mirror/source.py

# Copyright 2009 Google Inc. Released under the GPL v2

import os, re, time, urllib2, urlparse, HTMLParser

from autotest_lib.mirror import database
from autotest_lib.client.common_lib import utils


class source(object):
    """
    Abstract Base Class for the source classes.
    """
    def __init__(self, database):
        self.database = database


    def _get_new_files(self, files):
        """
        Return a copy of "files" after filtering out known old files
        from "files".
        """
        old_files = self.database.get_dictionary()
        return dict(filter(lambda x: x[0] not in old_files, files.iteritems()))


    def get_new_files(self):
        raise NotImplemented('get_new_files not implemented')


    def store_files(self, files):
        self.database.merge_dictionary(files)


class rsync_source(source):
    _cmd_template = '/usr/bin/rsync -rltz --no-motd %s %s/%s'

    def __init__(self, database, prefix, excludes = []):
        super(rsync_source, self).__init__(database)

        self.prefix = prefix
        self.exclude = ' '.join(['--exclude "' + x + '"' for x in excludes])
        self.sources = []


    def _parse_output(self, output, prefix):
        """
        Parse rsync's "ls -l" style output and return a dictionary of
        database.item indexed by the "name" field.
        """
        regex = re.compile(
            '-[rwx-]{9} +(\d+) (\d{4}/\d\d/\d\d \d\d:\d\d:\d\d) (.*)')
        res = {}
        for line in output.splitlines():
            match = regex.match(line)
            if match:
                groups = match.groups()
                timestamp = time.mktime(time.strptime(groups[1],
                                        '%Y/%m/%d %H:%M:%S'))
                if prefix:
                    fname = '%s/%s' % (prefix, groups[2])
                else:
                    fname = groups[2]

                item = database.item(fname, int(groups[0]), int(timestamp))
                res[item.name] = item

        return res


    def add_path(self, src, prefix=''):
        """
        Add paths to synchronize from the source.
        """
        self.sources.append((src, prefix))


    def get_new_files(self):
        """
        Implement source.get_new_files by using rsync listing feature.
        """
        files = {}
        for src, prefix in self.sources:
            output = utils.system_output(self._cmd_template %
                (self.exclude, self.prefix, src))
            files.update(self._parse_output(output, prefix))

        return self._get_new_files(files)


class _ahref_parser(HTMLParser.HTMLParser):
    def reset(self, url=None, pattern=None):
        HTMLParser.HTMLParser.reset(self)
        self.url = url
        self.pattern = pattern
        self.links = []


    def handle_starttag(self, tag, attrs):
        if tag == 'a':
            for name, value in attrs:
                if name == 'href':
                    # compose absolute URL if relative "href" found
                    url = urlparse.urljoin(self.url, value)
                    if self.pattern.match(url):
                        self.links.append(url)


    def get_ahref_list(self, url, pattern):
        self.reset(url, pattern)
        self.feed(urllib2.urlopen(url).read())
        self.close()

        return self.links


class url_source(source):
    """
    A simple URL based source that parses HTML to find references to
    kernel files.
    """
    _extension_pattern = re.compile(r'.*\.[^/.]+$')

    def __init__(self, database, prefix):
        super(url_source, self).__init__(database)
        self.prefix = prefix
        self.urls = []


    def add_url(self, url, pattern):
        """
        Add a URL path to a HTML document with links to kernel files.

        @param url: URL path to a HTML file with links to kernel files
                (can be either an absolute URL or one relative to self.prefix)
        @param pattern: regex pattern to filter kernel files links out of
                all othe links found in the HTML document
        """
        # if it does not have an extension then it's a directory and it needs
        # a trailing '/'. NOTE: there are some false positives such as
        # directories named "v2.6" where ".6" will be assumed to be extension.
        # In order for these to work the caller must provide a trailing /
        if url[-1:] != '/' and not self._extension_pattern.match(url):
            url = url + '/'
        self.urls.append((url, re.compile(pattern)))


    @staticmethod
    def _get_item(url):
        """
        Get a database.item object by fetching relevant HTTP information
        from the document pointed to by the given url.
        """
        try:
            info = urllib2.urlopen(url).info()
        except IOError, err:
            # file is referenced but does not exist
            print 'WARNING: %s' % err
            return None

        size = info.get('content-length')
        if size:
            size = int(size)
        else:
            size = -1

        timestamp = int(time.mktime(info.getdate('date')))
        if not timestamp:
            timestamp = 0

        return database.item(url, size, timestamp)


    def get_new_files(self):
        parser = _ahref_parser()

        files = {}
        for url, pattern in self.urls:
            links = parser.get_ahref_list(urlparse.urljoin(self.prefix, url),
                                          pattern)
            for link in links:
                item = self._get_item(link)
                if item:
                    files[item.name] = item

        return self._get_new_files(files)


class directory_source(source):
    """
    Source that finds kernel files by listing the contents of a directory.
    """
    def __init__(self, database, path):
        """
        Initialize a directory_source instance.

        @param database: Persistent database with known kernels information.
        @param path: Path to the directory with the kernel files found by
                this source.
        """
        super(directory_source, self).__init__(database)

        self._path = path


    def get_new_files(self, _stat_func=os.stat):
        """
        Main function, see source.get_new_files().

        @param _stat_func: Used for unit testing, if we stub os.stat in the
                unit test then unit test failures get reported confusingly
                because the unit test framework tries to stat() the unit test
                file.
        """
        all_files = {}
        for filename in os.listdir(self._path):
            full_filename = os.path.join(self._path, filename)
            try:
                stat_data = _stat_func(full_filename)
            except OSError:
                # File might have been removed/renamed since we listed the
                # directory so skip it.
                continue

            item = database.item(full_filename, stat_data.st_size,
                                 int(stat_data.st_mtime))
            all_files[filename] = item

        return self._get_new_files(all_files)