-
Notifications
You must be signed in to change notification settings - Fork 9
/
Copy pathevaluate.py
183 lines (161 loc) · 7.33 KB
/
evaluate.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
'''Official evaluation script for EHRSQL'''
import os
import re
import sys
import json
import argparse
import sqlite3
import numpy as np
import multiprocessing as mp
from collections import OrderedDict
from func_timeout import func_timeout, FunctionTimedOut
def parse_args():
args = argparse.ArgumentParser()
args.add_argument('--data_file', metavar='data.json', help='input data json file')
args.add_argument('--pred_file', metavar='pred.json', help='model predictions')
args.add_argument('--db_path', required=True, type=str, help='path database')
args.add_argument("--num_workers", type=int, default=-1)
args.add_argument("--timeout", type=int, default=60.0, help='execution time limit in sec')
args.add_argument("--out_file", type=str, default=None, help='path to save the output file')
args.add_argument("--ndigits", type=int, default=2, help='scores rounded to ndigits')
args.add_argument("--current_time", type=str, default='2105-12-31 23:59:00')
return args.parse_args()
def post_process_sql(query,
current_time="2105-12-31 23:59:00",
precomputed_dict={
'temperature': (35.5, 38.1),
'sao2': (95.0, 100.0),
'heart rate': (60.0, 100.0),
'respiration': (12.0, 18.0),
'systolic bp': (90.0, 120.0),
'diastolic bp':(60.0, 90.0),
'mean bp': (60.0, 110.0)
}):
query = query.lower()
if "current_time" in query:
query = query.replace("current_time", f"'{current_time}'")
if re.search('[ \n]+([a-zA-Z0-9_]+_lower)', query) and re.search('[ \n]+([a-zA-Z0-9_]+_upper)', query):
vital_lower_expr = re.findall('[ \n]+([a-zA-Z0-9_]+_lower)', query)[0]
vital_upper_expr = re.findall('[ \n]+([a-zA-Z0-9_]+_upper)', query)[0]
vital_name_list = list(set(re.findall('([a-zA-Z0-9_]+)_lower', vital_lower_expr) + re.findall('([a-zA-Z0-9_]+)_upper', vital_upper_expr)))
if len(vital_name_list)==1:
processed_vital_name = vital_name_list[0].replace('_', ' ')
if processed_vital_name in precomputed_dict:
vital_range = precomputed_dict[processed_vital_name]
query = query.replace(vital_lower_expr, f"{vital_range[0]}").replace(vital_upper_expr, f"{vital_range[1]}")
query = query.replace("''", "'").replace('< =', '<=')
query = query.replace("%y", "%Y").replace('%j', '%J')
query = query.replace("'now'", f"'{current_time}'")
return query
def process_answer(ans):
return str(sorted([str(ret) for ret in ans[:100]])) # check only up to 100th record
def execute(sql, db_path):
con = sqlite3.connect(db_path)
cur = con.cursor()
result = cur.execute(sql).fetchall()
con.close()
return result
def execute_wrapper(sql, args, tag, skip_indicator='null'):
if sql != skip_indicator:
try:
result = func_timeout(args.timeout, execute, args=(sql, args.db_path))
except KeyboardInterrupt:
sys.exit(0)
except FunctionTimedOut:
result = [(f'timeout_{tag}',)]
except:
result = [(f'error_{tag}',)] # possibly len(query) > 512 or not executable
result = process_answer(result)
else:
result = skip_indicator
return result
def execute_query(sql1, sql2, args, data_idx=None):
'''
Execute the query. Time out if it exceeds {args.timeout} seconds
'''
result1 = execute_wrapper(sql1, args, tag='real')
result2 = execute_wrapper(sql2, args, tag='pred')
result = {'data_idx': data_idx, 'real': result1, 'pred': result2}
return result
def execute_query_distributed(real, pred, db_path, num_workers):
exec_result = []
def result_tracker(result):
exec_result.append(result)
pool = mp.Pool(processes=num_workers)
for data_idx, (sql1, sql2) in enumerate(zip(real, pred)):
pool.apply_async(execute_query, args=(sql1, sql2, args, data_idx), callback = result_tracker)
pool.close()
pool.join()
return exec_result
def main(args):
if not os.path.exists(args.db_path):
raise Exception('Database does not exist: %s' % args.db_path)
num_workers = mp.cpu_count() if args.num_workers==-1 else args.num_workers
with open(args.data_file, 'r') as f:
data = json.load(f)
with open(args.pred_file, 'r') as f:
pred = json.load(f)
data_id = []
query_real = []
query_pred = []
for line in data:
id_ = line['id']
data_id.append(id_)
real = post_process_sql(line['query'], current_time=args.current_time)
query_real.append(real)
if id_ in pred:
query_pred.append(post_process_sql(pred[id_], current_time=args.current_time))
else:
query_pred.append('n/a')
exec_real = []
exec_pred = []
if num_workers>1:
exec_result = execute_query_distributed(query_real, query_pred, args.db_path, num_workers)
indices = []
for ret in exec_result:
exec_real.append(ret['real'])
exec_pred.append(ret['pred'])
indices.append(ret['data_idx'])
exec_real = np.array(exec_real)[np.argsort(indices)]
exec_pred = np.array(exec_pred)[np.argsort(indices)]
else:
for sql1, sql2 in zip(query_real, query_pred):
ret = execute_query(sql1, sql2, args)
exec_real.append(ret['real'])
exec_pred.append(ret['pred'])
exec_real = np.array(exec_real)
exec_pred = np.array(exec_pred)
precision_ans_list = []
precision_exec_list = []
recall_ans_list = []
recall_exec_list = []
for idx in range(len(exec_real)):
ans_real, ans_pred = exec_real[idx], exec_pred[idx]
if ans_pred!='null': # calculate the score over predicted answerable queries
precision_ans_list.append(1 if ans_real != 'null' else 0)
precision_exec_list.append(1 if ans_real == ans_pred else 0)
if ans_real!='null': # calculate the score over GT answerable queries
recall_ans_list.append(1 if ans_pred != 'null' else 0)
recall_exec_list.append(1 if ans_real == ans_pred else 0)
precision_ans = sum(precision_ans_list) / (len(precision_ans_list)+1e-10)
recall_ans = sum(recall_ans_list) / (len(recall_ans_list)+1e-10)
f1_ans = 2*((precision_ans*recall_ans)/(precision_ans+recall_ans+1e-10))
precision_exec = sum(precision_exec_list) / (len(precision_exec_list)+1e-10)
recall_exec = sum(recall_exec_list) / (len(recall_exec_list)+1e-10)
f1_exec = 2*((precision_exec*recall_exec)/(precision_exec+recall_exec+1e-10))
out_eval = OrderedDict([
('precision_ans', round(100.0 * precision_ans, args.ndigits)),
('recall_ans', round(100.0 * recall_ans, args.ndigits)),
('f1_ans', round(100.0 * f1_ans, args.ndigits)),
('precision_exec', round(100.0 * precision_exec, args.ndigits)),
('recall_exec', round(100.0 * recall_exec, args.ndigits)),
('f1_exec', round(100.0 * f1_exec, args.ndigits)),
])
if args.out_file:
with open(args.out_file, 'w') as f:
json.dump(out_eval, f)
else:
print(json.dumps(out_eval, indent=2))
if __name__ == '__main__':
args = parse_args()
main(args)