mirror of
https://github.com/systemd/systemd.git
synced 2024-11-01 09:21:26 +03:00
11a1589223
Files which are installed as-is (any .service and other unit files, .conf files, .policy files, etc), are left as is. My assumption is that SPDX identifiers are not yet that well known, so it's better to retain the extended header to avoid any doubt. I also kept any copyright lines. We can probably remove them, but it'd nice to obtain explicit acks from all involved authors before doing that.
126 lines
3.4 KiB
Python
Executable File
126 lines
3.4 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
# -*- Mode: python; coding: utf-8; indent-tabs-mode: nil -*- */
|
|
# SPDX-License-Identifier: LGPL-2.1+
|
|
#
|
|
# This file is part of systemd.
|
|
#
|
|
# Copyright 2012 Lennart Poettering
|
|
# Copyright 2013 Zbigniew Jędrzejewski-Szmek
|
|
|
|
import collections
|
|
import sys
|
|
import re
|
|
from xml_helper import xml_parse, xml_print, tree
|
|
|
|
MDASH = ' — ' if sys.version_info.major >= 3 else ' -- '
|
|
|
|
TEMPLATE = '''\
|
|
<refentry id="systemd.index" conditional="HAVE_PYTHON">
|
|
|
|
<refentryinfo>
|
|
<title>systemd.index</title>
|
|
<productname>systemd</productname>
|
|
|
|
<authorgroup>
|
|
<author>
|
|
<contrib>Developer</contrib>
|
|
<firstname>Lennart</firstname>
|
|
<surname>Poettering</surname>
|
|
<email>lennart@poettering.net</email>
|
|
</author>
|
|
</authorgroup>
|
|
</refentryinfo>
|
|
|
|
<refmeta>
|
|
<refentrytitle>systemd.index</refentrytitle>
|
|
<manvolnum>7</manvolnum>
|
|
</refmeta>
|
|
|
|
<refnamediv>
|
|
<refname>systemd.index</refname>
|
|
<refpurpose>List all manpages from the systemd project</refpurpose>
|
|
</refnamediv>
|
|
</refentry>
|
|
'''
|
|
|
|
SUMMARY = '''\
|
|
<refsect1>
|
|
<title>See Also</title>
|
|
<para>
|
|
<citerefentry><refentrytitle>systemd.directives</refentrytitle><manvolnum>7</manvolnum></citerefentry>
|
|
</para>
|
|
|
|
<para id='counts' />
|
|
</refsect1>
|
|
'''
|
|
|
|
COUNTS = '\
|
|
This index contains {count} entries, referring to {pages} individual manual pages.'
|
|
|
|
|
|
def check_id(page, t):
|
|
id = t.getroot().get('id')
|
|
if not re.search('/' + id + '[.]', page):
|
|
raise ValueError("id='{}' is not the same as page name '{}'".format(id, page))
|
|
|
|
def make_index(pages):
|
|
index = collections.defaultdict(list)
|
|
for p in pages:
|
|
t = xml_parse(p)
|
|
check_id(p, t)
|
|
section = t.find('./refmeta/manvolnum').text
|
|
refname = t.find('./refnamediv/refname').text
|
|
purpose = ' '.join(t.find('./refnamediv/refpurpose').text.split())
|
|
for f in t.findall('./refnamediv/refname'):
|
|
infos = (f.text, section, purpose, refname)
|
|
index[f.text[0].upper()].append(infos)
|
|
return index
|
|
|
|
def add_letter(template, letter, pages):
|
|
refsect1 = tree.SubElement(template, 'refsect1')
|
|
title = tree.SubElement(refsect1, 'title')
|
|
title.text = letter
|
|
para = tree.SubElement(refsect1, 'para')
|
|
for info in sorted(pages, key=lambda info: str.lower(info[0])):
|
|
refname, section, purpose, realname = info
|
|
|
|
b = tree.SubElement(para, 'citerefentry')
|
|
c = tree.SubElement(b, 'refentrytitle')
|
|
c.text = refname
|
|
d = tree.SubElement(b, 'manvolnum')
|
|
d.text = section
|
|
|
|
b.tail = MDASH + purpose # + ' (' + p + ')'
|
|
|
|
tree.SubElement(para, 'sbr')
|
|
|
|
def add_summary(template, indexpages):
|
|
count = 0
|
|
pages = set()
|
|
for group in indexpages:
|
|
count += len(group)
|
|
for info in group:
|
|
refname, section, purpose, realname = info
|
|
pages.add((realname, section))
|
|
|
|
refsect1 = tree.fromstring(SUMMARY)
|
|
template.append(refsect1)
|
|
|
|
para = template.find(".//para[@id='counts']")
|
|
para.text = COUNTS.format(count=count, pages=len(pages))
|
|
|
|
def make_page(*xml_files):
|
|
template = tree.fromstring(TEMPLATE)
|
|
index = make_index(xml_files)
|
|
|
|
for letter in sorted(index):
|
|
add_letter(template, letter, index[letter])
|
|
|
|
add_summary(template, index.values())
|
|
|
|
return template
|
|
|
|
if __name__ == '__main__':
|
|
with open(sys.argv[1], 'wb') as f:
|
|
f.write(xml_print(make_page(*sys.argv[2:])))
|