1
|
|
|
from itertools import chain |
2
|
|
|
from numbers import Real, Integral |
3
|
|
|
from ..data.value import Value, Unknown |
4
|
|
|
from math import isnan |
5
|
|
|
import numpy as np |
|
|
|
|
6
|
|
|
|
7
|
|
|
|
8
|
|
|
class Instance: |
9
|
|
|
def __init__(self, domain, data=None, id=None): |
|
|
|
|
10
|
|
|
""" |
11
|
|
|
Construct a new data instance. |
12
|
|
|
|
13
|
|
|
:param domain: domain that describes the instance's variables |
14
|
|
|
:type domain: Orange.data.Domain |
15
|
|
|
:param data: instance's values |
16
|
|
|
:type data: Orange.data.Instance or a sequence of values |
17
|
|
|
:param id: instance id |
18
|
|
|
:type id: hashable value |
19
|
|
|
""" |
20
|
|
|
if data is None and isinstance(domain, Instance): |
21
|
|
|
data = domain |
22
|
|
|
domain = data.domain |
23
|
|
|
|
24
|
|
|
self._domain = domain |
25
|
|
|
if data is None: |
26
|
|
|
self._x = np.repeat(Unknown, len(domain.attributes)) |
27
|
|
|
self._y = np.repeat(Unknown, len(domain.class_vars)) |
28
|
|
|
self._metas = np.array([var.Unknown for var in domain.metas], |
29
|
|
|
dtype=object) |
30
|
|
|
self._weight = 1 |
31
|
|
|
elif isinstance(data, Instance) and data.domain == domain: |
32
|
|
|
self._x = np.array(data._x) |
|
|
|
|
33
|
|
|
self._y = np.array(data._y) |
|
|
|
|
34
|
|
|
self._metas = np.array(data._metas) |
|
|
|
|
35
|
|
|
self._weight = data._weight |
|
|
|
|
36
|
|
|
else: |
37
|
|
|
self._x, self._y, self._metas = domain.convert(data) |
38
|
|
|
self._weight = 1 |
39
|
|
|
|
40
|
|
|
if id is not None: |
41
|
|
|
self.id = id |
42
|
|
|
else: |
43
|
|
|
from Orange.data import Table |
44
|
|
|
self.id = Table.new_id() |
45
|
|
|
|
46
|
|
|
@property |
47
|
|
|
def domain(self): |
48
|
|
|
"""The domain describing the instance's values.""" |
49
|
|
|
return self._domain |
50
|
|
|
|
51
|
|
|
@property |
52
|
|
|
def x(self): |
53
|
|
|
""" |
54
|
|
|
Instance's attributes as a 1-dimensional numpy array whose length |
55
|
|
|
equals `len(self.domain.attributes)`. |
56
|
|
|
""" |
57
|
|
|
return self._x |
58
|
|
|
|
59
|
|
|
@property |
60
|
|
|
def y(self): |
61
|
|
|
""" |
62
|
|
|
Instance's classes as a 1-dimensional numpy array whose length |
63
|
|
|
equals `len(self.domain.attributes)`. |
64
|
|
|
""" |
65
|
|
|
return self._y |
66
|
|
|
|
67
|
|
|
@property |
68
|
|
|
def metas(self): |
69
|
|
|
""" |
70
|
|
|
Instance's meta attributes as a 1-dimensional numpy array whose length |
71
|
|
|
equals `len(self.domain.attributes)`. |
72
|
|
|
""" |
73
|
|
|
return self._metas |
74
|
|
|
|
75
|
|
|
@property |
76
|
|
|
def list(self): |
77
|
|
|
""" |
78
|
|
|
All instance's values, including attributes, classes and meta |
79
|
|
|
attributes, as a list whose length equals `len(self.domain.attributes) |
80
|
|
|
+ len(self.domain.class_vars) + len(self.domain.metas)`. |
81
|
|
|
""" |
82
|
|
|
n_self, n_metas = len(self), len(self._metas) |
83
|
|
|
return [self[i].value if i < n_self else self[n_self - i - 1].value |
84
|
|
|
for i in range(n_self + n_metas)] |
85
|
|
|
|
86
|
|
|
@property |
87
|
|
|
def weight(self): |
88
|
|
|
"""The weight of the data instance. Default is 1.""" |
89
|
|
|
return self._weight |
90
|
|
|
|
91
|
|
|
@weight.setter |
92
|
|
|
def weight(self, weight): |
93
|
|
|
self._weight = weight |
94
|
|
|
|
95
|
|
|
def __setitem__(self, key, value): |
96
|
|
|
if not isinstance(key, Integral): |
97
|
|
|
key = self._domain.index(key) |
98
|
|
|
value = self._domain[key].to_val(value) |
99
|
|
|
if key >= 0 and not isinstance(value, (int, float)): |
100
|
|
|
raise TypeError("Expected primitive value, got '%s'" % |
101
|
|
|
type(value).__name__) |
102
|
|
|
|
103
|
|
|
if 0 <= key < len(self._domain.attributes): |
104
|
|
|
self._x[key] = value |
105
|
|
|
elif len(self._domain.attributes) <= key: |
106
|
|
|
self._y[key - len(self.domain.attributes)] = value |
107
|
|
|
else: |
108
|
|
|
self._metas[-1 - key] = value |
109
|
|
|
|
110
|
|
|
def __getitem__(self, key, key_id=None, key_var=None): |
111
|
|
|
# key_id can explicitly be given to prevent the extra dictionary |
112
|
|
|
# lookup. The derived LazyRowInstance passes the parameter, because |
113
|
|
|
# the key_id is already looked up there. key_var is included for |
114
|
|
|
# compatibility. |
115
|
|
|
if key_id is None: |
116
|
|
|
if not isinstance(key, Integral): |
117
|
|
|
key = self._domain.index(key) |
118
|
|
|
else: |
119
|
|
|
key = key_id |
120
|
|
|
|
121
|
|
|
if 0 <= key < len(self._domain.attributes): |
122
|
|
|
value = self._x[key] |
123
|
|
|
elif key >= len(self._domain.attributes): |
124
|
|
|
value = self._y[key - len(self.domain.attributes)] |
125
|
|
|
else: |
126
|
|
|
value = self._metas[-1 - key] |
127
|
|
|
return Value(self._domain[key], value) |
128
|
|
|
|
129
|
|
|
#TODO Should we return an instance of `object` if we have a meta attribute |
|
|
|
|
130
|
|
|
# that is not Discrete or Continuous? E.g. when we have strings, we'd |
131
|
|
|
# like to be able to use startswith, lower etc... |
132
|
|
|
# Or should we even return Continuous as floats and use Value only |
133
|
|
|
# for discrete attributes?! |
134
|
|
|
# Same in Table.__getitem__ |
135
|
|
|
|
136
|
|
|
@staticmethod |
137
|
|
|
def str_values(data, variables, limit=True): |
138
|
|
|
if limit: |
139
|
|
|
s = ", ".join(var.str_val(val) |
140
|
|
|
for var, val in zip(variables, data[:5])) |
141
|
|
|
if len(data) > 5: |
142
|
|
|
s += ", ..." |
143
|
|
|
return s |
144
|
|
|
else: |
145
|
|
|
return ", ".join(var.str_val(val) |
146
|
|
|
for var, val in zip(variables, data)) |
147
|
|
|
|
148
|
|
|
def _str(self, limit): |
149
|
|
|
s = "[" + self.str_values(self._x, self._domain.attributes, limit) |
150
|
|
|
if self._domain.class_vars: |
151
|
|
|
s += " | " + \ |
152
|
|
|
self.str_values(self._y, self._domain.class_vars, limit) |
153
|
|
|
s += "]" |
154
|
|
|
if self._domain.metas: |
155
|
|
|
s += " {" + \ |
156
|
|
|
self.str_values(self._metas, self._domain.metas, limit) + \ |
157
|
|
|
"}" |
158
|
|
|
return s |
159
|
|
|
|
160
|
|
|
def __str__(self): |
161
|
|
|
return self._str(False) |
162
|
|
|
|
163
|
|
|
def __repr__(self): |
164
|
|
|
return self._str(True) |
165
|
|
|
|
166
|
|
|
def __eq__(self, other): |
167
|
|
|
if not isinstance(other, Instance): |
168
|
|
|
other = Instance(self._domain, other) |
169
|
|
|
|
170
|
|
|
def same(x1, x2): |
171
|
|
|
nan1 = np.isnan(x1) |
172
|
|
|
nan2 = np.isnan(x2) |
173
|
|
|
return np.array_equal(nan1, nan2) and \ |
174
|
|
|
np.array_equal(x1[~nan1], x2[~nan2]) |
175
|
|
|
|
176
|
|
|
return same(self._x, other._x) and same(self._y, other._y) \ |
|
|
|
|
177
|
|
|
and all(m1 == m2 or |
178
|
|
|
type(m1) == type(m2) == float and isnan(m1) and isnan(m2) |
179
|
|
|
for m1, m2 in zip(self._metas, other._metas)) |
|
|
|
|
180
|
|
|
|
181
|
|
|
def __iter__(self): |
182
|
|
|
return chain(iter(self._x), iter(self._y)) |
183
|
|
|
|
184
|
|
|
def values(self): |
185
|
|
|
return (Value(var, val) |
186
|
|
|
for var, val in zip(self.domain.variables, self)) |
187
|
|
|
|
188
|
|
|
def __len__(self): |
189
|
|
|
return len(self._x) + len(self._y) |
190
|
|
|
|
191
|
|
|
def attributes(self): |
192
|
|
|
"""Return iterator over the instance's attributes""" |
193
|
|
|
return iter(self._x) |
194
|
|
|
|
195
|
|
|
def classes(self): |
196
|
|
|
"""Return iterator over the instance's class attributes""" |
197
|
|
|
return iter(self._y) |
198
|
|
|
|
199
|
|
|
# A helper function for get_class and set_class |
200
|
|
|
def _check_single_class(self): |
201
|
|
|
if not self._domain.class_vars: |
202
|
|
|
raise TypeError("Domain has no class variable") |
203
|
|
|
elif len(self._domain.class_vars) > 1: |
204
|
|
|
raise TypeError("Domain has multiple class variables") |
205
|
|
|
|
206
|
|
|
def get_class(self): |
207
|
|
|
""" |
208
|
|
|
Return the class value as an instance of :obj:`Orange.data.Value`. |
209
|
|
|
Throws an exception if there are multiple classes. |
210
|
|
|
""" |
211
|
|
|
self._check_single_class() |
212
|
|
|
return Value(self._domain.class_var, self._y[0]) |
213
|
|
|
|
214
|
|
|
def get_classes(self): |
215
|
|
|
""" |
216
|
|
|
Return the class value as a list of instances of |
217
|
|
|
:obj:`Orange.data.Value`. |
218
|
|
|
""" |
219
|
|
|
return (Value(var, value) |
220
|
|
|
for var, value in zip(self._domain.class_vars, self._y)) |
221
|
|
|
|
222
|
|
|
def set_class(self, value): |
223
|
|
|
""" |
224
|
|
|
Set the instance's class. Throws an exception if there are multiple |
225
|
|
|
classes. |
226
|
|
|
""" |
227
|
|
|
self._check_single_class() |
228
|
|
|
if not isinstance(value, Real): |
229
|
|
|
self._y[0] = self._domain.class_var.to_val(value) |
230
|
|
|
else: |
231
|
|
|
self._y[0] = value |
232
|
|
|
|
This can be caused by one of the following:
1. Missing Dependencies
This error could indicate a configuration issue of Pylint. Make sure that your libraries are available by adding the necessary commands.
2. Missing __init__.py files
This error could also result from missing
__init__.py
files in your module folders. Make sure that you place one file in each sub-folder.