1
0
mirror of https://github.com/arsenetar/dupeguru.git synced 2024-11-04 23:39:02 +00:00
dupeguru/core/util.py
Andrew Senetar 7ba8aa3514
Format files with black
- Format all files with black
- Update tox.ini flake8 arguments to be compatible
- Add black to requirements-extra.txt
- Reduce ignored flake8 rules and fix a few violations
2019-12-31 20:16:27 -06:00

61 lines
2.1 KiB
Python

# Copyright 2016 Hardcoded Software (http://www.hardcoded.net)
#
# This software is licensed under the "GPLv3" License as described in the "LICENSE" file,
# which should be included with this package. The terms are also available at
# http://www.gnu.org/licenses/gpl-3.0.html
import time
from hscommon.util import format_time_decimal
def format_timestamp(t, delta):
if delta:
return format_time_decimal(t)
else:
if t > 0:
return time.strftime("%Y/%m/%d %H:%M:%S", time.localtime(t))
else:
return "---"
def format_words(w):
def do_format(w):
if isinstance(w, list):
return "(%s)" % ", ".join(do_format(item) for item in w)
else:
return w.replace("\n", " ")
return ", ".join(do_format(item) for item in w)
def format_perc(p):
return "%0.0f" % p
def format_dupe_count(c):
return str(c) if c else "---"
def cmp_value(dupe, attrname):
value = getattr(dupe, attrname, "")
return value.lower() if isinstance(value, str) else value
def fix_surrogate_encoding(s, encoding="utf-8"):
# ref #210. It's possible to end up with file paths that, while correct unicode strings, are
# decoded with the 'surrogateescape' option, which make the string unencodable to utf-8. We fix
# these strings here by trying to encode them and, if it fails, we do an encode/decode dance
# to remove the problematic characters. This dance is *lossy* but there's not much we can do
# because if we end up with this type of string, it means that we don't know the encoding of the
# underlying filesystem that brought them. Don't use this for strings you're going to re-use in
# fs-related functions because you're going to lose your path (it's going to change). Use this
# if you need to export the path somewhere else, outside of the unicode realm.
# See http://lucumr.pocoo.org/2013/7/2/the-updated-guide-to-unicode/
try:
s.encode(encoding)
except UnicodeEncodeError:
return s.encode(encoding, "replace").decode(encoding)
else:
return s