert isinstance(v, self.dtype.type) self.data[k] = v def __len__(self) -> int: return len(self.data) def __eq__(self, other): return NotImplemented def __ne__(self, other): return NotImplemented def __array__(self, dtype=None): if dtype is None: dtype = object if dtype == object: # on py38 builds it looks like numpy is inferring to a non-1D array return construct_1d_object_array_from_listlike(list(self)) return np.asarray(self.data, dtype=dtype) @property def nbytes(self) -> int: return sys.getsizeof(self.data) def isna(self): return np.array([x == self.dtype.na_value for x in self.data], dtype=bool) def take(self, indexer, allow_fill=False, fill_value=None): # re-implement here, since NumPy has trouble setting # sized objects like UserDicts into scalar slots of # an ndarary. indexer = np.asarray(indexer) msg = ( "Index is out of bounds or cannot do a " "non-empty take from an empty array." ) if allow_fill: if fill_value is None: fill_value = self.dtype.na_value # bounds check if (indexer < -1).any(): raise ValueError try: output = [ self.data[loc] if loc != -1 else fill_value for loc in indexer ] except IndexError as err: raise IndexError(msg) from err else: try: output = [self.data[loc] for loc in indexer] except IndexError as err: raise IndexError(msg) from err return self._from_sequence(output) def copy(self): return type(self)(self.data[:]) def astype(self, dtype, copy=True): # NumPy has issues when all the dicts are the same length. # np.array([UserDict(...), UserDict(...)]) fails, # but np.array([{...}, {...}]) works, so cast. from pandas.core.arrays.string_ import StringDtype dtype = pandas_dtype(dtype) # needed to add this check for the Series constructor if isinstance(dtype, type(self.dtype)) and dtype == self.dtype: if copy: return self.copy() return self elif isinstance(dtype, StringDtype): value = self.astype(str) # numpy doesn't like nested dicts return dtype.construct_array_type()._from_sequence(value, copy=False) return np.array([dict(x) for x in self], dtype=dtype, copy=copy) def unique(self): # Parent method doesn't work since np.array will try to infer # a 2-dim object. return type(self)([dict(x) for x in {tuple(d.items()) for d in self.data}]) @classmethod def _concat_same_type(cls, to_concat): data = list(itertools.chain.from_iterable(x.data for x in to_concat)) return cls(data) def _values_for_factorize(self): frozen = self._values_for_argsort() if len(frozen) == 0: # factorize_array expects 1-d array, this is a len-0 2-d array. frozen = frozen.ravel() return frozen, () def _values_for_argsort(self): # Bypass NumPy's shape inference to get a (N,) array of tuples. frozen = [tuple(x.items()) for x in self] return construct_1d_object_array_from_listlike(frozen) def make_data(): # TODO: Use a regular dict. See _NDFrameIndexer._setitem_with_indexer return [ UserDict( [ (random.choice(string.ascii_letters), random.randint(0, 100)) for _ in range(random.randint(0, 10)) ] ) for _ in range(100) ]