mirror of
https://github.com/saitohirga/WSJT-X.git
synced 2024-11-19 10:32:02 -05:00
72 lines
2.3 KiB
Python
72 lines
2.3 KiB
Python
|
#!/usr/bin/env python
|
||
|
from subprocess import check_output as run
|
||
|
from datetime import datetime
|
||
|
from itertools import groupby
|
||
|
from operator import itemgetter
|
||
|
import re
|
||
|
import magic
|
||
|
|
||
|
def authors(filename):
|
||
|
log = run(['git', 'log', '--follow',
|
||
|
'--date=short','--format=%aN%x09%ad', filename],
|
||
|
universal_newlines=True)
|
||
|
for line in log.splitlines():
|
||
|
author, date = line.split('\t')
|
||
|
if author != 'fix-copyright.py':
|
||
|
yield author, datetime.strptime(date, '%Y-%m-%d')
|
||
|
|
||
|
def new_copyright(filename, previous):
|
||
|
def f():
|
||
|
au = list(authors(filename))
|
||
|
alldates = map(itemgetter(1), au)
|
||
|
aup = sorted(au + map(lambda a: (a, None), previous), key=itemgetter(0))
|
||
|
for author, records in groupby(aup, itemgetter(0)):
|
||
|
dates = filter(None, map(itemgetter(1), records))
|
||
|
if not dates: dates = alldates
|
||
|
start = min(dates)
|
||
|
end = max(dates)
|
||
|
fmt = '{0}' if start.year == end.year else '{0}-{1}'
|
||
|
line = 'Copyright ' + fmt.format(start.year, end.year) + ' ' + author
|
||
|
key = (start, author)
|
||
|
yield key, line
|
||
|
return map(itemgetter(1), sorted(f()))
|
||
|
|
||
|
def fix_copyright(filename):
|
||
|
# Find copyright block in original file
|
||
|
prefix = set()
|
||
|
names = []
|
||
|
lines = []
|
||
|
with open(filename, 'r') as f:
|
||
|
content = list(f)
|
||
|
for i, line in enumerate(content[:15]):
|
||
|
m = re.match(r'^(?P<prefix>\W*)(\(c\))?\s*?copyright\s*(\(c\))?\s+\d{4}(\s*-\s*\d{4})?\s+(?P<name>.+?)\s*$', line, re.IGNORECASE)
|
||
|
if m:
|
||
|
d = m.groupdict()
|
||
|
prefix.add(d['prefix'])
|
||
|
lines.append(i)
|
||
|
names.append(d['name'].strip())
|
||
|
if len(prefix) != 1:
|
||
|
print 'Not found:', filename
|
||
|
return
|
||
|
prefix = list(prefix)[0]
|
||
|
|
||
|
print filename
|
||
|
new = iter(new_copyright(filename, names))
|
||
|
with open(filename, 'w') as f:
|
||
|
for i, line in enumerate(content):
|
||
|
if i in lines:
|
||
|
for repl in new:
|
||
|
print >>f, prefix + repl
|
||
|
else:
|
||
|
print >>f, line,
|
||
|
pass
|
||
|
|
||
|
def all_files():
|
||
|
ls = run(['git', 'ls-files'], universal_newlines=True)
|
||
|
for filename in ls.splitlines():
|
||
|
if magic.from_file(filename, mime=True).split('/')[0] == 'text':
|
||
|
yield filename
|
||
|
|
||
|
for f in all_files():
|
||
|
fix_copyright(f)
|