Warning
This document is for an in-development version of Galaxy. You can alternatively view this page in the latest release if it exists or view the top of the latest release's documentation.
Source code for galaxy.model.deferred
import abc
import logging
import os
import shutil
from typing import (
cast,
NamedTuple,
Optional,
Union,
)
from sqlalchemy.orm import (
object_session,
Session,
)
from sqlalchemy.orm.exc import DetachedInstanceError
from galaxy.datatypes.sniff import (
convert_function,
stream_url_to_file,
)
from galaxy.exceptions import ObjectAttributeInvalidException
from galaxy.files import ConfiguredFileSources
from galaxy.model import (
Dataset,
DatasetCollection,
DatasetCollectionElement,
DatasetSource,
History,
HistoryDatasetAssociation,
HistoryDatasetCollectionAssociation,
LibraryDatasetDatasetAssociation,
)
from galaxy.model.base import transaction
from galaxy.objectstore import (
ObjectStore,
ObjectStorePopulator,
)
log = logging.getLogger(__name__)
[docs]class TransientDatasetPaths(NamedTuple):
external_filename: str
external_extra_files_path: str
metadata_files_dir: str
[docs]class TransientPathMapper:
[docs] @abc.abstractmethod
def transient_paths_for(self, old_dataset: Dataset) -> TransientDatasetPaths:
"""Map dataset to transient paths for detached HDAs.
Decide external_filename and external_extra_files_path that the supplied dataset's
materialized dataset should have its files written to.
"""
[docs]class SimpleTransientPathMapper(TransientPathMapper):
[docs] def transient_paths_for(self, old_dataset: Dataset) -> TransientDatasetPaths:
external_filename_basename = "dataset_%s.dat" % str(old_dataset.uuid)
external_filename = os.path.join(self._staging_directory, external_filename_basename)
external_extras_basename = "dataset_%s_files" % str(old_dataset.uuid)
external_extras = os.path.join(self._staging_directory, external_extras_basename)
return TransientDatasetPaths(external_filename, external_extras, self._staging_directory)
[docs]class DatasetInstanceMaterializer:
"""This class is responsible for ensuring dataset instances are not deferred."""
[docs] def __init__(
self,
attached: bool,
object_store_populator: Optional[ObjectStorePopulator] = None,
transient_path_mapper: Optional[TransientPathMapper] = None,
file_sources: Optional[ConfiguredFileSources] = None,
sa_session: Optional[Session] = None,
):
"""Constructor for DatasetInstanceMaterializer.
If attached is true, these objects should be created in a supplied object store.
If not, this class produces transient HDAs with external_filename and
external_extra_files_path set.
"""
self._attached = attached
self._transient_path_mapper = transient_path_mapper
self._object_store_populator = object_store_populator
self._file_sources = file_sources
self._sa_session = sa_session
[docs] def ensure_materialized(
self,
dataset_instance: Union[HistoryDatasetAssociation, LibraryDatasetDatasetAssociation],
target_history: Optional[History] = None,
) -> HistoryDatasetAssociation:
"""Create a new detached dataset instance from the supplied instance.
There will be times we want it usable as is without an objectstore and times
we want to place it in an objectstore.
"""
attached = self._attached
dataset = dataset_instance.dataset
if dataset.state != Dataset.states.DEFERRED and isinstance(dataset_instance, HistoryDatasetAssociation):
return dataset_instance
materialized_dataset = Dataset()
materialized_dataset.state = Dataset.states.OK
materialized_dataset.deleted = False
materialized_dataset.purged = False
materialized_dataset.sources = [s.copy() for s in dataset.sources]
materialized_dataset.hashes = [h.copy() for h in dataset.hashes]
target_source = self._find_closest_dataset_source(dataset)
transient_paths = None
if attached:
object_store_populator = self._object_store_populator
assert object_store_populator
object_store = object_store_populator.object_store
store_by = object_store.get_store_by(dataset)
if store_by == "id":
# we need a flush...
sa_session = self._sa_session
if sa_session is None:
sa_session = object_session(dataset_instance)
assert sa_session
sa_session.add(materialized_dataset)
with transaction(sa_session):
sa_session.commit()
object_store_populator.set_dataset_object_store_id(materialized_dataset)
path = self._stream_source(target_source, datatype=dataset_instance.datatype)
object_store.update_from_file(materialized_dataset, file_name=path)
else:
transient_path_mapper = self._transient_path_mapper
assert transient_path_mapper
transient_paths = transient_path_mapper.transient_paths_for(dataset)
# TODO: optimize this by streaming right to this path...
# TODO: take into acount transform and ensure we are and are not modifying the file as appropriate.
path = self._stream_source(target_source, datatype=dataset_instance.datatype)
shutil.move(path, transient_paths.external_filename)
materialized_dataset.external_filename = transient_paths.external_filename
history = target_history
if history is None and isinstance(dataset_instance, HistoryDatasetAssociation):
try:
history = dataset_instance.history
except DetachedInstanceError:
history = None
materialized_dataset_instance = HistoryDatasetAssociation(
create_dataset=False, # is the default but lets make this really clear...
history=history,
)
if attached:
sa_session = self._sa_session
if sa_session is None:
sa_session = object_session(dataset_instance)
assert sa_session
sa_session.add(materialized_dataset_instance)
materialized_dataset_instance.copy_from(
dataset_instance, new_dataset=materialized_dataset, include_tags=attached, include_metadata=True
)
require_metadata_regeneration = (
materialized_dataset_instance.has_metadata_files or materialized_dataset_instance.metadata_deferred
)
if require_metadata_regeneration:
materialized_dataset_instance.init_meta()
if transient_paths:
metadata_tmp_files_dir = transient_paths.metadata_files_dir
else:
# If metadata_tmp_files_dir is set we generate a MetdataTempFile,
# which we don't want when we're generating an attached materialized dataset instance
metadata_tmp_files_dir = None
materialized_dataset_instance.set_meta(metadata_tmp_files_dir=metadata_tmp_files_dir)
materialized_dataset_instance.metadata_deferred = False
return materialized_dataset_instance
def _stream_source(self, target_source: DatasetSource, datatype) -> str:
path = stream_url_to_file(target_source.source_uri, file_sources=self._file_sources)
transform = target_source.transform or []
to_posix_lines = False
spaces_to_tabs = False
datatype_groom = False
for transform_action in transform:
action = transform_action["action"]
if action == "to_posix_lines":
to_posix_lines = True
elif action == "spaces_to_tabs":
spaces_to_tabs = True
elif action == "datatype_groom":
datatype_groom = True
else:
raise Exception(f"Failed to materialize dataset, unknown transformation action {action} applied.")
if to_posix_lines or spaces_to_tabs:
convert_fxn = convert_function(to_posix_lines, spaces_to_tabs)
convert_result = convert_fxn(path, False)
assert convert_result.converted_path
path = convert_result.converted_path
if datatype_groom:
datatype.groom_dataset_content(path)
return path
def _find_closest_dataset_source(self, dataset: Dataset) -> DatasetSource:
best_source = None
for source in dataset.sources:
if source.extra_files_path:
continue
best_source = source
break
if best_source is None:
# TODO: implement test case...
raise ObjectAttributeInvalidException("dataset does not contain any valid dataset sources")
return best_source
CollectionInputT = Union[HistoryDatasetCollectionAssociation, DatasetCollectionElement]
[docs]def materialize_collection_input(
collection_input: CollectionInputT, materializer: DatasetInstanceMaterializer
) -> CollectionInputT:
if isinstance(collection_input, HistoryDatasetCollectionAssociation):
return materialize_collection_instance(
cast(HistoryDatasetCollectionAssociation, collection_input), materializer
)
else:
return _materialize_collection_element(cast(DatasetCollectionElement, collection_input), materializer)
[docs]def materialize_collection_instance(
hdca: HistoryDatasetCollectionAssociation, materializer: DatasetInstanceMaterializer
) -> HistoryDatasetCollectionAssociation:
if materializer._attached:
raise NotImplementedError("Materializing collections to attached collections unimplemented")
if not hdca.has_deferred_data:
return hdca
materialized_instance = HistoryDatasetCollectionAssociation()
materialized_instance.name = hdca.name
materialized_instance.collection = _materialize_collection(hdca.collection, materializer)
# TODO: tags
return materialized_instance
def _materialize_collection(
dataset_collection: DatasetCollection, materializer: DatasetInstanceMaterializer
) -> DatasetCollection:
materialized_collection = DatasetCollection()
materialized_elements = []
for element in dataset_collection.elements:
materialized_elements.append(_materialize_collection_element(element, materializer))
materialized_collection.elements = materialized_elements
return materialized_collection
def _materialize_collection_element(
element: DatasetCollectionElement, materializer: DatasetInstanceMaterializer
) -> DatasetCollectionElement:
materialized_object: Union[DatasetCollection, HistoryDatasetAssociation, LibraryDatasetDatasetAssociation]
if element.is_collection:
assert element.child_collection
materialized_object = _materialize_collection(element.child_collection, materializer)
else:
element_object = element.dataset_instance
assert element_object
materialized_object = materializer.ensure_materialized(element_object)
materialized_element = DatasetCollectionElement(
element=materialized_object,
element_index=element.element_index,
element_identifier=element.element_identifier,
)
return materialized_element
[docs]def materializer_factory(
attached: bool,
object_store: Optional[ObjectStore] = None,
object_store_populator: Optional[ObjectStorePopulator] = None,
transient_path_mapper: Optional[TransientPathMapper] = None,
transient_directory: Optional[str] = None,
file_sources: Optional[ConfiguredFileSources] = None,
sa_session: Optional[Session] = None,
) -> DatasetInstanceMaterializer:
if object_store_populator is None and object_store is not None:
object_store_populator = ObjectStorePopulator(object_store, None)
if transient_path_mapper is None and transient_directory is not None:
transient_path_mapper = SimpleTransientPathMapper(transient_directory)
return DatasetInstanceMaterializer(
attached,
object_store_populator=object_store_populator,
transient_path_mapper=transient_path_mapper,
file_sources=file_sources,
sa_session=sa_session,
)