class FormatterFactoryTestCase(unittest.TestCase, DatasetTestHelper):
"""Tests of the formatter factory infrastructure.
"""
def setUp(self):
self.id = 0
self.factory = FormatterFactory()
# Dummy FileDescriptor for testing getFormatter
self.fileDescriptor = FileDescriptor(
Location("/a/b/c", "d"),
StorageClass("DummyStorageClass", dict, None))
def assertIsFormatter(self, formatter):
"""Check that the supplied parameter is either a Formatter instance
or Formatter class."""
if inspect.isclass(formatter):
self.assertTrue(issubclass(formatter, Formatter),
f"Is {formatter} a Formatter")
else:
self.assertIsInstance(formatter, Formatter)
def testRegistry(self):
"""Check that formatters can be stored in the registry.
"""
formatterTypeName = "lsst.daf.butler.formatters.fitsCatalogFormatter.FitsCatalogFormatter"
storageClassName = "Image"
self.factory.registerFormatter(storageClassName, formatterTypeName)
f = self.factory.getFormatter(storageClassName, self.fileDescriptor)
self.assertIsFormatter(f)
self.assertEqual(f.name(), formatterTypeName)
self.assertIn(formatterTypeName, str(f))
self.assertIn(self.fileDescriptor.location.path, str(f))
fcls = self.factory.getFormatterClass(storageClassName)
self.assertIsFormatter(fcls)
# Defer the import so that we ensure that the infrastructure loaded
# it on demand previously
from lsst.daf.butler.formatters.fitsCatalogFormatter import FitsCatalogFormatter
self.assertEqual(type(f), FitsCatalogFormatter)
with self.assertRaises(TypeError):
# Requires a constructor parameter
self.factory.getFormatter(storageClassName)
with self.assertRaises(KeyError):
self.factory.getFormatter("Missing", self.fileDescriptor)
def testRegistryWithStorageClass(self):
"""Test that the registry can be given a StorageClass object.
"""
formatterTypeName = "lsst.daf.butler.formatters.yamlFormatter.YamlFormatter"
storageClassName = "TestClass"
sc = StorageClass(storageClassName, dict, None)
universe = DimensionUniverse()
datasetType = DatasetType("calexp", universe.empty, sc)
# Store using an instance
self.factory.registerFormatter(sc, formatterTypeName)
# Retrieve using the class
f = self.factory.getFormatter(sc, self.fileDescriptor)
self.assertIsFormatter(f)
self.assertEqual(f.fileDescriptor, self.fileDescriptor)
# Retrieve using the DatasetType
f2 = self.factory.getFormatter(datasetType, self.fileDescriptor)
self.assertIsFormatter(f2)
self.assertEqual(f.name(), f2.name())
# Class directly
f2cls = self.factory.getFormatterClass(datasetType)
self.assertIsFormatter(f2cls)
# This might defer the import, pytest may have already loaded it
from lsst.daf.butler.formatters.yamlFormatter import YamlFormatter
self.assertEqual(type(f), YamlFormatter)
with self.assertRaises(KeyError):
# Attempt to overwrite using a different value
self.factory.registerFormatter(
storageClassName,
"lsst.daf.butler.formatters.jsonFormatter.JsonFormatter")
def testRegistryConfig(self):
configFile = os.path.join(TESTDIR, "config", "basic",
"posixDatastore.yaml")
config = Config(configFile)
universe = DimensionUniverse()
self.factory.registerFormatters(config["datastore", "formatters"],
universe=universe)
# Create a DatasetRef with and without instrument matching the
# one in the config file.
dimensions = universe.extract(
("visit", "physical_filter", "instrument"))
sc = StorageClass("DummySC", dict, None)
refPviHsc = self.makeDatasetRef("pvi",
dimensions,
sc, {
"instrument": "DummyHSC",
"physical_filter": "v"
},
conform=False)
refPviHscFmt = self.factory.getFormatterClass(refPviHsc)
self.assertIsFormatter(refPviHscFmt)
self.assertIn("JsonFormatter", refPviHscFmt.name())
refPviNotHsc = self.makeDatasetRef("pvi",
dimensions,
sc, {
"instrument": "DummyNotHSC",
"physical_filter": "v"
},
conform=False)
refPviNotHscFmt = self.factory.getFormatterClass(refPviNotHsc)
self.assertIsFormatter(refPviNotHscFmt)
self.assertIn("PickleFormatter", refPviNotHscFmt.name())
# Create a DatasetRef that should fall back to using Dimensions
refPvixHsc = self.makeDatasetRef("pvix",
dimensions,
sc, {
"instrument": "DummyHSC",
"physical_filter": "v"
},
conform=False)
refPvixNotHscFmt = self.factory.getFormatterClass(refPvixHsc)
self.assertIsFormatter(refPvixNotHscFmt)
self.assertIn("PickleFormatter", refPvixNotHscFmt.name())
# Create a DatasetRef that should fall back to using StorageClass
dimensionsNoV = DimensionGraph(universe,
names=("physical_filter", "instrument"))
refPvixNotHscDims = self.makeDatasetRef("pvix",
dimensionsNoV,
sc, {
"instrument": "DummyHSC",
"physical_filter": "v"
},
conform=False)
refPvixNotHscDims_fmt = self.factory.getFormatterClass(
refPvixNotHscDims)
self.assertIsFormatter(refPvixNotHscDims_fmt)
self.assertIn("YamlFormatter", refPvixNotHscDims_fmt.name())
class PosixDatastore(Datastore):
"""Basic POSIX filesystem backed Datastore.
Attributes
----------
config : `DatastoreConfig`
Configuration used to create Datastore.
registry : `Registry`
`Registry` to use when recording the writing of Datasets.
root : `str`
Root directory of this `Datastore`.
locationFactory : `LocationFactory`
Factory for creating locations relative to this root.
formatterFactory : `FormatterFactory`
Factory for creating instances of formatters.
storageClassFactory : `StorageClassFactory`
Factory for creating storage class instances from name.
templates : `FileTemplates`
File templates that can be used by this `Datastore`.
name : `str`
Label associated with this Datastore.
Parameters
----------
config : `DatastoreConfig` or `str`
Configuration.
Raises
------
ValueError
If root location does not exist and ``create`` is `False` in the
configuration.
"""
defaultConfigFile = "datastores/posixDatastore.yaml"
"""Path to configuration defaults. Relative to $DAF_BUTLER_DIR/config or
absolute path. Can be None if no defaults specified.
"""
RecordTuple = namedtuple(
"PosixDatastoreRecord",
["formatter", "path", "storage_class", "checksum", "file_size"])
@classmethod
def setConfigRoot(cls, root, config, full, overwrite=True):
"""Set any filesystem-dependent config options for this Datastore to
be appropriate for a new empty repository with the given root.
Parameters
----------
root : `str`
Filesystem path to the root of the data repository.
config : `Config`
A `Config` to update. Only the subset understood by
this component will be updated. Will not expand
defaults.
full : `Config`
A complete config with all defaults expanded that can be
converted to a `DatastoreConfig`. Read-only and will not be
modified by this method.
Repository-specific options that should not be obtained
from defaults when Butler instances are constructed
should be copied from ``full`` to ``config``.
overwrite : `bool`, optional
If `False`, do not modify a value in ``config`` if the value
already exists. Default is always to overwrite with the provided
``root``.
Notes
-----
If a keyword is explicitly defined in the supplied ``config`` it
will not be overridden by this method if ``overwrite`` is `False`.
This allows explicit values set in external configs to be retained.
"""
Config.updateParameters(DatastoreConfig,
config,
full,
toUpdate={"root": root},
toCopy=("cls", ("records", "table")),
overwrite=overwrite)
def __init__(self, config, registry, butlerRoot=None):
super().__init__(config, registry)
if "root" not in self.config:
raise ValueError("No root directory specified in configuration")
# Name ourselves either using an explicit name or a name
# derived from the (unexpanded) root
if "name" in self.config:
self.name = self.config["name"]
else:
self.name = "POSIXDatastore@{}".format(self.config["root"])
# Support repository relocation in config
self.root = replaceRoot(self.config["root"], butlerRoot)
if not os.path.isdir(self.root):
if "create" not in self.config or not self.config["create"]:
raise ValueError(f"No valid root at: {self.root}")
safeMakeDir(self.root)
self.locationFactory = LocationFactory(self.root)
self.formatterFactory = FormatterFactory()
self.storageClassFactory = StorageClassFactory()
# Now associate formatters with storage classes
self.formatterFactory.registerFormatters(
self.config["formatters"], universe=self.registry.dimensions)
# Read the file naming templates
self.templates = FileTemplates(self.config["templates"],
universe=self.registry.dimensions)
# And read the constraints list
constraintsConfig = self.config.get("constraints")
self.constraints = Constraints(constraintsConfig,
universe=self.registry.dimensions)
# Storage of paths and formatters, keyed by dataset_id
types = {
"path": str,
"formatter": str,
"storage_class": str,
"file_size": int,
"checksum": str,
"dataset_id": int
}
lengths = {
"path": 256,
"formatter": 128,
"storage_class": 64,
"checksum": 128
}
self.records = DatabaseDict.fromConfig(self.config["records"],
types=types,
value=self.RecordTuple,
key="dataset_id",
lengths=lengths,
registry=registry)
def __str__(self):
return self.root
def addStoredFileInfo(self, ref, info):
"""Record internal storage information associated with this
`DatasetRef`
Parameters
----------
ref : `DatasetRef`
The Dataset that has been stored.
info : `StoredFileInfo`
Metadata associated with the stored Dataset.
"""
self.records[ref.id] = self.RecordTuple(
formatter=info.formatter,
path=info.path,
storage_class=info.storageClass.name,
checksum=info.checksum,
file_size=info.size)
def removeStoredFileInfo(self, ref):
"""Remove information about the file associated with this dataset.
Parameters
----------
ref : `DatasetRef`
The Dataset that has been removed.
"""
del self.records[ref.id]
def getStoredFileInfo(self, ref):
"""Retrieve information associated with file stored in this
`Datastore`.
Parameters
----------
ref : `DatasetRef`
The Dataset that is to be queried.
Returns
-------
info : `StoredFileInfo`
Stored information about this file and its formatter.
Raises
------
KeyError
Dataset with that id can not be found.
"""
record = self.records.get(ref.id, None)
if record is None:
raise KeyError(
"Unable to retrieve formatter associated with Dataset {}".
format(ref.id))
# Convert name of StorageClass to instance
storageClass = self.storageClassFactory.getStorageClass(
record.storage_class)
return StoredFileInfo(record.formatter,
record.path,
storageClass,
checksum=record.checksum,
size=record.file_size)
def exists(self, ref):
"""Check if the dataset exists in the datastore.
Parameters
----------
ref : `DatasetRef`
Reference to the required dataset.
Returns
-------
exists : `bool`
`True` if the entity exists in the `Datastore`.
"""
# Get the file information (this will fail if no file)
try:
storedFileInfo = self.getStoredFileInfo(ref)
except KeyError:
return False
# Use the path to determine the location
location = self.locationFactory.fromPath(storedFileInfo.path)
return os.path.exists(location.path)
def get(self, ref, parameters=None):
"""Load an InMemoryDataset from the store.
Parameters
----------
ref : `DatasetRef`
Reference to the required Dataset.
parameters : `dict`
`StorageClass`-specific parameters that specify, for example,
a slice of the Dataset to be loaded.
Returns
-------
inMemoryDataset : `object`
Requested Dataset or slice thereof as an InMemoryDataset.
Raises
------
FileNotFoundError
Requested dataset can not be retrieved.
TypeError
Return value from formatter has unexpected type.
ValueError
Formatter failed to process the dataset.
"""
log.debug("Retrieve %s from %s with parameters %s", ref, self.name,
parameters)
# Get file metadata and internal metadata
try:
storedFileInfo = self.getStoredFileInfo(ref)
except KeyError:
raise FileNotFoundError(
"Could not retrieve Dataset {}".format(ref))
# Use the path to determine the location
location = self.locationFactory.fromPath(storedFileInfo.path)
# Too expensive to recalculate the checksum on fetch
# but we can check size and existence
if not os.path.exists(location.path):
raise FileNotFoundError(
"Dataset with Id {} does not seem to exist at"
" expected location of {}".format(ref.id, location.path))
stat = os.stat(location.path)
size = stat.st_size
if size != storedFileInfo.size:
raise RuntimeError(
"Integrity failure in Datastore. Size of file {} ({}) does not"
" match recorded size of {}".format(location.path, size,
storedFileInfo.size))
# We have a write storage class and a read storage class and they
# can be different for concrete composites.
readStorageClass = ref.datasetType.storageClass
writeStorageClass = storedFileInfo.storageClass
# Check that the supplied parameters are suitable for the type read
readStorageClass.validateParameters(parameters)
# Is this a component request?
component = ref.datasetType.component()
formatter = getInstanceOf(storedFileInfo.formatter)
formatterParams, assemblerParams = formatter.segregateParameters(
parameters)
try:
result = formatter.read(FileDescriptor(
location,
readStorageClass=readStorageClass,
storageClass=writeStorageClass,
parameters=parameters),
component=component)
except Exception as e:
raise ValueError(
"Failure from formatter for Dataset {}: {}".format(ref.id, e))
# Process any left over parameters
if parameters:
result = readStorageClass.assembler().handleParameters(
result, assemblerParams)
# Validate the returned data type matches the expected data type
pytype = readStorageClass.pytype
if pytype and not isinstance(result, pytype):
raise TypeError(
"Got type {} from formatter but expected {}".format(
type(result), pytype))
return result
@transactional
def put(self, inMemoryDataset, ref):
"""Write a InMemoryDataset with a given `DatasetRef` to the store.
Parameters
----------
inMemoryDataset : `object`
The Dataset to store.
ref : `DatasetRef`
Reference to the associated Dataset.
Raises
------
TypeError
Supplied object and storage class are inconsistent.
DatasetTypeNotSupportedError
The associated `DatasetType` is not handled by this datastore.
Notes
-----
If the datastore is configured to reject certain dataset types it
is possible that the put will fail and raise a
`DatasetTypeNotSupportedError`. The main use case for this is to
allow `ChainedDatastore` to put to multiple datastores without
requiring that every datastore accepts the dataset.
"""
datasetType = ref.datasetType
storageClass = datasetType.storageClass
# Sanity check
if not isinstance(inMemoryDataset, storageClass.pytype):
raise TypeError("Inconsistency between supplied object ({}) "
"and storage class type ({})".format(
type(inMemoryDataset), storageClass.pytype))
# Confirm that we can accept this dataset
if not self.constraints.isAcceptable(ref):
# Raise rather than use boolean return value.
raise DatasetTypeNotSupportedError(
f"Dataset {ref} has been rejected by this datastore via"
" configuration.")
# Work out output file name
try:
template = self.templates.getTemplate(ref)
except KeyError as e:
raise DatasetTypeNotSupportedError(
f"Unable to find template for {ref}") from e
location = self.locationFactory.fromPath(template.format(ref))
# Get the formatter based on the storage class
try:
formatter = self.formatterFactory.getFormatter(ref)
except KeyError as e:
raise DatasetTypeNotSupportedError(
f"Unable to find formatter for {ref}") from e
storageDir = os.path.dirname(location.path)
if not os.path.isdir(storageDir):
with self._transaction.undoWith("mkdir", os.rmdir, storageDir):
safeMakeDir(storageDir)
# Write the file
predictedFullPath = os.path.join(self.root,
formatter.predictPath(location))
if os.path.exists(predictedFullPath):
raise FileExistsError(
f"Cannot write file for ref {ref} as "
f"output file {predictedFullPath} already exists")
with self._transaction.undoWith("write", os.remove, predictedFullPath):
path = formatter.write(
inMemoryDataset,
FileDescriptor(location, storageClass=storageClass))
assert predictedFullPath == os.path.join(self.root, path)
log.debug("Wrote file to %s", path)
self.ingest(path, ref, formatter=formatter)
@transactional
def ingest(self, path, ref, formatter=None, transfer=None):
"""Add an on-disk file with the given `DatasetRef` to the store,
possibly transferring it.
The caller is responsible for ensuring that the given (or predicted)
Formatter is consistent with how the file was written; `ingest` will
in general silently ignore incorrect formatters (as it cannot
efficiently verify their correctness), deferring errors until ``get``
is first called on the ingested dataset.
Parameters
----------
path : `str`
File path. Treated as relative to the repository root if not
absolute.
ref : `DatasetRef`
Reference to the associated Dataset.
formatter : `Formatter` (optional)
Formatter that should be used to retreive the Dataset. If not
provided, the formatter will be constructed according to
Datastore configuration.
transfer : str (optional)
If not None, must be one of 'move', 'copy', 'hardlink', or
'symlink' indicating how to transfer the file. The new
filename and location will be determined via template substitution,
as with ``put``. If the file is outside the datastore root, it
must be transferred somehow.
Raises
------
RuntimeError
Raised if ``transfer is None`` and path is outside the repository
root.
FileNotFoundError
Raised if the file at ``path`` does not exist.
FileExistsError
Raised if ``transfer is not None`` but a file already exists at the
location computed from the template.
DatasetTypeNotSupportedError
The associated `DatasetType` is not handled by this datastore.
"""
# Confirm that we can accept this dataset
if not self.constraints.isAcceptable(ref):
# Raise rather than use boolean return value.
raise DatasetTypeNotSupportedError(
f"Dataset {ref} has been rejected by this datastore via"
" configuration.")
if formatter is None:
formatter = self.formatterFactory.getFormatter(ref)
fullPath = os.path.normpath(os.path.join(self.root, path))
if not os.path.exists(fullPath):
raise FileNotFoundError(
"File at '{}' does not exist; note that paths to ingest are "
"assumed to be relative to self.root unless they are absolute."
.format(fullPath))
if transfer is None:
if os.path.isabs(path):
absRoot = os.path.abspath(self.root)
if os.path.commonpath([absRoot, path]) != absRoot:
raise RuntimeError(
"'{}' is not inside repository root '{}'".format(
path, self.root))
path = os.path.relpath(path, absRoot)
elif path.startswith(os.path.pardir):
raise RuntimeError(
f"'{path}' is outside repository root '{self.root}'")
else:
template = self.templates.getTemplate(ref)
location = self.locationFactory.fromPath(template.format(ref))
newPath = formatter.predictPath(location)
newFullPath = os.path.join(self.root, newPath)
if os.path.exists(newFullPath):
raise FileExistsError(
"File '{}' already exists".format(newFullPath))
storageDir = os.path.dirname(newFullPath)
if not os.path.isdir(storageDir):
with self._transaction.undoWith("mkdir", os.rmdir, storageDir):
safeMakeDir(storageDir)
if transfer == "move":
with self._transaction.undoWith("move", shutil.move,
newFullPath, fullPath):
shutil.move(fullPath, newFullPath)
elif transfer == "copy":
with self._transaction.undoWith("copy", os.remove,
newFullPath):
shutil.copy(fullPath, newFullPath)
elif transfer == "hardlink":
with self._transaction.undoWith("hardlink", os.unlink,
newFullPath):
os.link(fullPath, newFullPath)
elif transfer == "symlink":
with self._transaction.undoWith("symlink", os.unlink,
newFullPath):
os.symlink(fullPath, newFullPath)
else:
raise NotImplementedError(
"Transfer type '{}' not supported.".format(transfer))
path = newPath
fullPath = newFullPath
# Create Storage information in the registry
checksum = self.computeChecksum(fullPath)
stat = os.stat(fullPath)
size = stat.st_size
self.registry.addDatasetLocation(ref, self.name)
# Associate this dataset with the formatter for later read.
fileInfo = StoredFileInfo(formatter,
path,
ref.datasetType.storageClass,
size=size,
checksum=checksum)
# TODO: this is only transactional if the DatabaseDict uses
# self.registry internally. Probably need to add
# transactions to DatabaseDict to do better than that.
self.addStoredFileInfo(ref, fileInfo)
# Register all components with same information
for compRef in ref.components.values():
self.registry.addDatasetLocation(compRef, self.name)
self.addStoredFileInfo(compRef, fileInfo)
def getUri(self, ref, predict=False):
"""URI to the Dataset.
Parameters
----------
ref : `DatasetRef`
Reference to the required Dataset.
predict : `bool`
If `True`, allow URIs to be returned of datasets that have not
been written.
Returns
-------
uri : `str`
URI string pointing to the Dataset within the datastore. If the
Dataset does not exist in the datastore, and if ``predict`` is
`True`, the URI will be a prediction and will include a URI
fragment "#predicted".
If the datastore does not have entities that relate well
to the concept of a URI the returned URI string will be
descriptive. The returned URI is not guaranteed to be obtainable.
Raises
------
FileNotFoundError
A URI has been requested for a dataset that does not exist and
guessing is not allowed.
"""
# if this has never been written then we have to guess
if not self.exists(ref):
if not predict:
raise FileNotFoundError(
"Dataset {} not in this datastore".format(ref))
template = self.templates.getTemplate(ref)
location = self.locationFactory.fromPath(
template.format(ref) + "#predicted")
else:
# If this is a ref that we have written we can get the path.
# Get file metadata and internal metadata
storedFileInfo = self.getStoredFileInfo(ref)
# Use the path to determine the location
location = self.locationFactory.fromPath(storedFileInfo.path)
return location.uri
def remove(self, ref):
"""Indicate to the Datastore that a Dataset can be removed.
.. warning::
This method does not support transactions; removals are
immediate, cannot be undone, and are not guaranteed to
be atomic if deleting either the file or the internal
database records fails.
Parameters
----------
ref : `DatasetRef`
Reference to the required Dataset.
Raises
------
FileNotFoundError
Attempt to remove a dataset that does not exist.
"""
# Get file metadata and internal metadata
try:
storedFileInfo = self.getStoredFileInfo(ref)
except KeyError:
raise FileNotFoundError(
"Requested dataset ({}) does not exist".format(ref))
location = self.locationFactory.fromPath(storedFileInfo.path)
if not os.path.exists(location.path):
raise FileNotFoundError("No such file: {0}".format(location.uri))
os.remove(location.path)
# Remove rows from registries
self.removeStoredFileInfo(ref)
self.registry.removeDatasetLocation(self.name, ref)
for compRef in ref.components.values():
self.registry.removeDatasetLocation(self.name, compRef)
self.removeStoredFileInfo(compRef)
def transfer(self, inputDatastore, ref):
"""Retrieve a Dataset from an input `Datastore`,
and store the result in this `Datastore`.
Parameters
----------
inputDatastore : `Datastore`
The external `Datastore` from which to retreive the Dataset.
ref : `DatasetRef`
Reference to the required Dataset in the input data store.
"""
assert inputDatastore is not self # unless we want it for renames?
inMemoryDataset = inputDatastore.get(ref)
return self.put(inMemoryDataset, ref)
def validateConfiguration(self, entities, logFailures=False):
"""Validate some of the configuration for this datastore.
Parameters
----------
entities : iterable of `DatasetRef`, `DatasetType`, or `StorageClass`
Entities to test against this configuration. Can be differing
types.
logFailures : `bool`, optional
If `True`, output a log message for every validation error
detected.
Raises
------
DatastoreValidationError
Raised if there is a validation problem with a configuration.
All the problems are reported in a single exception.
Notes
-----
This method checks that all the supplied entities have valid file
templates and also have formatters defined.
"""
templateFailed = None
try:
self.templates.validateTemplates(entities, logFailures=logFailures)
except FileTemplateValidationError as e:
templateFailed = str(e)
formatterFailed = []
for entity in entities:
try:
self.formatterFactory.getFormatter(entity)
except KeyError as e:
formatterFailed.append(str(e))
if logFailures:
log.fatal("Formatter failure: %s", e)
if templateFailed or formatterFailed:
messages = []
if templateFailed:
messages.append(templateFailed)
if formatterFailed:
messages.append(",".join(formatterFailed))
msg = ";\n".join(messages)
raise DatastoreValidationError(msg)
def getLookupKeys(self):
# Docstring is inherited from base class
return self.templates.getLookupKeys() | self.formatterFactory.getLookupKeys() | \
self.constraints.getLookupKeys()
def validateKey(self, lookupKey, entity):
# Docstring is inherited from base class
# The key can be valid in either formatters or templates so we can
# only check the template if it exists
if lookupKey in self.templates:
try:
self.templates[lookupKey].validateTemplate(entity)
except FileTemplateValidationError as e:
raise DatastoreValidationError(e) from e
@staticmethod
def computeChecksum(filename, algorithm="blake2b", block_size=8192):
"""Compute the checksum of the supplied file.
Parameters
----------
filename : `str`
Name of file to calculate checksum from.
algorithm : `str`, optional
Name of algorithm to use. Must be one of the algorithms supported
by :py:class`hashlib`.
block_size : `int`
Number of bytes to read from file at one time.
Returns
-------
hexdigest : `str`
Hex digest of the file.
"""
if algorithm not in hashlib.algorithms_guaranteed:
raise NameError(
"The specified algorithm '{}' is not supported by hashlib".
format(algorithm))
hasher = hashlib.new(algorithm)
with open(filename, "rb") as f:
for chunk in iter(lambda: f.read(block_size), b""):
hasher.update(chunk)
return hasher.hexdigest()
class FileLikeDatastore(GenericBaseDatastore):
"""Generic Datastore for file-based implementations.
Should always be sub-classed since key abstract methods are missing.
Parameters
----------
config : `DatastoreConfig` or `str`
Configuration as either a `Config` object or URI to file.
Raises
------
ValueError
If root location does not exist and ``create`` is `False` in the
configuration.
"""
defaultConfigFile = None
"""Path to configuration defaults. Relative to $DAF_BUTLER_DIR/config or
absolute path. Can be None if no defaults specified.
"""
root: str
"""Root directory or URI of this `Datastore`."""
locationFactory: LocationFactory
"""Factory for creating locations relative to the datastore root."""
formatterFactory: FormatterFactory
"""Factory for creating instances of formatters."""
templates: FileTemplates
"""File templates that can be used by this `Datastore`."""
@classmethod
def setConfigRoot(cls, root, config, full, overwrite=True):
"""Set any filesystem-dependent config options for this Datastore to
be appropriate for a new empty repository with the given root.
Parameters
----------
root : `str`
URI to the root of the data repository.
config : `Config`
A `Config` to update. Only the subset understood by
this component will be updated. Will not expand
defaults.
full : `Config`
A complete config with all defaults expanded that can be
converted to a `DatastoreConfig`. Read-only and will not be
modified by this method.
Repository-specific options that should not be obtained
from defaults when Butler instances are constructed
should be copied from ``full`` to ``config``.
overwrite : `bool`, optional
If `False`, do not modify a value in ``config`` if the value
already exists. Default is always to overwrite with the provided
``root``.
Notes
-----
If a keyword is explicitly defined in the supplied ``config`` it
will not be overridden by this method if ``overwrite`` is `False`.
This allows explicit values set in external configs to be retained.
"""
Config.updateParameters(DatastoreConfig,
config,
full,
toUpdate={"root": root},
toCopy=("cls", ("records", "table")),
overwrite=overwrite)
@classmethod
def makeTableSpec(cls):
return ddl.TableSpec(
fields=NamedValueSet([
ddl.FieldSpec(name="dataset_id",
dtype=Integer,
primaryKey=True),
ddl.FieldSpec(name="path",
dtype=String,
length=256,
nullable=False),
ddl.FieldSpec(name="formatter",
dtype=String,
length=128,
nullable=False),
ddl.FieldSpec(name="storage_class",
dtype=String,
length=64,
nullable=False),
# TODO: should checksum be Base64Bytes instead?
ddl.FieldSpec(name="checksum",
dtype=String,
length=128,
nullable=True),
ddl.FieldSpec(name="file_size", dtype=Integer, nullable=True),
]),
unique=frozenset(),
foreignKeys=[
ddl.ForeignKeySpec(table="dataset",
source=("dataset_id", ),
target=("dataset_id", ),
onDelete="CASCADE")
])
def __init__(self, config, registry, butlerRoot=None):
super().__init__(config, registry)
if "root" not in self.config:
raise ValueError("No root directory specified in configuration")
# Name ourselves either using an explicit name or a name
# derived from the (unexpanded) root
if "name" in self.config:
self.name = self.config["name"]
else:
# We use the unexpanded root in the name to indicate that this
# datastore can be moved without having to update registry.
self.name = "{}@{}".format(
type(self).__name__, self.config["root"])
# Support repository relocation in config
# Existence of self.root is checked in subclass
self.root = replaceRoot(self.config["root"], butlerRoot)
self.locationFactory = LocationFactory(self.root)
self.formatterFactory = FormatterFactory()
# Now associate formatters with storage classes
self.formatterFactory.registerFormatters(
self.config["formatters"], universe=self.registry.dimensions)
# Read the file naming templates
self.templates = FileTemplates(self.config["templates"],
universe=self.registry.dimensions)
# Storage of paths and formatters, keyed by dataset_id
self._tableName = self.config["records", "table"]
try:
registry.registerOpaqueTable(self._tableName, self.makeTableSpec())
except ReadOnlyDatabaseError:
# If the database is read only and we just tried and failed to
# create a table, it means someone is trying to create a read-only
# butler client for an empty repo. That should be okay, as long
# as they then try to get any datasets before some other client
# creates the table. Chances are they'rejust validating
# configuration.
pass
# Determine whether checksums should be used
self.useChecksum = self.config.get("checksum", True)
def __str__(self):
return self.root
def addStoredItemInfo(self, refs, infos):
# Docstring inherited from GenericBaseDatastore
records = []
for ref, info in zip(refs, infos):
records.append(
dict(dataset_id=ref.id,
formatter=info.formatter,
path=info.path,
storage_class=info.storageClass.name,
checksum=info.checksum,
file_size=info.file_size))
self.registry.insertOpaqueData(self._tableName, *records)
def getStoredItemInfo(self, ref):
# Docstring inherited from GenericBaseDatastore
records = list(
self.registry.fetchOpaqueData(self._tableName, dataset_id=ref.id))
if len(records) == 0:
raise KeyError(
f"Unable to retrieve location associated with Dataset {ref}.")
assert len(
records
) == 1, "Primary key constraint should make more than one result impossible."
record = records[0]
# Convert name of StorageClass to instance
storageClass = self.storageClassFactory.getStorageClass(
record["storage_class"])
return StoredFileInfo(formatter=record["formatter"],
path=record["path"],
storageClass=storageClass,
checksum=record["checksum"],
file_size=record["file_size"])
def _registered_refs_per_artifact(self, pathInStore):
"""Return all dataset refs associated with the supplied path.
Parameters
----------
pathInStore : `str`
Path of interest in the data store.
Returns
-------
ids : `set` of `int`
All `DatasetRef` IDs associated with this path.
"""
records = list(
self.registry.fetchOpaqueData(self._tableName, path=pathInStore))
ids = {r["dataset_id"] for r in records}
return ids
def removeStoredItemInfo(self, ref):
# Docstring inherited from GenericBaseDatastore
self.registry.deleteOpaqueData(self._tableName, dataset_id=ref.id)
def _get_dataset_location_info(self, ref):
"""Find the `Location` of the requested dataset in the
`Datastore` and the associated stored file information.
Parameters
----------
ref : `DatasetRef`
Reference to the required `Dataset`.
Returns
-------
location : `Location`
Location of the dataset within the datastore.
Returns `None` if the dataset can not be located.
info : `StoredFileInfo`
Stored information about this file and its formatter.
"""
# Get the file information (this will fail if no file)
try:
storedFileInfo = self.getStoredItemInfo(ref)
except KeyError:
return None, None
# Use the path to determine the location
location = self.locationFactory.fromPath(storedFileInfo.path)
return location, storedFileInfo
def _can_remove_dataset_artifact(self, ref):
"""Check that there is only one dataset associated with the
specified artifact.
Parameters
----------
ref : `DatasetRef`
Dataset to be removed.
Returns
-------
can_remove : `Bool`
True if the artifact can be safely removed.
"""
storedFileInfo = self.getStoredItemInfo(ref)
# Get all entries associated with this path
allRefs = self._registered_refs_per_artifact(storedFileInfo.path)
if not allRefs:
raise RuntimeError(
f"Datastore inconsistency error. {storedFileInfo.path} not in registry"
)
# Get all the refs associated with this dataset if it is a composite
theseRefs = {
r.id
for r in itertools.chain([ref], ref.components.values())
}
# Remove these refs from all the refs and if there is nothing left
# then we can delete
remainingRefs = allRefs - theseRefs
if remainingRefs:
return False
return True
def _prepare_for_get(self, ref, parameters=None):
"""Check parameters for ``get`` and obtain formatter and
location.
Parameters
----------
ref : `DatasetRef`
Reference to the required Dataset.
parameters : `dict`
`StorageClass`-specific parameters that specify, for example,
a slice of the Dataset to be loaded.
Returns
-------
getInfo : `DatastoreFileGetInformation`
Parameters needed to retrieve the file.
"""
log.debug("Retrieve %s from %s with parameters %s", ref, self.name,
parameters)
# Get file metadata and internal metadata
location, storedFileInfo = self._get_dataset_location_info(ref)
if location is None:
raise FileNotFoundError(f"Could not retrieve Dataset {ref}.")
# We have a write storage class and a read storage class and they
# can be different for concrete composites.
readStorageClass = ref.datasetType.storageClass
writeStorageClass = storedFileInfo.storageClass
# Check that the supplied parameters are suitable for the type read
readStorageClass.validateParameters(parameters)
# Is this a component request?
component = ref.datasetType.component()
formatter = getInstanceOf(
storedFileInfo.formatter,
FileDescriptor(location,
readStorageClass=readStorageClass,
storageClass=writeStorageClass,
parameters=parameters), ref.dataId)
formatterParams, assemblerParams = formatter.segregateParameters()
return DatastoreFileGetInformation(location, formatter, storedFileInfo,
assemblerParams, component,
readStorageClass)
def _prepare_for_put(self, inMemoryDataset, ref):
"""Check the arguments for ``put`` and obtain formatter and
location.
Parameters
----------
inMemoryDataset : `object`
The Dataset to store.
ref : `DatasetRef`
Reference to the associated Dataset.
Returns
-------
location : `Location`
The location to write the dataset.
formatter : `Formatter`
The `Formatter` to use to write the dataset.
Raises
------
TypeError
Supplied object and storage class are inconsistent.
DatasetTypeNotSupportedError
The associated `DatasetType` is not handled by this datastore.
"""
self._validate_put_parameters(inMemoryDataset, ref)
# Work out output file name
try:
template = self.templates.getTemplate(ref)
except KeyError as e:
raise DatasetTypeNotSupportedError(
f"Unable to find template for {ref}") from e
location = self.locationFactory.fromPath(template.format(ref))
# Get the formatter based on the storage class
storageClass = ref.datasetType.storageClass
try:
formatter = self.formatterFactory.getFormatter(
ref, FileDescriptor(location, storageClass=storageClass),
ref.dataId)
except KeyError as e:
raise DatasetTypeNotSupportedError(
f"Unable to find formatter for {ref}") from e
return location, formatter
@abstractmethod
def _standardizeIngestPath(self,
path: str,
*,
transfer: Optional[str] = None) -> str:
"""Standardize the path of a to-be-ingested file.
Parameters
----------
path : `str`
Path of a file to be ingested.
transfer : `str`, optional
How (and whether) the dataset should be added to the datastore.
If `None` (default), the file must already be in a location
appropriate for the datastore (e.g. within its root directory),
and will not be moved. Other choices include "move", "copy",
"symlink", and "hardlink". This is provided only so
`NotImplementedError` can be raised if the mode is not supported;
actual transfers are deferred to `_extractIngestInfo`.
Returns
-------
path : `str`
New path in what the datastore considers standard form.
Notes
-----
Subclasses of `FileLikeDatastore` should implement this method instead
of `_prepIngest`. It should not modify the data repository or given
file in any way.
Raises
------
NotImplementedError
Raised if the datastore does not support the given transfer mode
(including the case where ingest is not supported at all).
FileNotFoundError
Raised if one of the given files does not exist.
"""
raise NotImplementedError("Must be implemented by subclasses.")
@abstractmethod
def _extractIngestInfo(self,
path: str,
ref: DatasetRef,
*,
formatter: Type[Formatter],
transfer: Optional[str] = None) -> StoredFileInfo:
"""Relocate (if necessary) and extract `StoredFileInfo` from a
to-be-ingested file.
Parameters
----------
path : `str`
Path of a file to be ingested.
ref : `DatasetRef`
Reference for the dataset being ingested. Guaranteed to have
``dataset_id not None`.
formatter : `type`
`Formatter` subclass to use for this dataset.
transfer : `str`, optional
How (and whether) the dataset should be added to the datastore.
If `None` (default), the file must already be in a location
appropriate for the datastore (e.g. within its root directory),
and will not be modified. Other choices include "move", "copy",
"symlink", and "hardlink".
Returns
-------
info : `StoredFileInfo`
Internal datastore record for this file. This will be inserted by
the caller; the `_extractIngestInfo` is only resposible for
creating and populating the struct.
Raises
------
FileNotFoundError
Raised if one of the given files does not exist.
FileExistsError
Raised if transfer is not `None` but the (internal) location the
file would be moved to is already occupied.
"""
raise NotImplementedError("Must be implemented by subclasses.")
def _prepIngest(self,
*datasets: FileDataset,
transfer: Optional[str] = None) -> _IngestPrepData:
# Docstring inherited from Datastore._prepIngest.
filtered = []
for dataset in datasets:
acceptable = [
ref for ref in dataset.refs
if self.constraints.isAcceptable(ref)
]
if not acceptable:
continue
else:
dataset.refs = acceptable
if dataset.formatter is None:
dataset.formatter = self.formatterFactory.getFormatterClass(
dataset.refs[0])
else:
dataset.formatter = getClassOf(dataset.formatter)
dataset.path = self._standardizeIngestPath(dataset.path,
transfer=transfer)
filtered.append(dataset)
return _IngestPrepData(filtered)
@transactional
def _finishIngest(self,
prepData: Datastore.IngestPrepData,
*,
transfer: Optional[str] = None):
# Docstring inherited from Datastore._finishIngest.
refsAndInfos = []
for dataset in prepData.datasets:
# Do ingest as if the first dataset ref is associated with the file
info = self._extractIngestInfo(dataset.path,
dataset.refs[0],
formatter=dataset.formatter,
transfer=transfer)
refsAndInfos.extend([(ref, info) for ref in dataset.refs])
self._register_datasets(refsAndInfos)
def getUri(self, ref, predict=False):
"""URI to the Dataset.
Parameters
----------
ref : `DatasetRef`
Reference to the required Dataset.
predict : `bool`
If `True`, allow URIs to be returned of datasets that have not
been written.
Returns
-------
uri : `str`
URI string pointing to the Dataset within the datastore. If the
Dataset does not exist in the datastore, and if ``predict`` is
`True`, the URI will be a prediction and will include a URI
fragment "#predicted".
If the datastore does not have entities that relate well
to the concept of a URI the returned URI string will be
descriptive. The returned URI is not guaranteed to be obtainable.
Raises
------
FileNotFoundError
A URI has been requested for a dataset that does not exist and
guessing is not allowed.
Notes
-----
When a predicted URI is requested an attempt will be made to form
a reasonable URI based on file templates and the expected formatter.
"""
# if this has never been written then we have to guess
if not self.exists(ref):
if not predict:
raise FileNotFoundError(
"Dataset {} not in this datastore".format(ref))
template = self.templates.getTemplate(ref)
location = self.locationFactory.fromPath(template.format(ref))
storageClass = ref.datasetType.storageClass
formatter = self.formatterFactory.getFormatter(
ref, FileDescriptor(location, storageClass=storageClass))
# Try to use the extension attribute but ignore problems if the
# formatter does not define one.
try:
location = formatter.makeUpdatedLocation(location)
except Exception:
# Use the default extension
pass
# Add a URI fragment to indicate this is a guess
return location.uri + "#predicted"
# If this is a ref that we have written we can get the path.
# Get file metadata and internal metadata
storedFileInfo = self.getStoredItemInfo(ref)
# Use the path to determine the location
location = self.locationFactory.fromPath(storedFileInfo.path)
return location.uri
def validateConfiguration(self, entities, logFailures=False):
"""Validate some of the configuration for this datastore.
Parameters
----------
entities : iterable of `DatasetRef`, `DatasetType`, or `StorageClass`
Entities to test against this configuration. Can be differing
types.
logFailures : `bool`, optional
If `True`, output a log message for every validation error
detected.
Raises
------
DatastoreValidationError
Raised if there is a validation problem with a configuration.
All the problems are reported in a single exception.
Notes
-----
This method checks that all the supplied entities have valid file
templates and also have formatters defined.
"""
templateFailed = None
try:
self.templates.validateTemplates(entities, logFailures=logFailures)
except FileTemplateValidationError as e:
templateFailed = str(e)
formatterFailed = []
for entity in entities:
try:
self.formatterFactory.getFormatterClass(entity)
except KeyError as e:
formatterFailed.append(str(e))
if logFailures:
log.fatal("Formatter failure: %s", e)
if templateFailed or formatterFailed:
messages = []
if templateFailed:
messages.append(templateFailed)
if formatterFailed:
messages.append(",".join(formatterFailed))
msg = ";\n".join(messages)
raise DatastoreValidationError(msg)
def getLookupKeys(self):
# Docstring is inherited from base class
return self.templates.getLookupKeys() | self.formatterFactory.getLookupKeys() | \
self.constraints.getLookupKeys()
def validateKey(self, lookupKey, entity):
# Docstring is inherited from base class
# The key can be valid in either formatters or templates so we can
# only check the template if it exists
if lookupKey in self.templates:
try:
self.templates[lookupKey].validateTemplate(entity)
except FileTemplateValidationError as e:
raise DatastoreValidationError(e) from e
class FormatterFactoryTestCase(unittest.TestCase, DatasetTestHelper):
"""Tests of the formatter factory infrastructure.
"""
def setUp(self):
self.id = 0
self.factory = FormatterFactory()
# Dummy FileDescriptor for testing getFormatter
self.fileDescriptor = FileDescriptor(Location("/a/b/c", "d"),
StorageClass("DummyStorageClass", dict, None))
def assertIsFormatter(self, formatter):
"""Check that the supplied parameter is either a Formatter instance
or Formatter class."""
if inspect.isclass(formatter):
self.assertTrue(issubclass(formatter, Formatter), f"Is {formatter} a Formatter")
else:
self.assertIsInstance(formatter, Formatter)
def testFormatter(self):
"""Check basic parameter exceptions"""
f = DoNothingFormatter(self.fileDescriptor)
self.assertEqual(f.writeRecipes, {})
self.assertEqual(f.writeParameters, {})
self.assertIn("DoNothingFormatter", repr(f))
with self.assertRaises(TypeError):
DoNothingFormatter()
with self.assertRaises(ValueError):
DoNothingFormatter(self.fileDescriptor, writeParameters={"param1": 0})
with self.assertRaises(RuntimeError):
DoNothingFormatter(self.fileDescriptor, writeRecipes={"label": "value"})
with self.assertRaises(NotImplementedError):
f.makeUpdatedLocation(Location("a", "b"))
with self.assertRaises(NotImplementedError):
f.write("str")
def testExtensionValidation(self):
"""Test extension validation"""
for file, single_ok, multi_ok in (("e.fits", True, True),
("e.fit", False, True),
("e.fits.fz", False, True),
("e.txt", False, False),
("e.1.4.fits", True, True),
("e.3.fit", False, True),
("e.1.4.fits.gz", False, True),
):
loc = Location("/a/b/c", file)
for formatter, passes in ((SingleExtensionFormatter, single_ok),
(MultipleExtensionsFormatter, multi_ok)):
if passes:
formatter.validateExtension(loc)
else:
with self.assertRaises(ValueError):
formatter.validateExtension(loc)
def testRegistry(self):
"""Check that formatters can be stored in the registry.
"""
formatterTypeName = "lsst.daf.butler.tests.deferredFormatter.DeferredFormatter"
storageClassName = "Image"
self.factory.registerFormatter(storageClassName, formatterTypeName)
f = self.factory.getFormatter(storageClassName, self.fileDescriptor)
self.assertIsFormatter(f)
self.assertEqual(f.name(), formatterTypeName)
self.assertIn(formatterTypeName, str(f))
self.assertIn(self.fileDescriptor.location.path, str(f))
fcls = self.factory.getFormatterClass(storageClassName)
self.assertIsFormatter(fcls)
# Defer the import so that we ensure that the infrastructure loaded
# it on demand previously
from lsst.daf.butler.tests.deferredFormatter import DeferredFormatter
self.assertEqual(type(f), DeferredFormatter)
with self.assertRaises(TypeError):
# Requires a constructor parameter
self.factory.getFormatter(storageClassName)
with self.assertRaises(KeyError):
self.factory.getFormatter("Missing", self.fileDescriptor)
# Check that a bad formatter path fails
storageClassName = "BadImage"
self.factory.registerFormatter(storageClassName, "lsst.daf.butler.tests.deferredFormatter.Unknown")
with self.assertRaises(ImportError):
self.factory.getFormatter(storageClassName, self.fileDescriptor)
def testRegistryWithStorageClass(self):
"""Test that the registry can be given a StorageClass object.
"""
formatterTypeName = "lsst.daf.butler.formatters.yaml.YamlFormatter"
storageClassName = "TestClass"
sc = StorageClass(storageClassName, dict, None)
universe = DimensionUniverse()
datasetType = DatasetType("calexp", universe.empty, sc)
# Store using an instance
self.factory.registerFormatter(sc, formatterTypeName)
# Retrieve using the class
f = self.factory.getFormatter(sc, self.fileDescriptor)
self.assertIsFormatter(f)
self.assertEqual(f.fileDescriptor, self.fileDescriptor)
# Retrieve using the DatasetType
f2 = self.factory.getFormatter(datasetType, self.fileDescriptor)
self.assertIsFormatter(f2)
self.assertEqual(f.name(), f2.name())
# Class directly
f2cls = self.factory.getFormatterClass(datasetType)
self.assertIsFormatter(f2cls)
# This might defer the import, pytest may have already loaded it
from lsst.daf.butler.formatters.yaml import YamlFormatter
self.assertEqual(type(f), YamlFormatter)
with self.assertRaises(KeyError):
# Attempt to overwrite using a different value
self.factory.registerFormatter(storageClassName,
"lsst.daf.butler.formatters.json.JsonFormatter")
def testRegistryConfig(self):
configFile = os.path.join(TESTDIR, "config", "basic", "posixDatastore.yaml")
config = Config(configFile)
universe = DimensionUniverse()
self.factory.registerFormatters(config["datastore", "formatters"], universe=universe)
# Create a DatasetRef with and without instrument matching the
# one in the config file.
dimensions = universe.extract(("visit", "physical_filter", "instrument"))
sc = StorageClass("DummySC", dict, None)
refPviHsc = self.makeDatasetRef("pvi", dimensions, sc, {"instrument": "DummyHSC",
"physical_filter": "v"},
conform=False)
refPviHscFmt = self.factory.getFormatterClass(refPviHsc)
self.assertIsFormatter(refPviHscFmt)
self.assertIn("JsonFormatter", refPviHscFmt.name())
refPviNotHsc = self.makeDatasetRef("pvi", dimensions, sc, {"instrument": "DummyNotHSC",
"physical_filter": "v"},
conform=False)
refPviNotHscFmt = self.factory.getFormatterClass(refPviNotHsc)
self.assertIsFormatter(refPviNotHscFmt)
self.assertIn("PickleFormatter", refPviNotHscFmt.name())
# Create a DatasetRef that should fall back to using Dimensions
refPvixHsc = self.makeDatasetRef("pvix", dimensions, sc, {"instrument": "DummyHSC",
"physical_filter": "v"},
conform=False)
refPvixNotHscFmt = self.factory.getFormatterClass(refPvixHsc)
self.assertIsFormatter(refPvixNotHscFmt)
self.assertIn("PickleFormatter", refPvixNotHscFmt.name())
# Create a DatasetRef that should fall back to using StorageClass
dimensionsNoV = DimensionGraph(universe, names=("physical_filter", "instrument"))
refPvixNotHscDims = self.makeDatasetRef("pvix", dimensionsNoV, sc, {"instrument": "DummyHSC",
"physical_filter": "v"},
conform=False)
refPvixNotHscDims_fmt = self.factory.getFormatterClass(refPvixNotHscDims)
self.assertIsFormatter(refPvixNotHscDims_fmt)
self.assertIn("YamlFormatter", refPvixNotHscDims_fmt.name())
# Check that parameters are stored
refParam = self.makeDatasetRef("paramtest", dimensions, sc, {"instrument": "DummyNotHSC",
"physical_filter": "v"},
conform=False)
lookup, refParam_fmt, kwargs = self.factory.getFormatterClassWithMatch(refParam)
self.assertIn("writeParameters", kwargs)
expected = {"max": 5, "min": 2, "comment": "Additional commentary", "recipe": "recipe1"}
self.assertEqual(kwargs["writeParameters"], expected)
self.assertIn("FormatterTest", refParam_fmt.name())
f = self.factory.getFormatter(refParam, self.fileDescriptor)
self.assertEqual(f.writeParameters, expected)
f = self.factory.getFormatter(refParam, self.fileDescriptor, writeParameters={"min": 22,
"extra": 50})
self.assertEqual(f.writeParameters, {"max": 5, "min": 22, "comment": "Additional commentary",
"extra": 50, "recipe": "recipe1"})
self.assertIn("recipe1", f.writeRecipes)
self.assertEqual(f.writeParameters["recipe"], "recipe1")
with self.assertRaises(ValueError):
# "new" is not allowed as a write parameter
self.factory.getFormatter(refParam, self.fileDescriptor, writeParameters={"new": 1})
with self.assertRaises(RuntimeError):
# "mode" is a required recipe parameter
self.factory.getFormatter(refParam, self.fileDescriptor, writeRecipes={"recipe3": {"notmode": 1}})
class FormatterFactoryTestCase(unittest.TestCase, DatasetTestHelper):
"""Tests of the formatter factory infrastructure.
"""
def setUp(self):
self.id = 0
self.factory = FormatterFactory()
def testRegistry(self):
"""Check that formatters can be stored in the registry.
"""
formatterTypeName = "lsst.daf.butler.formatters.fitsCatalogFormatter.FitsCatalogFormatter"
storageClassName = "Image"
self.factory.registerFormatter(storageClassName, formatterTypeName)
f = self.factory.getFormatter(storageClassName)
self.assertIsInstance(f, Formatter)
# Defer the import so that we ensure that the infrastructure loaded
# it on demand previously
from lsst.daf.butler.formatters.fitsCatalogFormatter import FitsCatalogFormatter
self.assertEqual(type(f), FitsCatalogFormatter)
with self.assertRaises(KeyError):
f = self.factory.getFormatter("Missing")
def testRegistryWithStorageClass(self):
"""Test that the registry can be given a StorageClass object.
"""
formatterTypeName = "lsst.daf.butler.formatters.yamlFormatter.YamlFormatter"
storageClassName = "TestClass"
sc = StorageClass(storageClassName, dict, None)
universe = DimensionUniverse.fromConfig()
datasetType = DatasetType("calexp", universe.extract([]), sc)
# Store using an instance
self.factory.registerFormatter(sc, formatterTypeName)
# Retrieve using the class
f = self.factory.getFormatter(sc)
self.assertIsInstance(f, Formatter)
# Retrieve using the DatasetType
f2 = self.factory.getFormatter(datasetType)
self.assertIsInstance(f, Formatter)
self.assertEqual(f.name(), f2.name())
# This might defer the import, pytest may have already loaded it
from lsst.daf.butler.formatters.yamlFormatter import YamlFormatter
self.assertEqual(type(f), YamlFormatter)
with self.assertRaises(KeyError):
# Attempt to overwrite using a different value
self.factory.registerFormatter(
storageClassName,
"lsst.daf.butler.formatters.jsonFormatter.JsonFormatter")
def testRegistryConfig(self):
configFile = os.path.join(TESTDIR, "config", "basic",
"posixDatastore.yaml")
config = Config(configFile)
universe = DimensionUniverse.fromConfig()
self.factory.registerFormatters(config["datastore", "formatters"],
universe=universe)
# Create a DatasetRef with and without instrument matching the
# one in the config file.
dimensions = universe.extract(
("visit", "physical_filter", "instrument"))
sc = StorageClass("DummySC", dict, None)
refPviHsc = self.makeDatasetRef("pvi", dimensions, sc, {
"instrument": "DummyHSC",
"physical_filter": "v"
})
refPviHscFmt = self.factory.getFormatter(refPviHsc)
self.assertIsInstance(refPviHscFmt, Formatter)
self.assertIn("JsonFormatter", refPviHscFmt.name())
refPviNotHsc = self.makeDatasetRef("pvi", dimensions, sc, {
"instrument": "DummyNotHSC",
"physical_filter": "v"
})
refPviNotHscFmt = self.factory.getFormatter(refPviNotHsc)
self.assertIsInstance(refPviNotHscFmt, Formatter)
self.assertIn("PickleFormatter", refPviNotHscFmt.name())
# Create a DatasetRef that should fall back to using Dimensions
refPvixHsc = self.makeDatasetRef("pvix", dimensions, sc, {
"instrument": "DummyHSC",
"physical_filter": "v"
})
refPvixNotHscFmt = self.factory.getFormatter(refPvixHsc)
self.assertIsInstance(refPvixNotHscFmt, Formatter)
self.assertIn("PickleFormatter", refPvixNotHscFmt.name())
# Create a DatasetRef that should fall back to using StorageClass
dimensionsNoV = universe.extract(("physical_filter", "instrument"))
refPvixNotHscDims = self.makeDatasetRef("pvix", dimensionsNoV, sc, {
"instrument": "DummyHSC",
"physical_filter": "v"
})
refPvixNotHscDims_fmt = self.factory.getFormatter(refPvixNotHscDims)
self.assertIsInstance(refPvixNotHscDims_fmt, Formatter)
self.assertIn("YamlFormatter", refPvixNotHscDims_fmt.name())
class FileLikeDatastore(GenericBaseDatastore):
"""Generic Datastore for file-based implementations.
Should always be sub-classed since key abstract methods are missing.
Parameters
----------
config : `DatastoreConfig` or `str`
Configuration as either a `Config` object or URI to file.
Raises
------
ValueError
If root location does not exist and ``create`` is `False` in the
configuration.
"""
defaultConfigFile = None
"""Path to configuration defaults. Relative to $DAF_BUTLER_DIR/config or
absolute path. Can be None if no defaults specified.
"""
Record: ClassVar[Type] = DatastoreRecord
"""Class to use to represent datastore records."""
root: str
"""Root directory or URI of this `Datastore`."""
locationFactory: LocationFactory
"""Factory for creating locations relative to the datastore root."""
formatterFactory: FormatterFactory
"""Factory for creating instances of formatters."""
templates: FileTemplates
"""File templates that can be used by this `Datastore`."""
records: DatabaseDict
"""Place to store internal records about datasets."""
@classmethod
def setConfigRoot(cls, root, config, full, overwrite=True):
"""Set any filesystem-dependent config options for this Datastore to
be appropriate for a new empty repository with the given root.
Parameters
----------
root : `str`
URI to the root of the data repository.
config : `Config`
A `Config` to update. Only the subset understood by
this component will be updated. Will not expand
defaults.
full : `Config`
A complete config with all defaults expanded that can be
converted to a `DatastoreConfig`. Read-only and will not be
modified by this method.
Repository-specific options that should not be obtained
from defaults when Butler instances are constructed
should be copied from ``full`` to ``config``.
overwrite : `bool`, optional
If `False`, do not modify a value in ``config`` if the value
already exists. Default is always to overwrite with the provided
``root``.
Notes
-----
If a keyword is explicitly defined in the supplied ``config`` it
will not be overridden by this method if ``overwrite`` is `False`.
This allows explicit values set in external configs to be retained.
"""
Config.updateParameters(DatastoreConfig, config, full,
toUpdate={"root": root},
toCopy=("cls", ("records", "table")), overwrite=overwrite)
def __init__(self, config, registry, butlerRoot=None):
super().__init__(config, registry)
if "root" not in self.config:
raise ValueError("No root directory specified in configuration")
# Name ourselves either using an explicit name or a name
# derived from the (unexpanded) root
if "name" in self.config:
self.name = self.config["name"]
else:
# We use the unexpanded root in the name to indicate that this
# datastore can be moved without having to update registry.
self.name = "{}@{}".format(type(self).__name__,
self.config["root"])
# Support repository relocation in config
# Existence of self.root is checked in subclass
self.root = replaceRoot(self.config["root"], butlerRoot)
self.locationFactory = LocationFactory(self.root)
self.formatterFactory = FormatterFactory()
# Now associate formatters with storage classes
self.formatterFactory.registerFormatters(self.config["formatters"],
universe=self.registry.dimensions)
# Read the file naming templates
self.templates = FileTemplates(self.config["templates"],
universe=self.registry.dimensions)
# Storage of paths and formatters, keyed by dataset_id
self.records = DatabaseDict.fromConfig(self.config["records"],
value=self.Record, key="dataset_id",
registry=registry)
def __str__(self):
return self.root
def _info_to_record(self, info):
"""Convert a `StoredFileInfo` to a suitable database record.
Parameters
----------
info : `StoredFileInfo`
Metadata associated with the stored Dataset.
Returns
-------
record : `DatastoreRecord`
Record to be stored.
"""
return self.Record(formatter=info.formatter, path=info.path,
storage_class=info.storageClass.name,
checksum=info.checksum, file_size=info.file_size)
def _record_to_info(self, record):
"""Convert a record associated with this dataset to a `StoredItemInfo`
Parameters
----------
record : `DatastoreRecord`
Object stored in the record table.
Returns
-------
info : `StoredFileInfo`
The information associated with this dataset record as a Python
class.
"""
# Convert name of StorageClass to instance
storageClass = self.storageClassFactory.getStorageClass(record.storage_class)
return StoredFileInfo(formatter=record.formatter,
path=record.path,
storageClass=storageClass,
checksum=record.checksum,
file_size=record.file_size)
def _get_dataset_location_info(self, ref):
"""Find the `Location` of the requested dataset in the
`Datastore` and the associated stored file information.
Parameters
----------
ref : `DatasetRef`
Reference to the required `Dataset`.
Returns
-------
location : `Location`
Location of the dataset within the datastore.
Returns `None` if the dataset can not be located.
info : `StoredFileInfo`
Stored information about this file and its formatter.
"""
# Get the file information (this will fail if no file)
try:
storedFileInfo = self.getStoredItemInfo(ref)
except KeyError:
return None, None
# Use the path to determine the location
location = self.locationFactory.fromPath(storedFileInfo.path)
return location, storedFileInfo
def _prepare_for_get(self, ref, parameters=None):
"""Check parameters for ``get`` and obtain formatter and
location.
Parameters
----------
ref : `DatasetRef`
Reference to the required Dataset.
parameters : `dict`
`StorageClass`-specific parameters that specify, for example,
a slice of the Dataset to be loaded.
Returns
-------
getInfo : `DatastoreFileGetInformation`
Parameters needed to retrieve the file.
"""
log.debug("Retrieve %s from %s with parameters %s", ref, self.name, parameters)
# Get file metadata and internal metadata
location, storedFileInfo = self._get_dataset_location_info(ref)
if location is None:
raise FileNotFoundError(f"Could not retrieve Dataset {ref}.")
# We have a write storage class and a read storage class and they
# can be different for concrete composites.
readStorageClass = ref.datasetType.storageClass
writeStorageClass = storedFileInfo.storageClass
# Check that the supplied parameters are suitable for the type read
readStorageClass.validateParameters(parameters)
# Is this a component request?
component = ref.datasetType.component()
formatter = getInstanceOf(storedFileInfo.formatter,
FileDescriptor(location, readStorageClass=readStorageClass,
storageClass=writeStorageClass, parameters=parameters))
formatterParams, assemblerParams = formatter.segregateParameters()
return DatastoreFileGetInformation(location, formatter, storedFileInfo,
assemblerParams, component, readStorageClass)
def _prepare_for_put(self, inMemoryDataset, ref):
"""Check the arguments for ``put`` and obtain formatter and
location.
Parameters
----------
inMemoryDataset : `object`
The Dataset to store.
ref : `DatasetRef`
Reference to the associated Dataset.
Returns
-------
location : `Location`
The location to write the dataset.
formatter : `Formatter`
The `Formatter` to use to write the dataset.
Raises
------
TypeError
Supplied object and storage class are inconsistent.
DatasetTypeNotSupportedError
The associated `DatasetType` is not handled by this datastore.
"""
self._validate_put_parameters(inMemoryDataset, ref)
# Work out output file name
try:
template = self.templates.getTemplate(ref)
except KeyError as e:
raise DatasetTypeNotSupportedError(f"Unable to find template for {ref}") from e
location = self.locationFactory.fromPath(template.format(ref))
# Get the formatter based on the storage class
storageClass = ref.datasetType.storageClass
try:
formatter = self.formatterFactory.getFormatter(ref,
FileDescriptor(location,
storageClass=storageClass))
except KeyError as e:
raise DatasetTypeNotSupportedError(f"Unable to find formatter for {ref}") from e
return location, formatter
def _register_dataset_file(self, ref, formatter, path, size, checksum=None):
"""Update registry to indicate that this dataset has been stored,
specifying file metadata.
Parameters
----------
ref : `DatasetRef`
Dataset to register.
formatter : `Formatter`
Formatter to use to read this dataset.
path : `str`
Path to dataset relative to datastore root.
size : `int`
Size of the serialized dataset.
checksum : `str`, optional
Checksum of the serialized dataset. Can be `None`.
"""
# Associate this dataset with the formatter for later read.
fileInfo = StoredFileInfo(formatter, path, ref.datasetType.storageClass,
file_size=size, checksum=checksum)
self._register_dataset(ref, fileInfo)
def getUri(self, ref, predict=False):
"""URI to the Dataset.
Parameters
----------
ref : `DatasetRef`
Reference to the required Dataset.
predict : `bool`
If `True`, allow URIs to be returned of datasets that have not
been written.
Returns
-------
uri : `str`
URI string pointing to the Dataset within the datastore. If the
Dataset does not exist in the datastore, and if ``predict`` is
`True`, the URI will be a prediction and will include a URI
fragment "#predicted".
If the datastore does not have entities that relate well
to the concept of a URI the returned URI string will be
descriptive. The returned URI is not guaranteed to be obtainable.
Raises
------
FileNotFoundError
A URI has been requested for a dataset that does not exist and
guessing is not allowed.
"""
# if this has never been written then we have to guess
if not self.exists(ref):
if not predict:
raise FileNotFoundError("Dataset {} not in this datastore".format(ref))
template = self.templates.getTemplate(ref)
location = self.locationFactory.fromPath(template.format(ref) + "#predicted")
else:
# If this is a ref that we have written we can get the path.
# Get file metadata and internal metadata
storedFileInfo = self.getStoredItemInfo(ref)
# Use the path to determine the location
location = self.locationFactory.fromPath(storedFileInfo.path)
return location.uri
def validateConfiguration(self, entities, logFailures=False):
"""Validate some of the configuration for this datastore.
Parameters
----------
entities : iterable of `DatasetRef`, `DatasetType`, or `StorageClass`
Entities to test against this configuration. Can be differing
types.
logFailures : `bool`, optional
If `True`, output a log message for every validation error
detected.
Raises
------
DatastoreValidationError
Raised if there is a validation problem with a configuration.
All the problems are reported in a single exception.
Notes
-----
This method checks that all the supplied entities have valid file
templates and also have formatters defined.
"""
templateFailed = None
try:
self.templates.validateTemplates(entities, logFailures=logFailures)
except FileTemplateValidationError as e:
templateFailed = str(e)
formatterFailed = []
for entity in entities:
try:
self.formatterFactory.getFormatterClass(entity)
except KeyError as e:
formatterFailed.append(str(e))
if logFailures:
log.fatal("Formatter failure: %s", e)
if templateFailed or formatterFailed:
messages = []
if templateFailed:
messages.append(templateFailed)
if formatterFailed:
messages.append(",".join(formatterFailed))
msg = ";\n".join(messages)
raise DatastoreValidationError(msg)
def getLookupKeys(self):
# Docstring is inherited from base class
return self.templates.getLookupKeys() | self.formatterFactory.getLookupKeys() | \
self.constraints.getLookupKeys()
def validateKey(self, lookupKey, entity):
# Docstring is inherited from base class
# The key can be valid in either formatters or templates so we can
# only check the template if it exists
if lookupKey in self.templates:
try:
self.templates[lookupKey].validateTemplate(entity)
except FileTemplateValidationError as e:
raise DatastoreValidationError(e) from e