1
|
|
|
""" |
2
|
|
|
ETLT |
3
|
|
|
|
4
|
|
|
Copyright 2016 Set Based IT Consultancy |
5
|
|
|
|
6
|
|
|
Licence MIT |
7
|
|
|
""" |
8
|
1 |
|
from etlt.helper.Allen import Allen |
9
|
1 |
|
from etlt.helper.Type2Helper import Type2Helper |
10
|
|
|
|
11
|
|
|
|
12
|
1 |
|
class Type2JoinHelper(Type2Helper): |
13
|
|
|
""" |
14
|
|
|
A helper class for joining data sets with date intervals. |
15
|
|
|
""" |
16
|
|
|
|
17
|
|
|
# ------------------------------------------------------------------------------------------------------------------ |
18
|
1 |
|
def _equal(self, row1, row2): |
19
|
|
|
""" |
20
|
|
|
Returns True if two rows are identical excluding start and end date. Returns False otherwise. |
21
|
|
|
|
22
|
|
|
:param dict[str,T] row1: The first row. |
23
|
|
|
:param dict[str,T] row2: The second row. |
24
|
|
|
|
25
|
|
|
:rtype: bool |
26
|
|
|
""" |
27
|
1 |
|
for key in row1.keys(): |
28
|
1 |
|
if key not in [self._key_start_date, self._key_end_date]: |
29
|
1 |
|
if row1[key] != row2[key]: |
30
|
1 |
|
return False |
31
|
|
|
|
32
|
1 |
|
return True |
33
|
|
|
|
34
|
|
|
# ------------------------------------------------------------------------------------------------------------------ |
35
|
1 |
|
@staticmethod |
36
|
|
|
def _intersect(start1, end1, start2, end2): |
37
|
|
|
""" |
38
|
|
|
Returns the intersection of two intervals. Returns (None,None) if the intersection is empty. |
39
|
|
|
|
40
|
|
|
:param int start1: The start date of the first interval. |
41
|
|
|
:param int end1: The end date of the first interval. |
42
|
|
|
:param int start2: The start date of the second interval. |
43
|
|
|
:param int end2: The end date of the second interval. |
44
|
|
|
|
45
|
|
|
:rtype: tuple[int|None,int|None] |
46
|
|
|
""" |
47
|
1 |
|
start = max(start1, start2) |
48
|
1 |
|
end = min(end1, end2) |
49
|
|
|
|
50
|
1 |
|
if start > end: |
51
|
1 |
|
return None, None |
52
|
|
|
|
53
|
1 |
|
return start, end |
54
|
|
|
|
55
|
|
|
# ------------------------------------------------------------------------------------------------------------------ |
56
|
1 |
|
def _additional_rows_date2int(self, keys, rows): |
57
|
|
|
""" |
58
|
|
|
Replaces start and end dates of the additional date intervals in the row set with their integer representation |
59
|
|
|
|
60
|
|
|
:param list[tuple[str,str]] keys: The other keys with start and end date. |
61
|
|
|
:param list[dict[str,T]] rows: The list of rows. |
62
|
|
|
|
63
|
|
|
:rtype: list[dict[str,T]] |
64
|
|
|
""" |
65
|
1 |
|
for row in rows: |
66
|
1 |
|
for key_start_date, key_end_date in keys: |
67
|
1 |
|
if key_start_date not in [self._key_start_date, self._key_end_date]: |
68
|
1 |
|
row[key_start_date] = self._date2int(row[key_start_date]) |
69
|
1 |
|
if key_end_date not in [self._key_start_date, self._key_end_date]: |
70
|
1 |
|
row[key_end_date] = self._date2int(row[key_end_date]) |
71
|
|
|
|
72
|
|
|
# ------------------------------------------------------------------------------------------------------------------ |
73
|
1 |
|
def _intersection(self, keys, rows): |
74
|
|
|
""" |
75
|
|
|
Computes the intersection of the date intervals of two or more reference data sets. If the intersection is empty |
76
|
|
|
the row is removed from the group. |
77
|
|
|
|
78
|
|
|
:param list[tuple[str,str]] keys: The other keys with start and end date. |
79
|
|
|
:param list[dict[str,T]] rows: The list of rows. |
80
|
|
|
|
81
|
|
|
:rtype: list[dict[str,T]] |
82
|
|
|
""" |
83
|
1 |
|
ret = list() |
84
|
1 |
|
for row in rows: |
85
|
1 |
|
start_date = row[self._key_start_date] |
86
|
1 |
|
end_date = row[self._key_end_date] |
87
|
1 |
|
for key_start_date, key_end_date in keys: |
88
|
1 |
|
start_date, end_date = Type2JoinHelper._intersect(start_date, |
89
|
|
|
end_date, |
90
|
|
|
row[key_start_date], |
91
|
|
|
row[key_end_date]) |
92
|
1 |
|
if not start_date: |
93
|
1 |
|
break |
94
|
1 |
|
if key_start_date not in [self._key_start_date, self._key_end_date]: |
95
|
1 |
|
del row[key_start_date] |
96
|
1 |
|
if key_end_date not in [self._key_start_date, self._key_end_date]: |
97
|
1 |
|
del row[key_end_date] |
98
|
|
|
|
99
|
1 |
|
if start_date: |
100
|
1 |
|
row[self._key_start_date] = start_date |
101
|
1 |
|
row[self._key_end_date] = end_date |
102
|
1 |
|
ret.append(row) |
103
|
|
|
|
104
|
1 |
|
return ret |
105
|
|
|
|
106
|
|
|
# ------------------------------------------------------------------------------------------------------------------ |
107
|
1 |
|
def _merge_adjacent_rows(self, rows): |
108
|
|
|
""" |
109
|
|
|
Resolves adjacent and overlapping rows. With proper reference data overlapping rows MUST not occur. However, |
110
|
|
|
this method can handle overlapping rows. Overlapping rows are resolved as follows: |
111
|
|
|
* The interval with the most recent begin date prevails for the overlapping period. |
112
|
|
|
* If the begin dates are the same the interval with the most recent end date prevails. |
113
|
|
|
* If the begin and end dates are equal the last row in the data set prevails. |
114
|
|
|
Identical (excluding begin and end date) adjacent rows are replace with a single row. |
115
|
|
|
|
116
|
|
|
:param list[dict[str,T]] rows: The rows in a group (i.e. with the same natural key). |
117
|
|
|
. |
118
|
|
|
:rtype: list[dict[str,T]] |
119
|
|
|
""" |
120
|
1 |
|
ret = list() |
121
|
|
|
|
122
|
1 |
|
prev_row = None |
123
|
1 |
|
for row in rows: |
124
|
1 |
|
if prev_row: |
125
|
1 |
|
relation = Allen.relation(prev_row[self._key_start_date], |
126
|
|
|
prev_row[self._key_end_date], |
127
|
|
|
row[self._key_start_date], |
128
|
|
|
row[self._key_end_date]) |
129
|
1 |
|
if relation == Allen.X_BEFORE_Y: |
130
|
|
|
# Two rows with distinct intervals. |
131
|
|
|
# prev_row: |----| |
132
|
|
|
# row: |-----| |
133
|
1 |
|
ret.append(prev_row) |
134
|
1 |
|
prev_row = row |
135
|
|
|
|
136
|
1 |
|
elif relation == Allen.X_MEETS_Y: |
137
|
|
|
# The two rows are adjacent. |
138
|
|
|
# prev_row: |-------| |
139
|
|
|
# row: |-------| |
140
|
1 |
|
if self._equal(prev_row, row): |
141
|
|
|
# The two rows are identical (except for start and end date) and adjacent. Combine the two rows |
142
|
|
|
# into one row. |
143
|
1 |
|
prev_row[self._key_end_date] = row[self._key_end_date] |
144
|
|
|
else: |
145
|
|
|
# Rows are adjacent but not identical. |
146
|
1 |
|
ret.append(prev_row) |
147
|
1 |
|
prev_row = row |
148
|
|
|
|
149
|
1 |
|
elif relation == Allen.X_OVERLAPS_WITH_Y: |
150
|
|
|
# prev_row overlaps row. Should not occur with proper reference data. |
151
|
|
|
# prev_row: |-----------| |
152
|
|
|
# row: |----------| |
153
|
1 |
|
if self._equal(prev_row, row): |
154
|
|
|
# The two rows are identical (except for start and end date) and overlapping. Combine the two |
155
|
|
|
# rows into one row. |
156
|
1 |
|
prev_row[self._key_end_date] = row[self._key_end_date] |
157
|
|
|
else: |
158
|
|
|
# Rows are overlapping but not identical. |
159
|
1 |
|
prev_row[self._key_end_date] = row[self._key_start_date] - 1 |
160
|
1 |
|
ret.append(prev_row) |
161
|
1 |
|
prev_row = row |
162
|
|
|
|
163
|
1 |
|
elif relation == Allen.X_STARTS_Y: |
164
|
|
|
# prev_row start row. Should not occur with proper reference data. |
165
|
|
|
# prev_row: |------| |
166
|
|
|
# row: |----------------| |
167
|
1 |
|
prev_row = row |
168
|
|
|
|
169
|
1 |
|
elif relation == Allen.X_EQUAL_Y: |
170
|
|
|
# Can happen when the reference data sets are joined without respect for date intervals. |
171
|
|
|
# prev_row: |----------------| |
172
|
|
|
# row: |----------------| |
173
|
1 |
|
prev_row = row |
174
|
|
|
|
175
|
1 |
|
elif relation == Allen.X_DURING_Y_INVERSE: |
176
|
|
|
# row during prev_row. Should not occur with proper reference data. |
177
|
|
|
# prev_row: |----------------| |
178
|
|
|
# row: |------| |
179
|
|
|
# Note: the interval with the most recent start date prevails. Hence, the interval after |
180
|
|
|
# row[self._key_end_date] is discarded. |
181
|
1 |
|
if self._equal(prev_row, row): |
182
|
1 |
|
prev_row[self._key_end_date] = row[self._key_end_date] |
183
|
|
|
else: |
184
|
1 |
|
prev_row[self._key_end_date] = row[self._key_start_date] - 1 |
185
|
1 |
|
ret.append(prev_row) |
186
|
1 |
|
prev_row = row |
187
|
|
|
|
188
|
1 |
|
elif relation == Allen.X_FINISHES_Y_INVERSE: |
189
|
|
|
# row finishes prev_row. Should not occur with proper reference data. |
190
|
|
|
# prev_row: |----------------| |
191
|
|
|
# row: |------| |
192
|
1 |
|
if not self._equal(prev_row, row): |
193
|
1 |
|
prev_row[self._key_end_date] = row[self._key_start_date] - 1 |
194
|
1 |
|
ret.append(prev_row) |
195
|
1 |
|
prev_row = row |
196
|
|
|
|
197
|
|
|
# Note: if the two rows are identical (except for start and end date) nothing to do. |
198
|
|
|
else: |
199
|
|
|
# Note: The rows are sorted such that prev_row[self._key_begin_date] <= row[self._key_begin_date]. |
200
|
|
|
# Hence the following relation should not occur: X_DURING_Y, X_FINISHES_Y, X_BEFORE_Y_INVERSE, |
201
|
|
|
# X_MEETS_Y_INVERSE, X_OVERLAPS_WITH_Y_INVERSE, and X_STARTS_Y_INVERSE. Hence, we covered all 13 |
202
|
|
|
# relations in Allen's interval algebra. |
203
|
|
|
raise ValueError('Data is not sorted properly. Relation: %d' % relation) |
204
|
|
|
else: |
205
|
1 |
|
prev_row = row |
206
|
|
|
|
207
|
1 |
|
if prev_row: |
208
|
1 |
|
ret.append(prev_row) |
209
|
|
|
|
210
|
1 |
|
return ret |
211
|
|
|
|
212
|
|
|
# ------------------------------------------------------------------------------------------------------------------ |
213
|
1 |
|
def merge(self, keys): |
214
|
|
|
""" |
215
|
|
|
Merges the join on natural keys of two or more reference data sets. |
216
|
|
|
|
217
|
|
|
:param list[tuple[str,str]] keys: For each data set the keys of the start and end date. |
218
|
|
|
""" |
219
|
1 |
|
deletes = [] |
220
|
1 |
|
for natural_key, rows in self.rows.items(): |
221
|
1 |
|
self._additional_rows_date2int(keys, rows) |
222
|
1 |
|
rows = self._intersection(keys, rows) |
223
|
1 |
|
if rows: |
224
|
1 |
|
rows = self._rows_sort(rows) |
225
|
1 |
|
self.rows[natural_key] = self._merge_adjacent_rows(rows) |
226
|
|
|
else: |
227
|
1 |
|
deletes.append(natural_key) |
228
|
|
|
|
229
|
1 |
|
for natural_key in deletes: |
230
|
1 |
|
del self.rows[natural_key] |
231
|
|
|
|
232
|
|
|
# ---------------------------------------------------------------------------------------------------------------------- |
233
|
|
|
|