|
|
# template-filters.py - common template expansion filters
|
|
|
#
|
|
|
# Copyright 2005-2008 Matt Mackall <mpm@selenic.com>
|
|
|
#
|
|
|
# This software may be used and distributed according to the terms of the
|
|
|
# GNU General Public License version 2, incorporated herein by reference.
|
|
|
|
|
|
import cgi, re, os, time, urllib, textwrap
|
|
|
import util, encoding
|
|
|
|
|
|
def stringify(thing):
|
|
|
'''turn nested template iterator into string.'''
|
|
|
if hasattr(thing, '__iter__') and not isinstance(thing, str):
|
|
|
return "".join([stringify(t) for t in thing if t is not None])
|
|
|
return str(thing)
|
|
|
|
|
|
agescales = [("second", 1),
|
|
|
("minute", 60),
|
|
|
("hour", 3600),
|
|
|
("day", 3600 * 24),
|
|
|
("week", 3600 * 24 * 7),
|
|
|
("month", 3600 * 24 * 30),
|
|
|
("year", 3600 * 24 * 365)]
|
|
|
|
|
|
agescales.reverse()
|
|
|
|
|
|
def age(date):
|
|
|
'''turn a (timestamp, tzoff) tuple into an age string.'''
|
|
|
|
|
|
def plural(t, c):
|
|
|
if c == 1:
|
|
|
return t
|
|
|
return t + "s"
|
|
|
def fmt(t, c):
|
|
|
return "%d %s" % (c, plural(t, c))
|
|
|
|
|
|
now = time.time()
|
|
|
then = date[0]
|
|
|
if then > now:
|
|
|
return 'in the future'
|
|
|
|
|
|
delta = max(1, int(now - then))
|
|
|
for t, s in agescales:
|
|
|
n = delta // s
|
|
|
if n >= 2 or s == 1:
|
|
|
return fmt(t, n)
|
|
|
|
|
|
para_re = None
|
|
|
space_re = None
|
|
|
|
|
|
def fill(text, width):
|
|
|
'''fill many paragraphs.'''
|
|
|
global para_re, space_re
|
|
|
if para_re is None:
|
|
|
para_re = re.compile('(\n\n|\n\\s*[-*]\\s*)', re.M)
|
|
|
space_re = re.compile(r' +')
|
|
|
|
|
|
def findparas():
|
|
|
start = 0
|
|
|
while True:
|
|
|
m = para_re.search(text, start)
|
|
|
if not m:
|
|
|
w = len(text)
|
|
|
while w > start and text[w-1].isspace(): w -= 1
|
|
|
yield text[start:w], text[w:]
|
|
|
break
|
|
|
yield text[start:m.start(0)], m.group(1)
|
|
|
start = m.end(1)
|
|
|
|
|
|
return "".join([space_re.sub(' ', textwrap.fill(para, width)) + rest
|
|
|
for para, rest in findparas()])
|
|
|
|
|
|
def firstline(text):
|
|
|
'''return the first line of text'''
|
|
|
try:
|
|
|
return text.splitlines(True)[0].rstrip('\r\n')
|
|
|
except IndexError:
|
|
|
return ''
|
|
|
|
|
|
def nl2br(text):
|
|
|
'''replace raw newlines with xhtml line breaks.'''
|
|
|
return text.replace('\n', '<br/>\n')
|
|
|
|
|
|
def obfuscate(text):
|
|
|
text = unicode(text, encoding.encoding, 'replace')
|
|
|
return ''.join(['&#%d;' % ord(c) for c in text])
|
|
|
|
|
|
def domain(author):
|
|
|
'''get domain of author, or empty string if none.'''
|
|
|
f = author.find('@')
|
|
|
if f == -1: return ''
|
|
|
author = author[f+1:]
|
|
|
f = author.find('>')
|
|
|
if f >= 0: author = author[:f]
|
|
|
return author
|
|
|
|
|
|
def person(author):
|
|
|
'''get name of author, or else username.'''
|
|
|
if not '@' in author: return author
|
|
|
f = author.find('<')
|
|
|
if f == -1: return util.shortuser(author)
|
|
|
return author[:f].rstrip()
|
|
|
|
|
|
def indent(text, prefix):
|
|
|
'''indent each non-empty line of text after first with prefix.'''
|
|
|
lines = text.splitlines()
|
|
|
num_lines = len(lines)
|
|
|
def indenter():
|
|
|
for i in xrange(num_lines):
|
|
|
l = lines[i]
|
|
|
if i and l.strip():
|
|
|
yield prefix
|
|
|
yield l
|
|
|
if i < num_lines - 1 or text.endswith('\n'):
|
|
|
yield '\n'
|
|
|
return "".join(indenter())
|
|
|
|
|
|
def permissions(flags):
|
|
|
if "l" in flags:
|
|
|
return "lrwxrwxrwx"
|
|
|
if "x" in flags:
|
|
|
return "-rwxr-xr-x"
|
|
|
return "-rw-r--r--"
|
|
|
|
|
|
def xmlescape(text):
|
|
|
text = (text
|
|
|
.replace('&', '&')
|
|
|
.replace('<', '<')
|
|
|
.replace('>', '>')
|
|
|
.replace('"', '"')
|
|
|
.replace("'", ''')) # ' invalid in HTML
|
|
|
return re.sub('[\x00-\x08\x0B\x0C\x0E-\x1F]', ' ', text)
|
|
|
|
|
|
_escapes = [
|
|
|
('\\', '\\\\'), ('"', '\\"'), ('\t', '\\t'), ('\n', '\\n'),
|
|
|
('\r', '\\r'), ('\f', '\\f'), ('\b', '\\b'),
|
|
|
]
|
|
|
|
|
|
def jsonescape(s):
|
|
|
for k, v in _escapes:
|
|
|
s = s.replace(k, v)
|
|
|
return s
|
|
|
|
|
|
def json(obj):
|
|
|
if obj is None or obj is False or obj is True:
|
|
|
return {None: 'null', False: 'false', True: 'true'}[obj]
|
|
|
elif isinstance(obj, int) or isinstance(obj, float):
|
|
|
return str(obj)
|
|
|
elif isinstance(obj, str):
|
|
|
return '"%s"' % jsonescape(obj)
|
|
|
elif isinstance(obj, unicode):
|
|
|
return json(obj.encode('utf-8'))
|
|
|
elif hasattr(obj, 'keys'):
|
|
|
out = []
|
|
|
for k, v in obj.iteritems():
|
|
|
s = '%s: %s' % (json(k), json(v))
|
|
|
out.append(s)
|
|
|
return '{' + ', '.join(out) + '}'
|
|
|
elif hasattr(obj, '__iter__'):
|
|
|
out = []
|
|
|
for i in obj:
|
|
|
out.append(json(i))
|
|
|
return '[' + ', '.join(out) + ']'
|
|
|
else:
|
|
|
raise TypeError('cannot encode type %s' % obj.__class__.__name__)
|
|
|
|
|
|
def stripdir(text):
|
|
|
'''Treat the text as path and strip a directory level, if possible.'''
|
|
|
dir = os.path.dirname(text)
|
|
|
if dir == "":
|
|
|
return os.path.basename(text)
|
|
|
else:
|
|
|
return dir
|
|
|
|
|
|
def nonempty(str):
|
|
|
return str or "(none)"
|
|
|
|
|
|
filters = {
|
|
|
"addbreaks": nl2br,
|
|
|
"basename": os.path.basename,
|
|
|
"stripdir": stripdir,
|
|
|
"age": age,
|
|
|
"date": lambda x: util.datestr(x),
|
|
|
"domain": domain,
|
|
|
"email": util.email,
|
|
|
"escape": lambda x: cgi.escape(x, True),
|
|
|
"fill68": lambda x: fill(x, width=68),
|
|
|
"fill76": lambda x: fill(x, width=76),
|
|
|
"firstline": firstline,
|
|
|
"tabindent": lambda x: indent(x, '\t'),
|
|
|
"hgdate": lambda x: "%d %d" % x,
|
|
|
"isodate": lambda x: util.datestr(x, '%Y-%m-%d %H:%M %1%2'),
|
|
|
"isodatesec": lambda x: util.datestr(x, '%Y-%m-%d %H:%M:%S %1%2'),
|
|
|
"json": json,
|
|
|
"jsonescape": jsonescape,
|
|
|
"localdate": lambda x: (x[0], util.makedate()[1]),
|
|
|
"nonempty": nonempty,
|
|
|
"obfuscate": obfuscate,
|
|
|
"permissions": permissions,
|
|
|
"person": person,
|
|
|
"rfc822date": lambda x: util.datestr(x, "%a, %d %b %Y %H:%M:%S %1%2"),
|
|
|
"rfc3339date": lambda x: util.datestr(x, "%Y-%m-%dT%H:%M:%S%1:%2"),
|
|
|
"short": lambda x: x[:12],
|
|
|
"shortdate": util.shortdate,
|
|
|
"stringify": stringify,
|
|
|
"strip": lambda x: x.strip(),
|
|
|
"urlescape": lambda x: urllib.quote(x),
|
|
|
"user": lambda x: util.shortuser(x),
|
|
|
"stringescape": lambda x: x.encode('string_escape'),
|
|
|
"xmlescape": xmlescape,
|
|
|
}
|
|
|
|