forked from kakao-arena/brunch-article-recommendation
-
Notifications
You must be signed in to change notification settings - Fork 0
/
database.py
48 lines (39 loc) · 1.55 KB
/
database.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
# -*- coding: utf-8 -*-
import os
import random
import six
import fire
import mmh3
import tqdm
from util import iterate_data_files
def groupby(from_dtm, to_dtm, tmp_dir, out_path, num_chunks=10):
from_dtm, to_dtm = map(str, [from_dtm, to_dtm])
fouts = {idx: open(os.path.join(tmp_dir, str(idx)), 'w')
for idx in range(num_chunks)}
files = sorted([path for path, _ in iterate_data_files(from_dtm, to_dtm)])
for path in tqdm.tqdm(files, mininterval=1):
for line in open(path):
user = line.strip().split()[0]
chunk_index = mmh3.hash(user, 17) % num_chunks
fouts[chunk_index].write(line)
map(lambda x: x.close(), fouts.values())
with open(out_path, 'w') as fout:
for chunk_idx in fouts.keys():
_groupby = {}
chunk_path = os.path.join(tmp_dir, str(chunk_idx))
for line in open(chunk_path):
tkns = line.strip().split()
userid, seen = tkns[0], tkns[1:]
_groupby.setdefault(userid, []).extend(seen)
os.remove(chunk_path)
for userid, seen in six.iteritems(_groupby):
fout.write('%s %s\n' % (userid, ' '.join(seen)))
def sample_users(data_path, out_path, num_users):
users = [data.strip().split()[0] for data in open(data_path)]
random.shuffle(users)
users = users[:num_users]
with open(out_path, 'w') as fout:
fout.write('\n'.join(users))
if __name__ == '__main__':
fire.Fire({'groupby': groupby,
'sample_users': sample_users})