##// END OF EJS Templates
tests: Fix --without-vcsserver parameter...
tests: Fix --without-vcsserver parameter Ensuring now that we always disable the vcsserver auto start for the test suite, since the test suite bootstraps its own vcsserver process. Also enhanced TestINI.

File last commit:

r613:af6adc7d default
r816:43fb99bf default
Show More
caches.py
234 lines | 8.1 KiB | text/x-python | PythonLexer
# -*- coding: utf-8 -*-
# Copyright (C) 2015-2016 RhodeCode GmbH
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License, version 3
# (only), as published by the Free Software Foundation.
#
# 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 Affero General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#
# This program is dual-licensed. If you wish to learn more about the
# RhodeCode Enterprise Edition, including its added features, Support services,
# and proprietary license terms, please see https://rhodecode.com/licenses/
import beaker
import logging
import threading
from beaker.cache import _cache_decorate, cache_regions, region_invalidate
from rhodecode.lib.utils import safe_str, md5
from rhodecode.model.db import Session, CacheKey, IntegrityError
log = logging.getLogger(__name__)
FILE_TREE = 'cache_file_tree'
FILE_TREE_META = 'cache_file_tree_metadata'
FILE_SEARCH_TREE_META = 'cache_file_search_metadata'
SUMMARY_STATS = 'cache_summary_stats'
# This list of caches gets purged when invalidation happens
USED_REPO_CACHES = (FILE_TREE, FILE_SEARCH_TREE_META)
DEFAULT_CACHE_MANAGER_CONFIG = {
'type': 'memorylru_base',
'max_items': 10240,
'key_length': 256,
'enabled': True
}
def configure_cache_region(
region_name, region_kw, default_cache_kw, default_expire=60):
default_type = default_cache_kw.get('type', 'memory')
default_lock_dir = default_cache_kw.get('lock_dir')
default_data_dir = default_cache_kw.get('data_dir')
region_kw['lock_dir'] = region_kw.get('lock_dir', default_lock_dir)
region_kw['data_dir'] = region_kw.get('data_dir', default_data_dir)
region_kw['type'] = region_kw.get('type', default_type)
region_kw['expire'] = int(region_kw.get('expire', default_expire))
beaker.cache.cache_regions[region_name] = region_kw
def get_cache_manager(region_name, cache_name, custom_ttl=None):
"""
Creates a Beaker cache manager. Such instance can be used like that::
_namespace = caches.get_repo_namespace_key(caches.XXX, repo_name)
cache_manager = caches.get_cache_manager('repo_cache_long', _namespace)
_cache_key = caches.compute_key_from_params(repo_name, commit.raw_id)
def heavy_compute():
...
result = cache_manager.get(_cache_key, createfunc=heavy_compute)
:param region_name: region from ini file
:param cache_name: custom cache name, usually prefix+repo_name. eg
file_switcher_repo1
:param custom_ttl: override .ini file timeout on this cache
:return: instance of cache manager
"""
cache_config = cache_regions.get(region_name, DEFAULT_CACHE_MANAGER_CONFIG)
if custom_ttl:
log.debug('Updating region %s with custom ttl: %s',
region_name, custom_ttl)
cache_config.update({'expire': custom_ttl})
return beaker.cache.Cache._get_cache(cache_name, cache_config)
def clear_cache_manager(cache_manager):
"""
namespace = 'foobar'
cache_manager = get_cache_manager('repo_cache_long', namespace)
clear_cache_manager(cache_manager)
"""
log.debug('Clearing all values for cache manager %s', cache_manager)
cache_manager.clear()
def clear_repo_caches(repo_name):
# invalidate cache manager for this repo
for prefix in USED_REPO_CACHES:
namespace = get_repo_namespace_key(prefix, repo_name)
cache_manager = get_cache_manager('repo_cache_long', namespace)
clear_cache_manager(cache_manager)
def compute_key_from_params(*args):
"""
Helper to compute key from given params to be used in cache manager
"""
return md5("_".join(map(safe_str, args)))
def get_repo_namespace_key(prefix, repo_name):
return '{0}_{1}'.format(prefix, compute_key_from_params(repo_name))
def conditional_cache(region, prefix, condition, func):
"""
Conditional caching function use like::
def _c(arg):
# heavy computation function
return data
# depending on the condition the compute is wrapped in cache or not
compute = conditional_cache('short_term', 'cache_desc',
condition=True, func=func)
return compute(arg)
:param region: name of cache region
:param prefix: cache region prefix
:param condition: condition for cache to be triggered, and
return data cached
:param func: wrapped heavy function to compute
"""
wrapped = func
if condition:
log.debug('conditional_cache: True, wrapping call of '
'func: %s into %s region cache', region, func)
cached_region = _cache_decorate((prefix,), None, None, region)
wrapped = cached_region(func)
return wrapped
class ActiveRegionCache(object):
def __init__(self, context):
self.context = context
def invalidate(self, *args, **kwargs):
return False
def compute(self):
log.debug('Context cache: getting obj %s from cache', self.context)
return self.context.compute_func(self.context.cache_key)
class FreshRegionCache(ActiveRegionCache):
def invalidate(self):
log.debug('Context cache: invalidating cache for %s', self.context)
region_invalidate(
self.context.compute_func, None, self.context.cache_key)
return True
class InvalidationContext(object):
def __repr__(self):
return '<InvalidationContext:{}[{}]>'.format(
safe_str(self.repo_name), safe_str(self.cache_type))
def __init__(self, compute_func, repo_name, cache_type,
raise_exception=False, thread_scoped=False):
self.compute_func = compute_func
self.repo_name = repo_name
self.cache_type = cache_type
self.cache_key = compute_key_from_params(
repo_name, cache_type)
self.raise_exception = raise_exception
# Append the thread id to the cache key if this invalidation context
# should be scoped to the current thread.
if thread_scoped:
thread_id = threading.current_thread().ident
self.cache_key = '{cache_key}_{thread_id}'.format(
cache_key=self.cache_key, thread_id=thread_id)
def get_cache_obj(self):
cache_key = CacheKey.get_cache_key(
self.repo_name, self.cache_type)
cache_obj = CacheKey.get_active_cache(cache_key)
if not cache_obj:
cache_obj = CacheKey(cache_key, self.repo_name)
return cache_obj
def __enter__(self):
"""
Test if current object is valid, and return CacheRegion function
that does invalidation and calculation
"""
self.cache_obj = self.get_cache_obj()
if self.cache_obj.cache_active:
# means our cache obj is existing and marked as it's
# cache is not outdated, we return BaseInvalidator
self.skip_cache_active_change = True
return ActiveRegionCache(self)
# the key is either not existing or set to False, we return
# the real invalidator which re-computes value. We additionally set
# the flag to actually update the Database objects
self.skip_cache_active_change = False
return FreshRegionCache(self)
def __exit__(self, exc_type, exc_val, exc_tb):
if self.skip_cache_active_change:
return
try:
self.cache_obj.cache_active = True
Session().add(self.cache_obj)
Session().commit()
except IntegrityError:
# if we catch integrity error, it means we inserted this object
# assumption is that's really an edge race-condition case and
# it's safe is to skip it
Session().rollback()
except Exception:
log.exception('Failed to commit on cache key update')
Session().rollback()
if self.raise_exception:
raise