##// END OF EJS Templates
release: Bump version 4.14.0 to 4.15.0
release: Bump version 4.14.0 to 4.15.0

File last commit:

r2883:f2837b35 default
r3207:dc2459c0 default
Show More
caching_query.py
298 lines | 10.9 KiB | text/x-python | PythonLexer
# -*- coding: utf-8 -*-
# Copyright (C) 2010-2018 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/
"""caching_query.py
Represent functions and classes
which allow the usage of Dogpile caching with SQLAlchemy.
Introduces a query option called FromCache.
The three new concepts introduced here are:
* CachingQuery - a Query subclass that caches and
retrieves results in/from dogpile.cache.
* FromCache - a query option that establishes caching
parameters on a Query
* RelationshipCache - a variant of FromCache which is specific
to a query invoked during a lazy load.
* _params_from_query - extracts value parameters from
a Query.
The rest of what's here are standard SQLAlchemy and
dogpile.cache constructs.
"""
from sqlalchemy.orm.interfaces import MapperOption
from sqlalchemy.orm.query import Query
from sqlalchemy.sql import visitors
from dogpile.cache.api import NO_VALUE
from rhodecode.lib.utils2 import safe_str
class CachingQuery(Query):
"""A Query subclass which optionally loads full results from a dogpile
cache region.
The CachingQuery optionally stores additional state that allows it to consult
a dogpile.cache cache before accessing the database, in the form
of a FromCache or RelationshipCache object. Each of these objects
refer to the name of a :class:`dogpile.cache.Region` that's been configured
and stored in a lookup dictionary. When such an object has associated
itself with the CachingQuery, the corresponding :class:`dogpile.cache.Region`
is used to locate a cached result. If none is present, then the
Query is invoked normally, the results being cached.
The FromCache and RelationshipCache mapper options below represent
the "public" method of configuring this state upon the CachingQuery.
"""
def _get_region(self):
from rhodecode.lib.rc_cache import region_meta
return region_meta.dogpile_cache_regions
def __init__(self, regions, *args, **kw):
self.cache_regions = regions or self._get_region()
Query.__init__(self, *args, **kw)
def __iter__(self):
"""override __iter__ to pull results from dogpile
if particular attributes have been configured.
Note that this approach does *not* detach the loaded objects from
the current session. If the cache backend is an in-process cache
(like "memory") and lives beyond the scope of the current session's
transaction, those objects may be expired. The method here can be
modified to first expunge() each loaded item from the current
session before returning the list of items, so that the items
in the cache are not the same ones in the current Session.
"""
super_ = super(CachingQuery, self)
if hasattr(self, '_cache_region'):
return self.get_value(createfunc=lambda: list(super_.__iter__()))
else:
return super_.__iter__()
def _execute_and_instances(self, context):
"""override _execute_and_instances to pull results from dogpile
if the query is invoked directly from an external context.
This method is necessary in order to maintain compatibility
with the "baked query" system now used by default in some
relationship loader scenarios. Note also the
RelationshipCache._generate_cache_key method which enables
the baked query to be used within lazy loads.
.. versionadded:: 1.2.7
"""
super_ = super(CachingQuery, self)
if context.query is not self and hasattr(self, '_cache_region'):
# special logic called when the Query._execute_and_instances()
# method is called directly from the baked query
return self.get_value(
createfunc=lambda: list(
super_._execute_and_instances(context)
)
)
else:
return super_._execute_and_instances(context)
def _get_cache_plus_key(self):
"""Return a cache region plus key."""
dogpile_region = self.cache_regions[self._cache_region.region]
if self._cache_region.cache_key:
key = self._cache_region.cache_key
else:
key = _key_from_query(self)
return dogpile_region, key
def invalidate(self):
"""Invalidate the cache value represented by this Query."""
dogpile_region, cache_key = self._get_cache_plus_key()
dogpile_region.delete(cache_key)
def get_value(self, merge=True, createfunc=None,
expiration_time=None, ignore_expiration=False):
"""Return the value from the cache for this query.
Raise KeyError if no value present and no
createfunc specified.
"""
dogpile_region, cache_key = self._get_cache_plus_key()
# ignore_expiration means, if the value is in the cache
# but is expired, return it anyway. This doesn't make sense
# with createfunc, which says, if the value is expired, generate
# a new value.
assert not ignore_expiration or not createfunc, \
"Can't ignore expiration and also provide createfunc"
if ignore_expiration or not createfunc:
cached_value = dogpile_region.get(cache_key,
expiration_time=expiration_time,
ignore_expiration=ignore_expiration)
else:
cached_value = dogpile_region.get_or_create(
cache_key,
createfunc,
expiration_time=expiration_time
)
if cached_value is NO_VALUE:
raise KeyError(cache_key)
if merge:
cached_value = self.merge_result(cached_value, load=False)
return cached_value
def set_value(self, value):
"""Set the value in the cache for this query."""
dogpile_region, cache_key = self._get_cache_plus_key()
dogpile_region.set(cache_key, value)
def query_callable(regions=None, query_cls=CachingQuery):
def query(*arg, **kw):
return query_cls(regions, *arg, **kw)
return query
def _key_from_query(query, qualifier=None):
"""Given a Query, create a cache key.
There are many approaches to this; here we use the simplest,
which is to create an md5 hash of the text of the SQL statement,
combined with stringified versions of all the bound parameters
within it. There's a bit of a performance hit with
compiling out "query.statement" here; other approaches include
setting up an explicit cache key with a particular Query,
then combining that with the bound parameter values.
"""
stmt = query.with_labels().statement
compiled = stmt.compile()
params = compiled.params
# here we return the key as a long string. our "key mangler"
# set up with the region will boil it down to an md5.
return " ".join(
[safe_str(compiled)] +
[safe_str(params[k]) for k in sorted(params)])
class FromCache(MapperOption):
"""Specifies that a Query should load results from a cache."""
propagate_to_loaders = False
def __init__(self, region="sql_cache_short", cache_key=None):
"""Construct a new FromCache.
:param region: the cache region. Should be a
region configured in the dictionary of dogpile
regions.
:param cache_key: optional. A string cache key
that will serve as the key to the query. Use this
if your query has a huge amount of parameters (such
as when using in_()) which correspond more simply to
some other identifier.
"""
self.region = region
self.cache_key = cache_key
def process_query(self, query):
"""Process a Query during normal loading operation."""
query._cache_region = self
class RelationshipCache(MapperOption):
"""Specifies that a Query as called within a "lazy load"
should load results from a cache."""
propagate_to_loaders = True
def __init__(self, attribute, region="sql_cache_short", cache_key=None):
"""Construct a new RelationshipCache.
:param attribute: A Class.attribute which
indicates a particular class relationship() whose
lazy loader should be pulled from the cache.
:param region: name of the cache region.
:param cache_key: optional. A string cache key
that will serve as the key to the query, bypassing
the usual means of forming a key from the Query itself.
"""
self.region = region
self.cache_key = cache_key
self._relationship_options = {
(attribute.property.parent.class_, attribute.property.key): self
}
def _generate_cache_key(self, path):
"""Indicate to the lazy-loader strategy that a "baked" query
may be used by returning ``None``.
If this method is omitted, the default implementation of
:class:`.MapperOption._generate_cache_key` takes place, which
returns ``False`` to disable the "baked" query from being used.
.. versionadded:: 1.2.7
"""
return None
def process_query_conditionally(self, query):
"""Process a Query that is used within a lazy loader.
(the process_query_conditionally() method is a SQLAlchemy
hook invoked only within lazyload.)
"""
if query._current_path:
mapper, prop = query._current_path[-2:]
key = prop.key
for cls in mapper.class_.__mro__:
if (cls, key) in self._relationship_options:
relationship_option = self._relationship_options[(cls, key)]
query._cache_region = relationship_option
break
def and_(self, option):
"""Chain another RelationshipCache option to this one.
While many RelationshipCache objects can be specified on a single
Query separately, chaining them together allows for a more efficient
lookup during load.
"""
self._relationship_options.update(option._relationship_options)
return self