|
1
|
|
|
import dataclasses |
|
|
|
|
|
|
2
|
|
|
import html |
|
3
|
|
|
from dataclasses import dataclass |
|
4
|
|
|
from datetime import datetime |
|
5
|
|
|
from typing import Optional, Sequence |
|
6
|
|
|
|
|
7
|
|
|
import pandas as pd |
|
|
|
|
|
|
8
|
|
|
from typeddfs import TypedDfs |
|
|
|
|
|
|
9
|
|
|
|
|
10
|
|
|
from mandos.model import ReflectionUtils |
|
11
|
|
|
|
|
12
|
|
|
HIT_FIELD_TYPE = frozenset([str, int, float, datetime]) |
|
13
|
|
|
|
|
14
|
|
|
|
|
15
|
|
|
@dataclass(frozen=True, repr=True, order=True) |
|
16
|
|
|
class Pair: |
|
17
|
|
|
""" |
|
18
|
|
|
Predicate, object pairs. |
|
19
|
|
|
""" |
|
20
|
|
|
|
|
21
|
|
|
pred: str |
|
22
|
|
|
obj: str |
|
23
|
|
|
|
|
24
|
|
|
|
|
25
|
|
|
@dataclass(frozen=True, repr=True, order=True) |
|
26
|
|
|
class Triple: |
|
27
|
|
|
""" |
|
28
|
|
|
Usually compound, predicate, object. |
|
29
|
|
|
""" |
|
30
|
|
|
|
|
31
|
|
|
sub: str |
|
32
|
|
|
pred: str |
|
33
|
|
|
obj: str |
|
34
|
|
|
|
|
35
|
|
|
@property |
|
36
|
|
|
def n_triples(self) -> str: |
|
37
|
|
|
""" |
|
38
|
|
|
Returns a simple text statement in n-triples format. |
|
39
|
|
|
""" |
|
40
|
|
|
s = self.sub |
|
|
|
|
|
|
41
|
|
|
p = html.escape(self.pred, quote=True) |
|
|
|
|
|
|
42
|
|
|
o = html.escape(self.obj, quote=True) |
|
|
|
|
|
|
43
|
|
|
return f'"{s}" "{p}" "{o}" .' |
|
44
|
|
|
|
|
45
|
|
|
|
|
46
|
|
|
@dataclass(frozen=True, order=True, repr=True) |
|
|
|
|
|
|
47
|
|
|
class AbstractHit: |
|
48
|
|
|
""" |
|
49
|
|
|
An abstract annotation (statement type), which may support additional fields. |
|
50
|
|
|
""" |
|
51
|
|
|
|
|
52
|
|
|
record_id: Optional[str] |
|
53
|
|
|
origin_inchikey: str |
|
54
|
|
|
matched_inchikey: str |
|
55
|
|
|
compound_id: str |
|
56
|
|
|
compound_name: str |
|
57
|
|
|
predicate: str |
|
58
|
|
|
statement: str |
|
59
|
|
|
object_id: str |
|
60
|
|
|
object_name: str |
|
61
|
|
|
value: float |
|
62
|
|
|
search_key: str |
|
63
|
|
|
search_class: str |
|
64
|
|
|
data_source: str |
|
65
|
|
|
run_date: datetime |
|
66
|
|
|
cache_date: Optional[datetime] |
|
67
|
|
|
# is_hit: Optional[bool] = None |
|
68
|
|
|
# score: Optional[float] = None |
|
69
|
|
|
# x_score_1: Optional[float] = None |
|
70
|
|
|
# x_score_2: Optional[float] = None |
|
71
|
|
|
|
|
72
|
|
|
@property |
|
73
|
|
|
def hit_class(self) -> str: |
|
|
|
|
|
|
74
|
|
|
return self.__class__.__name__ |
|
75
|
|
|
|
|
76
|
|
|
@property |
|
77
|
|
|
def to_triple(self) -> Triple: |
|
|
|
|
|
|
78
|
|
|
return Triple(sub=self.origin_inchikey, pred=self.predicate, obj=self.object_name) |
|
79
|
|
|
|
|
80
|
|
|
@property |
|
81
|
|
|
def to_pair(self) -> Pair: |
|
|
|
|
|
|
82
|
|
|
return Pair(pred=self.predicate, obj=self.object_name) |
|
83
|
|
|
|
|
84
|
|
|
def __hash__(self): |
|
85
|
|
|
return hash(self.record_id) |
|
86
|
|
|
|
|
87
|
|
|
@property |
|
88
|
|
|
def universal_id(self) -> str: |
|
89
|
|
|
""" |
|
90
|
|
|
Gets an identifier (a hex key) that uniquely identifies the record by its unique attributes. |
|
91
|
|
|
Does **NOT** distinguish between hits with duplicate information and does **NOT** |
|
92
|
|
|
include ``record_id``. |
|
93
|
|
|
|
|
94
|
|
|
Returns: |
|
95
|
|
|
A 16-character hexadecimal string |
|
96
|
|
|
""" |
|
97
|
|
|
# excluding record_id only because it's not available for some hit types |
|
98
|
|
|
# we'd rather immediately see duplicates if the exist |
|
99
|
|
|
fields = { |
|
100
|
|
|
field |
|
101
|
|
|
for field in self.fields() |
|
102
|
|
|
if field |
|
103
|
|
|
not in {"record_id", "origin_inchikey", "compound_name", "search_key", "search_class"} |
|
104
|
|
|
} |
|
105
|
|
|
hexed = hex(hash(tuple([getattr(self, f) for f in fields]))) |
|
106
|
|
|
# remove negative signs -- still unique |
|
107
|
|
|
return hexed.replace("-", "").replace("0x", "") |
|
108
|
|
|
|
|
109
|
|
|
@classmethod |
|
110
|
|
|
def fields(cls) -> Sequence[str]: |
|
111
|
|
|
""" |
|
112
|
|
|
Finds the list of fields in this class by reflection. |
|
113
|
|
|
""" |
|
114
|
|
|
return [f.name for f in dataclasses.fields(cls)] |
|
115
|
|
|
|
|
116
|
|
|
|
|
117
|
|
|
HitFrame = ( |
|
118
|
|
|
TypedDfs.typed("HitFrame") |
|
119
|
|
|
.require("record_id", dtype=str) |
|
120
|
|
|
.require("inchikey", "compound_id", "compound_name", dtype=str) |
|
121
|
|
|
.require("predicate", "statement", dtype=str) |
|
122
|
|
|
.require("object_id", "object_name", dtype=str) |
|
123
|
|
|
.require("search_key", "search_class", "data_source", dtype=str) |
|
124
|
|
|
.require("hit_class", dtype=str) |
|
125
|
|
|
.require("cache_date", "run_date", dtype=str) |
|
126
|
|
|
.reserve("is_hit", dtype=bool) |
|
127
|
|
|
.reserve("score", *[f"x_score_{i}" for i in range(1, 10)], dtype=float) |
|
128
|
|
|
).build() |
|
129
|
|
|
|
|
130
|
|
|
|
|
131
|
|
|
class HitUtils: |
|
|
|
|
|
|
132
|
|
|
@classmethod |
|
133
|
|
|
def hits_to_df(cls, hits: Sequence[AbstractHit]) -> HitFrame: |
|
|
|
|
|
|
134
|
|
|
data = [] |
|
135
|
|
|
for hit in hits: |
|
136
|
|
|
x = {f: getattr(hit, f) for f in hit.__class__.fields()} |
|
|
|
|
|
|
137
|
|
|
x["universal_id"] = hit.universal_id |
|
138
|
|
|
x["hit_class"] = hit.hit_class |
|
139
|
|
|
data.append(x) |
|
140
|
|
|
return HitFrame([pd.Series(x) for x in data]) |
|
141
|
|
|
|
|
142
|
|
|
@classmethod |
|
143
|
|
|
def df_to_hits(cls, self: HitFrame) -> Sequence[AbstractHit]: |
|
|
|
|
|
|
144
|
|
|
hits = [] |
|
145
|
|
|
for row in self.iterrows(): |
|
146
|
|
|
clazz = ReflectionUtils.injection(row.hit_class, AbstractHit) |
|
147
|
|
|
# ignore extra columns |
|
148
|
|
|
# if cols are missing, let it fail on clazz.__init__ |
|
149
|
|
|
data = {f: getattr(row, f) for f in self.columns if f in row.__dict__} |
|
150
|
|
|
# noinspection PyArgumentList |
|
151
|
|
|
hit = clazz(**data) |
|
152
|
|
|
hits.append(hit) |
|
153
|
|
|
return hits |
|
154
|
|
|
|
|
155
|
|
|
|
|
156
|
|
|
__all__ = ["AbstractHit", "HitFrame", "Pair", "Triple", "HIT_FIELD_TYPE", "HitUtils"] |
|
157
|
|
|
|
dmyersturnbull /
mandos
Loading history...