Files
Giacomo Vercesi dfb9a36bb2 pypeline: implement compression
Implement compression of objects before they are saved into the storage
provider. Each container type can specify which algorithm to use
(currently `none` or `zstd`) and the compression level.
2026-06-17 10:06:35 +02:00

129 lines
4.7 KiB
Python

#
# This file is distributed under the MIT License. See LICENSE.md for details.
#
from collections.abc import Buffer
from typing import Mapping, Sequence, final
from revng.pypeline.container import ConfigurationId, ContainerDeclaration, ContainerSet
from revng.pypeline.object import ObjectID
from revng.pypeline.storage.storage_provider import ContainerLocation, PipeDependencies
from revng.pypeline.storage.storage_provider import SavePointsRange, StorageProvider
from .requests import Requests
from .task import TaskArgument, TaskArgumentAccess
class SavePoint:
"""
A SavePoint is a task that satisfies its requests by looking up the requested objects in the
container it owns.
"""
__slots__: tuple = ("name", "arguments", "to_save")
def __init__(
self,
name: str,
to_save: Sequence[ContainerDeclaration],
):
self.name = name
self.arguments = [
TaskArgument(
name=declaration.name,
container_type=declaration.container_type,
access=TaskArgumentAccess.READ_WRITE,
)
for declaration in to_save
]
self.to_save = to_save
@final
def prerequisites_for(
self,
requests: Requests,
configuration_id: ConfigurationId,
storage_provider: StorageProvider,
savepoint_range: SavePointsRange,
) -> Requests:
"""
This method has the same semantics as Pipe.prerequisites_for, but it has
additional arguments that the Pipe should not have.
"""
result = requests.clone()
for decl, request in result.items():
if decl not in self.to_save:
# This savepoint does not care about this container
continue
# We have a match, so we can check if the objects are present
location = ContainerLocation(
savepoint_id=savepoint_range.start,
container_id=decl.name,
configuration_id=configuration_id,
)
found_objs = storage_provider.has(location, request)
# If so, we should REMOVE the found objects from the request
for found_obj in found_objs:
result[decl].remove(found_obj)
return result
def run(
self,
containers: ContainerSet,
incoming: Requests,
outgoing: Requests,
configuration_id: ConfigurationId,
storage_provider: StorageProvider,
savepoint_range: SavePointsRange,
pipes_dependencies: list[PipeDependencies],
):
"""
The savepoint will cache the incoming containers, and then fill the outgoing
containers with the cached data.
"""
assert set(containers.keys()).issuperset(
set(incoming.keys()) | set(outgoing.keys())
), "SavePoint containers must be a subset of incoming and outgoing requests."
objects: dict[ContainerLocation, Mapping[ObjectID, Buffer]] = {}
for decl in self.to_save:
location = ContainerLocation(
savepoint_id=savepoint_range.start,
container_id=decl.name,
configuration_id=configuration_id,
)
container_incoming = incoming.get(decl)
if len(container_incoming) > 0:
serialized_container = containers[decl].serialize(container_incoming)
compression = containers[decl].get_compression()
if compression is not None:
objects[location] = {
k: compression.compress(v) for k, v in serialized_container.items()
}
else:
objects[location] = serialized_container
storage_provider.add_objects(pipes_dependencies, objects)
for decl in self.to_save:
# Compute the actual set of objects to load, if an object has
# already been saved from incoming do not re-load it from storage
container_outgoing = outgoing.get(decl) - incoming.get(decl)
if len(container_outgoing) > 0:
location = ContainerLocation(
savepoint_id=savepoint_range.start,
container_id=decl.name,
configuration_id=configuration_id,
)
container_data = storage_provider.get(location, container_outgoing)
compression = containers[decl].get_compression()
if compression is not None:
containers[decl].deserialize(
{k: compression.decompress(v) for k, v in container_data.items()}
)
else:
containers[decl].deserialize(container_data)