Changeset - 82f818616265
[Not reviewed]
default
0 1 0
Søren Løvborg - 10 years ago 2016-03-08 12:28:06
sorenl@unity3d.com
db: cache SCM instance short-term (tied to SQLAlchemy session lifetime)

Repeatedly checking whether SCM instances are invalidated is slow, and
we don't actually _want_ SCM instances to invalidate half-way through
a request either.

Therefore cache them in on the db.Repository object, the lifetime
of which is directly tied to the lifetime of the SQLAlchemy session,
the lifetime of which is tied directly to the individual HTTP request.
This way, we only check for invalidation the first time the SCM instance
is accessed in a request.

This will improve performance in cases where we have (by definition) badly
written code that retrieves repo objects several times.
1 file changed with 5 insertions and 1 deletions:
0 comments (0 inline, 0 general)
kallithea/model/db.py
Show inline comments
 
@@ -1344,195 +1344,199 @@ class Repository(Base, BaseModel):
 
        """
 
        from kallithea.lib.vcs.backends.base import BaseChangeset
 
        if cs_cache is None:
 
            cs_cache = EmptyChangeset()
 
            # use no-cache version here
 
            scm_repo = self.scm_instance_no_cache()
 
            if scm_repo:
 
                cs_cache = scm_repo.get_changeset()
 

	
 
        if isinstance(cs_cache, BaseChangeset):
 
            cs_cache = cs_cache.__json__()
 

	
 
        if (not self.changeset_cache or cs_cache['raw_id'] != self.changeset_cache['raw_id']):
 
            _default = datetime.datetime.fromtimestamp(0)
 
            last_change = cs_cache.get('date') or _default
 
            log.debug('updated repo %s with new cs cache %s',
 
                      self.repo_name, cs_cache)
 
            self.updated_on = last_change
 
            self.changeset_cache = cs_cache
 
            Session().add(self)
 
            Session().commit()
 
        else:
 
            log.debug('changeset_cache for %s already up to date with %s',
 
                      self.repo_name, cs_cache['raw_id'])
 

	
 
    @property
 
    def tip(self):
 
        return self.get_changeset('tip')
 

	
 
    @property
 
    def author(self):
 
        return self.tip.author
 

	
 
    @property
 
    def last_change(self):
 
        return self.scm_instance.last_change
 

	
 
    def get_comments(self, revisions=None):
 
        """
 
        Returns comments for this repository grouped by revisions
 

	
 
        :param revisions: filter query by revisions only
 
        """
 
        cmts = ChangesetComment.query() \
 
            .filter(ChangesetComment.repo == self)
 
        if revisions is not None:
 
            if not revisions:
 
                return {} # don't use sql 'in' on empty set
 
            cmts = cmts.filter(ChangesetComment.revision.in_(revisions))
 
        grouped = collections.defaultdict(list)
 
        for cmt in cmts.all():
 
            grouped[cmt.revision].append(cmt)
 
        return grouped
 

	
 
    def statuses(self, revisions):
 
        """
 
        Returns statuses for this repository.
 
        PRs without any votes do _not_ show up as unreviewed.
 

	
 
        :param revisions: list of revisions to get statuses for
 
        """
 
        if not revisions:
 
            return {}
 

	
 
        statuses = ChangesetStatus.query() \
 
            .filter(ChangesetStatus.repo == self) \
 
            .filter(ChangesetStatus.version == 0) \
 
            .filter(ChangesetStatus.revision.in_(revisions))
 

	
 
        grouped = {}
 
        for stat in statuses.all():
 
            pr_id = pr_nice_id = pr_repo = None
 
            if stat.pull_request:
 
                pr_id = stat.pull_request.pull_request_id
 
                pr_nice_id = PullRequest.make_nice_id(pr_id)
 
                pr_repo = stat.pull_request.other_repo.repo_name
 
            grouped[stat.revision] = [str(stat.status), stat.status_lbl,
 
                                      pr_id, pr_repo, pr_nice_id,
 
                                      stat.author]
 
        return grouped
 

	
 
    def _repo_size(self):
 
        from kallithea.lib import helpers as h
 
        log.debug('calculating repository size...')
 
        return h.format_byte_size(self.scm_instance.size)
 

	
 
    #==========================================================================
 
    # SCM CACHE INSTANCE
 
    #==========================================================================
 

	
 
    def set_invalidate(self):
 
        """
 
        Mark caches of this repo as invalid.
 
        """
 
        CacheInvalidation.set_invalidate(self.repo_name)
 

	
 
    _scm_instance = None
 

	
 
    @property
 
    def scm_instance(self):
 
        return self.scm_instance_cached()
 
        if self._scm_instance is None:
 
            self._scm_instance = self.scm_instance_cached()
 
        return self._scm_instance
 

	
 
    def scm_instance_cached(self, valid_cache_keys=None):
 
        @cache_region('long_term', 'scm_instance_cached')
 
        def _c(repo_name): # repo_name is just for the cache key
 
            log.debug('Creating new %s scm_instance and populating cache', repo_name)
 
            return self.scm_instance_no_cache()
 
        rn = self.repo_name
 

	
 
        valid = CacheInvalidation.test_and_set_valid(rn, None, valid_cache_keys=valid_cache_keys)
 
        if not valid:
 
            log.debug('Cache for %s invalidated, getting new object', rn)
 
            region_invalidate(_c, None, 'scm_instance_cached', rn)
 
        else:
 
            log.debug('Trying to get scm_instance of %s from cache', rn)
 
        return _c(rn)
 

	
 
    def scm_instance_no_cache(self):
 
        repo_full_path = safe_str(self.repo_full_path)
 
        alias = get_scm(repo_full_path)[0]
 
        log.debug('Creating instance of %s repository from %s',
 
                  alias, self.repo_full_path)
 
        backend = get_backend(alias)
 

	
 
        if alias == 'hg':
 
            repo = backend(repo_full_path, create=False,
 
                           baseui=self._ui)
 
        else:
 
            repo = backend(repo_full_path, create=False)
 

	
 
        return repo
 

	
 
    def __json__(self):
 
        return dict(landing_rev = self.landing_rev)
 

	
 
class RepoGroup(Base, BaseModel):
 
    __tablename__ = 'groups'
 
    __table_args__ = (
 
        UniqueConstraint('group_name', 'group_parent_id'),
 
        CheckConstraint('group_id != group_parent_id'),
 
        _table_args_default_dict,
 
    )
 
    __mapper_args__ = {'order_by': 'group_name'}
 

	
 
    SEP = ' » '
 

	
 
    group_id = Column(Integer(), unique=True, primary_key=True)
 
    group_name = Column(Unicode(255), nullable=False, unique=True)
 
    group_parent_id = Column(Integer(), ForeignKey('groups.group_id'), nullable=True)
 
    group_description = Column(Unicode(10000), nullable=False)
 
    enable_locking = Column(Boolean(), nullable=False, default=False)
 
    user_id = Column(Integer(), ForeignKey('users.user_id'), nullable=False)
 
    created_on = Column(DateTime(timezone=False), nullable=False, default=datetime.datetime.now)
 

	
 
    repo_group_to_perm = relationship('UserRepoGroupToPerm', cascade='all', order_by='UserRepoGroupToPerm.group_to_perm_id')
 
    users_group_to_perm = relationship('UserGroupRepoGroupToPerm', cascade='all')
 
    parent_group = relationship('RepoGroup', remote_side=group_id)
 
    user = relationship('User')
 

	
 
    def __init__(self, group_name='', parent_group=None):
 
        self.group_name = group_name
 
        self.parent_group = parent_group
 

	
 
    def __unicode__(self):
 
        return u"<%s('id:%s:%s')>" % (self.__class__.__name__, self.group_id,
 
                                      self.group_name)
 

	
 
    @classmethod
 
    def _generate_choice(cls, repo_group):
 
        """Return tuple with group_id and name as html literal"""
 
        from webhelpers.html import literal
 
        if repo_group is None:
 
            return (-1, u'-- %s --' % _('top level'))
 
        return repo_group.group_id, literal(cls.SEP.join(repo_group.full_path_splitted))
 

	
 
    @classmethod
 
    def groups_choices(cls, groups):
 
        """Return tuples with group_id and name as html literal."""
 
        return sorted((cls._generate_choice(g) for g in groups),
 
                      key=lambda c: c[1].split(cls.SEP))
 

	
 
    @classmethod
 
    def url_sep(cls):
 
        return URL_SEP
 

	
 
    @classmethod
 
    def get_by_group_name(cls, group_name, cache=False, case_insensitive=False):
 
        if case_insensitive:
 
            gr = cls.query() \
 
                .filter(func.lower(cls.group_name) == func.lower(group_name))
 
        else:
 
            gr = cls.query() \
 
                .filter(cls.group_name == group_name)
 
        if cache:
 
            gr = gr.options(FromCache(
 
                            "sql_cache_short",
 
                            "get_group_%s" % _hash_key(group_name)
0 comments (0 inline, 0 general)