Files @ 51b70569c330
Branch filter:

Location: kallithea/rhodecode/controllers/summary.py

Marcin Kuzminski
extended admin rescan to show what repositories was added and what removed
fixed found bugs, added expunge to cached getter since it introduce some more problems.
Added some relationship to properly clean related objects from database.
# -*- coding: utf-8 -*-
"""
    rhodecode.controllers.summary
    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    Summary controller for Rhodecode
    
    :created_on: Apr 18, 2010
    :author: marcink
    :copyright: (C) 2009-2011 Marcin Kuzminski <marcin@python-works.com>    
    :license: GPLv3, see COPYING for more details.
"""
# 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; version 2
# of the License or (at your opinion) any later version of the license.
# 
# 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.

import calendar
import logging
from time import mktime
from datetime import datetime, timedelta, date

from vcs.exceptions import ChangesetError

from pylons import tmpl_context as c, request, url
from pylons.i18n.translation import _

from rhodecode.model.scm import ScmModel
from rhodecode.model.db import Statistics

from rhodecode.lib.auth import LoginRequired, HasRepoPermissionAnyDecorator
from rhodecode.lib.base import BaseController, render
from rhodecode.lib.utils import OrderedDict, EmptyChangeset

from rhodecode.lib.celerylib import run_task
from rhodecode.lib.celerylib.tasks import get_commits_stats

from webhelpers.paginate import Page

try:
    import json
except ImportError:
    #python 2.5 compatibility
    import simplejson as json
log = logging.getLogger(__name__)

class SummaryController(BaseController):

    @LoginRequired()
    @HasRepoPermissionAnyDecorator('repository.read', 'repository.write',
                                   'repository.admin')
    def __before__(self):
        super(SummaryController, self).__before__()

    def index(self):
        scm_model = ScmModel()
        c.repo, dbrepo = scm_model.get(c.repo_name)
        c.dbrepo = dbrepo
        c.following = scm_model.is_following_repo(c.repo_name,
                                             c.rhodecode_user.user_id)
        def url_generator(**kw):
            return url('shortlog_home', repo_name=c.repo_name, **kw)

        c.repo_changesets = Page(c.repo, page=1, items_per_page=10,
                                 url=url_generator)

        e = request.environ

        if self.rhodecode_user.username == 'default':
            #for default(anonymous) user we don't need to pass credentials
            username = ''
            password = ''
        else:
            username = str(c.rhodecode_user.username)
            password = '@'

        uri = u'%(protocol)s://%(user)s%(password)s%(host)s%(prefix)s/%(repo_name)s' % {
                                        'protocol': e.get('wsgi.url_scheme'),
                                        'user':username,
                                        'password':password,
                                        'host':e.get('HTTP_HOST'),
                                        'prefix':e.get('SCRIPT_NAME'),
                                        'repo_name':c.repo_name, }
        c.clone_repo_url = uri
        c.repo_tags = OrderedDict()
        for name, hash in c.repo.tags.items()[:10]:
            try:
                c.repo_tags[name] = c.repo.get_changeset(hash)
            except ChangesetError:
                c.repo_tags[name] = EmptyChangeset(hash)

        c.repo_branches = OrderedDict()
        for name, hash in c.repo.branches.items()[:10]:
            try:
                c.repo_branches[name] = c.repo.get_changeset(hash)
            except ChangesetError:
                c.repo_branches[name] = EmptyChangeset(hash)

        td = date.today() + timedelta(days=1)
        td_1m = td - timedelta(days=calendar.mdays[td.month])
        td_1y = td - timedelta(days=365)

        ts_min_m = mktime(td_1m.timetuple())
        ts_min_y = mktime(td_1y.timetuple())
        ts_max_y = mktime(td.timetuple())

        if dbrepo.enable_statistics:
            c.no_data_msg = _('No data loaded yet')
            run_task(get_commits_stats, c.repo.name, ts_min_y, ts_max_y)
        else:
            c.no_data_msg = _('Statistics update are disabled for this repository')
        c.ts_min = ts_min_m
        c.ts_max = ts_max_y

        stats = self.sa.query(Statistics)\
            .filter(Statistics.repository == dbrepo)\
            .scalar()


        if stats and stats.languages:
            c.no_data = False is dbrepo.enable_statistics
            lang_stats = json.loads(stats.languages)
            c.commit_data = stats.commit_activity
            c.overview_data = stats.commit_activity_combined
            c.trending_languages = json.dumps(OrderedDict(
                                       sorted(lang_stats.items(), reverse=True,
                                            key=lambda k: k[1])[:10]
                                        )
                                    )
        else:
            c.commit_data = json.dumps({})
            c.overview_data = json.dumps([[ts_min_y, 0], [ts_max_y, 10] ])
            c.trending_languages = json.dumps({})
            c.no_data = True

        c.enable_downloads = dbrepo.enable_downloads
        if c.enable_downloads:
            c.download_options = self._get_download_links(c.repo)

        return render('summary/summary.html')



    def _get_download_links(self, repo):

        download_l = []

        branches_group = ([], _("Branches"))
        tags_group = ([], _("Tags"))

        for name, chs in c.repository_branches.items():
            #chs = chs.split(':')[-1]
            branches_group[0].append((chs, name),)
        download_l.append(branches_group)

        for name, chs in c.repository_tags.items():
            #chs = chs.split(':')[-1]
            tags_group[0].append((chs, name),)
        download_l.append(tags_group)

        return download_l