1
0
mirror of https://github.com/wting/autojump synced 2024-10-27 20:34:07 +00:00
wting_autojump/bin/autojump
2016-10-01 10:10:48 -07:00

321 lines
10 KiB
Python
Executable File

#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
Copyright © 2008-2012 Joel Schaerer
Copyright © 2012-2016 William Ting
* This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3, or (at your option)
any later version.
* This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
* You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
"""
from __future__ import print_function
from itertools import chain
from math import sqrt
from operator import attrgetter
from operator import itemgetter
import os
import sys
if sys.version_info[0] == 3:
ifilter = filter
imap = map
os.getcwdu = os.getcwd
else:
from itertools import ifilter
from itertools import imap
# Vendorized argparse for Python 2.6 support
from autojump_argparse import ArgumentParser
# autojump is not a standard python package but rather installed as a mixed
# shell + Python app with no outside dependencies (except Python). As a
# consequence we use relative imports and depend on file prefixes to prevent
# module conflicts.
from autojump_data import dictify
from autojump_data import entriefy
from autojump_data import Entry
from autojump_data import load
from autojump_data import save
from autojump_match import match_anywhere
from autojump_match import match_consecutive
from autojump_match import match_fuzzy
from autojump_utils import first
from autojump_utils import get_pwd
from autojump_utils import get_tab_entry_info
from autojump_utils import has_uppercase
from autojump_utils import is_autojump_sourced
from autojump_utils import is_osx
from autojump_utils import is_windows
from autojump_utils import last
from autojump_utils import print_entry
from autojump_utils import print_local
from autojump_utils import print_tab_menu
from autojump_utils import sanitize
from autojump_utils import take
from autojump_utils import unico
VERSION = '22.5.0'
FUZZY_MATCH_THRESHOLD = 0.6
TAB_ENTRIES_COUNT = 9
TAB_SEPARATOR = '__'
def set_defaults():
config = {}
if is_osx():
data_home = os.path.join(
os.path.expanduser('~'),
'Library',
'autojump')
elif is_windows():
data_home = os.path.join(
os.getenv('APPDATA'),
'autojump')
else:
data_home = os.getenv(
'XDG_DATA_HOME',
os.path.join(
os.path.expanduser('~'),
'.local',
'share',
'autojump'))
config['data_path'] = os.path.join(data_home, 'autojump.txt')
config['backup_path'] = os.path.join(data_home, 'autojump.txt.bak')
return config
def parse_arguments():
parser = ArgumentParser(
description='Automatically jump to directory passed as an argument.',
epilog="Please see autojump(1) man pages for full documentation.")
parser.add_argument(
'directory', metavar='DIRECTORY', nargs='*', default='',
help='directory to jump to')
parser.add_argument(
'-a', '--add', metavar='DIRECTORY',
help='add path')
parser.add_argument(
'-i', '--increase', metavar='WEIGHT', nargs='?', type=int,
const=10, default=False,
help='increase current directory weight')
parser.add_argument(
'-d', '--decrease', metavar='WEIGHT', nargs='?', type=int,
const=15, default=False,
help='decrease current directory weight')
parser.add_argument(
'--complete', action="store_true", default=False,
help='used for tab completion')
parser.add_argument(
'--purge', action="store_true", default=False,
help='remove non-existent paths from database')
parser.add_argument(
'-s', '--stat', action="store_true", default=False,
help='show database entries and their key weights')
parser.add_argument(
'-v', '--version', action="version", version="%(prog)s v" +
VERSION, help='show version information')
return parser.parse_args()
def add_path(data, path, weight=10):
"""
Add a new path or increment an existing one.
os.path.realpath() is not used because it's preferable to use symlinks
with resulting duplicate entries in the database than a single canonical
path.
"""
path = unico(path).rstrip(os.sep)
if path == os.path.expanduser('~'):
return data, Entry(path, 0)
data[path] = sqrt((data.get(path, 0) ** 2) + (weight ** 2))
return data, Entry(path, data[path])
def decrease_path(data, path, weight=15):
"""Decrease or zero out a path."""
path = unico(path).rstrip(os.sep)
data[path] = max(0, data.get(path, 0) - weight)
return data, Entry(path, data[path])
def detect_smartcase(needles):
"""
If any needles contain an uppercase letter then use case sensitive
searching. Otherwise use case insensitive searching.
"""
return not any(imap(has_uppercase, needles))
def find_matches(entries, needles, check_entries=True):
"""Return an iterator to matching entries."""
# TODO(wting|2014-02-24): replace assertion with unit test
assert isinstance(needles, list), "Needles must be a list."
ignore_case = detect_smartcase(needles)
try:
pwd = os.getcwdu()
except OSError:
pwd = None
# using closure to prevent constantly hitting hdd
def is_cwd(entry):
return os.path.realpath(entry.path) == pwd
if check_entries:
path_exists = lambda entry: os.path.exists(entry.path)
else:
path_exists = lambda _: True
data = sorted(
entries,
key=attrgetter('weight', 'path'),
reverse=True)
return ifilter(
lambda entry: not is_cwd(entry) and path_exists(entry),
chain(
match_consecutive(needles, data, ignore_case),
match_fuzzy(needles, data, ignore_case),
match_anywhere(needles, data, ignore_case)))
def handle_tab_completion(needle, entries):
tab_needle, tab_index, tab_path = get_tab_entry_info(needle, TAB_SEPARATOR)
if tab_path:
print_local(tab_path)
elif tab_index:
get_ith_path = lambda i, iterable: last(take(i, iterable)).path
print_local(get_ith_path(
tab_index,
find_matches(entries, [tab_needle], check_entries=False)))
elif tab_needle:
# found partial tab completion entry
print_tab_menu(
tab_needle,
take(TAB_ENTRIES_COUNT, find_matches(
entries,
[tab_needle],
check_entries=False)),
TAB_SEPARATOR)
else:
print_tab_menu(
needle,
take(TAB_ENTRIES_COUNT, find_matches(
entries,
[needle],
check_entries=False)),
TAB_SEPARATOR)
def purge_missing_paths(entries):
"""Remove non-existent paths from a list of entries."""
exists = lambda entry: os.path.exists(entry.path)
return ifilter(exists, entries)
def print_stats(data, data_path):
for path, weight in sorted(data.items(), key=itemgetter(1)):
print_entry(Entry(path, weight))
print("________________________________________\n")
print("%d:\t total weight" % sum(data.values()))
print("%d:\t number of entries" % len(data))
try:
print_local(
"%.2f:\t current directory weight" % data.get(os.getcwdu(), 0))
except OSError:
# current directory no longer exists
pass
print("\ndata:\t %s" % data_path)
def main(args): # noqa
if not is_autojump_sourced() and not is_windows():
print("Please source the correct autojump file in your shell's")
print("startup file. For more information, please reinstall autojump")
print("and read the post installation instructions.")
return 1
config = set_defaults()
# all arguments are mutually exclusive
if args.add:
save(config, first(add_path(load(config), args.add)))
elif args.complete:
handle_tab_completion(
needle=first(chain(sanitize(args.directory), [''])),
entries=entriefy(load(config)))
elif args.decrease:
data, entry = decrease_path(load(config), get_pwd(), args.decrease)
save(config, data)
print_entry(entry)
elif args.increase:
data, entry = add_path(load(config), get_pwd(), args.increase)
save(config, data)
print_entry(entry)
elif args.purge:
old_data = load(config)
new_data = dictify(purge_missing_paths(entriefy(old_data)))
save(config, new_data)
print("Purged %d entries." % (len(old_data) - len(new_data)))
elif args.stat:
print_stats(load(config), config['data_path'])
elif not args.directory:
# Return best match.
entries = entriefy(load(config))
print_local(first(chain(
imap(attrgetter('path'), find_matches(entries, [''])),
# always return a path to calling shell functions
['.'])))
else:
entries = entriefy(load(config))
needles = sanitize(args.directory)
tab_needle, tab_index, tab_path = \
get_tab_entry_info(first(needles), TAB_SEPARATOR)
# Handle `j foo__`, assuming first index.
if not tab_path and not tab_index \
and tab_needle and needles[0] == tab_needle + TAB_SEPARATOR:
tab_index = 1
if tab_path:
print_local(tab_path)
elif tab_index:
get_ith_path = lambda i, iterable: last(take(i, iterable)).path
print_local(
get_ith_path(
tab_index,
find_matches(entries, [tab_needle])))
else:
print_local(first(chain(
imap(attrgetter('path'), find_matches(entries, needles)),
# always return a path to calling shell functions
['.'])))
return 0
if __name__ == "__main__":
sys.exit(main(parse_arguments()))