Remove the usertypes.enum from the source and use the standard enum module instead. Enum start number is available since python 3.5
170 lines
6.0 KiB
Python
170 lines
6.0 KiB
Python
# vim: ft=python fileencoding=utf-8 sts=4 sw=4 et:
|
|
|
|
# Copyright 2014-2017 Florian Bruhin (The Compiler) <mail@qutebrowser.org>
|
|
#
|
|
# This file is part of qutebrowser.
|
|
#
|
|
# qutebrowser 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 of the License, or
|
|
# (at your option) any later version.
|
|
#
|
|
# qutebrowser 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 qutebrowser. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
"""Utilities used for the documentation and built-in help."""
|
|
|
|
import re
|
|
import sys
|
|
import inspect
|
|
import os.path
|
|
import collections
|
|
import enum
|
|
|
|
import qutebrowser
|
|
from qutebrowser.utils import log, utils
|
|
|
|
|
|
def is_git_repo():
|
|
"""Check if we're running from a git repository."""
|
|
gitfolder = os.path.join(qutebrowser.basedir, os.path.pardir, '.git')
|
|
return os.path.isdir(gitfolder)
|
|
|
|
|
|
def docs_up_to_date(path):
|
|
"""Check if the generated html documentation is up to date.
|
|
|
|
Args:
|
|
path: The path of the document to check.
|
|
|
|
Return:
|
|
True if they are up to date or we couldn't check.
|
|
False if they are outdated.
|
|
"""
|
|
if hasattr(sys, 'frozen') or not is_git_repo():
|
|
return True
|
|
html_path = os.path.join(qutebrowser.basedir, 'html', 'doc', path)
|
|
filename = os.path.splitext(path)[0]
|
|
asciidoc_path = os.path.join(qutebrowser.basedir, os.path.pardir,
|
|
'doc', 'help', filename + '.asciidoc')
|
|
try:
|
|
html_time = os.path.getmtime(html_path)
|
|
asciidoc_time = os.path.getmtime(asciidoc_path)
|
|
except FileNotFoundError:
|
|
return True
|
|
return asciidoc_time <= html_time
|
|
|
|
|
|
class DocstringParser:
|
|
|
|
"""Generate documentation based on a docstring of a command handler.
|
|
|
|
The docstring needs to follow the format described in doc/contributing.
|
|
|
|
Attributes:
|
|
_state: The current state of the parser state machine.
|
|
_cur_arg_name: The name of the argument we're currently handling.
|
|
_short_desc_parts: The short description of the function as list.
|
|
_long_desc_parts: The long description of the function as list.
|
|
short_desc: The short description of the function.
|
|
long_desc: The long description of the function.
|
|
arg_descs: A dict of argument names to their descriptions
|
|
"""
|
|
|
|
State = enum.Enum('State', ['short', 'desc', 'desc_hidden',
|
|
'arg_start', 'arg_inside', 'misc'])
|
|
|
|
def __init__(self, func):
|
|
"""Constructor.
|
|
|
|
Args:
|
|
func: The function to parse the docstring for.
|
|
"""
|
|
self._state = self.State.short
|
|
self._cur_arg_name = None
|
|
self._short_desc_parts = []
|
|
self._long_desc_parts = []
|
|
self.arg_descs = collections.OrderedDict()
|
|
doc = inspect.getdoc(func)
|
|
handlers = {
|
|
self.State.short: self._parse_short,
|
|
self.State.desc: self._parse_desc,
|
|
self.State.desc_hidden: self._skip,
|
|
self.State.arg_start: self._parse_arg_start,
|
|
self.State.arg_inside: self._parse_arg_inside,
|
|
self.State.misc: self._skip,
|
|
}
|
|
if doc is None:
|
|
if sys.flags.optimize < 2:
|
|
log.commands.warning(
|
|
"Function {}() from {} has no docstring".format(
|
|
utils.qualname(func),
|
|
inspect.getsourcefile(func)))
|
|
self.long_desc = ""
|
|
self.short_desc = ""
|
|
return
|
|
for line in doc.splitlines():
|
|
handler = handlers[self._state]
|
|
stop = handler(line)
|
|
if stop:
|
|
break
|
|
for k, v in self.arg_descs.items():
|
|
desc = ' '.join(v)
|
|
desc = re.sub(r', or None($|\.)', r'\1', desc)
|
|
desc = re.sub(r', or None', r', or not given', desc)
|
|
self.arg_descs[k] = desc
|
|
self.long_desc = ' '.join(self._long_desc_parts)
|
|
self.short_desc = ' '.join(self._short_desc_parts)
|
|
|
|
def _process_arg(self, line):
|
|
"""Helper method to process a line like 'fooarg: Blah blub'."""
|
|
self._cur_arg_name, argdesc = line.split(':', maxsplit=1)
|
|
self._cur_arg_name = self._cur_arg_name.strip().lstrip('*')
|
|
self.arg_descs[self._cur_arg_name] = [argdesc.strip()]
|
|
|
|
def _skip(self, line):
|
|
"""Handler to ignore everything until we get 'Args:'."""
|
|
if line.startswith('Args:'):
|
|
self._state = self.State.arg_start
|
|
|
|
def _parse_short(self, line):
|
|
"""Parse the short description (first block) in the docstring."""
|
|
if not line:
|
|
self._state = self.State.desc
|
|
else:
|
|
self._short_desc_parts.append(line.strip())
|
|
|
|
def _parse_desc(self, line):
|
|
"""Parse the long description in the docstring."""
|
|
if line.startswith('Args:'):
|
|
self._state = self.State.arg_start
|
|
elif line.strip() == '//' or line.startswith('Attributes:'):
|
|
self._state = self.State.desc_hidden
|
|
elif line.strip():
|
|
self._long_desc_parts.append(line.strip())
|
|
|
|
def _parse_arg_start(self, line):
|
|
"""Parse first argument line."""
|
|
self._process_arg(line)
|
|
self._state = self.State.arg_inside
|
|
|
|
def _parse_arg_inside(self, line):
|
|
"""Parse subsequent argument lines."""
|
|
argname = self._cur_arg_name
|
|
if re.match(r'^[A-Z][a-z]+:$', line):
|
|
if not self.arg_descs[argname][-1].strip():
|
|
self.arg_descs[argname] = self.arg_descs[argname][:-1]
|
|
return True
|
|
elif not line.strip():
|
|
self.arg_descs[argname].append('\n\n')
|
|
elif line[4:].startswith(' '):
|
|
self.arg_descs[argname].append(line.strip() + '\n')
|
|
else:
|
|
self._process_arg(line)
|
|
return False
|