def getRecord(self, metadataPrefix, identifier):
"""Return a (header, metadata, about) tuple for the the record.
metadataPrefix - identifies metadata set to retrieve the record in
identifier - repository-unique identifier of record
Should raise error.CannotDisseminateFormatError if metadataPrefix
is unknown or not supported by identifier.
Should raise error.IdDoesNotExistError if identifier is unknown or
illegal.
"""
session = self.session
if (
metadataPrefix and not
(metadataPrefix in self.protocolMap.recordNamespaces)
):
raise CannotDisseminateFormatError()
if not self.metadataRegistry.hasWriter(metadataPrefix):
# Need to create a 'MetadataWriter' for this schema for oaipmh to
# use, and put in self.metadataRegister
schemaId = self.protocolMap.recordNamespaces[metadataPrefix]
txr = self.protocolMap.transformerHash.get(schemaId, None)
mdw = Cheshire3OaiMetadataWriter(txr)
self.metadataRegistry.registerWriter(metadataPrefix, mdw)
q = cqlparse('rec.identifier exact "%s"' % (identifier))
try:
rs = self.db.search(session, q)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException('Index map for rec.identifier required '
'in protocolMap: %s'
'' % self.db.get_path(session,
'protocolMap').id
)
if not len(rs) or len(rs) > 1:
raise IdDoesNotExistError('%s records exist for this identifier'
'' % (len(rs)))
r = rs[0]
rec = r.fetch_record(session)
# Now reverse lookup lastModificationDate
q = cqlparse('rec.lastModificationDate < "%s"'
'' % (datetime.datetime.utcnow())
)
pm = self.db.get_path(session, 'protocolMap') # Get CQL ProtocolMap
idx = pm.resolveIndex(session, q)
vector = idx.fetch_vector(session, rec)
term = idx.fetch_termById(session, vector[2][0][0])
try:
datestamp = datetime.datetime.strptime(term, '%Y-%m-%dT%H:%M:%S')
except ValueError:
datestamp = datetime.datetime.strptime(term, '%Y-%m-%d %H:%M:%S')
# Handle non-ascii characters in identifier
identifier = unicode(r.id, 'utf-8')
identifier = identifier.encode('ascii', 'xmlcharrefreplace')
return (Header(identifier, datestamp, [], None), rec, None)
def testCombineSumWeights(self):
"Test combining ResultSet scores by summation."
# A clause / boolean is required to combine ResultSets
# Use TF-IDF because it's most simple to calculate
clause = cqlparse('my.index '
'all/rel.algorithm=tfidf/rel.combine=sum '
'"foo bar"')
clause.addPrefix('rel', "info:srw/cql-context-set/2/relevance-1.2")
# A Database is required for relevance ranking
db = FakeDatabase(self.session, None, parent=None)
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
rs = rs.combine(self.session, [self.a, self.b], clause, db)
# Check return value is a Resultset
self.assertIsInstance(rs, SimpleResultSet)
# Check merged ResultSet has 1 item
self.assertEqual(len(rs), 1)
# Check that merged ResultSet contains the correct item
self.assertIn(self.rsi1, rs)
for rsi in rs:
# Check that each ResultSetItem has a score (weight)
self.assertTrue(hasattr(rsi, 'weight'))
# Check that each ResultSetItem has a scaled score less than 1
self.assertLessEqual(rsi.scaledWeight, 1.0)
# Check combined scores correct
matches = len(self.b)
self.assertEqual(rs[0].weight,
sum([5 * math.log(db.totalItems / matches),
3 * math.log(db.totalItems / matches)
]
)
)
def listMetadataFormats(self, identifier=None):
"""Return a list of (metadataPrefix, schema, metadataNamespace) tuples (tuple items are strings).
identifier - identify record for which we want to know all
supported metadata formats. if absent, list all metadata
formats supported by repository. (optional)
Should raise error.IdDoesNotExistError if record with identifier does not exist.
Should raise error.NoMetadataFormatsError if no formats are available for the indicated record.
N.B.: Cheshire3 should supply same formats to all records in a database
"""
if identifier is not None:
q = cqlparse('rec.identifier exact "%s"' % (identifier))
try:
rs = self.db.search(session, q)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException(
'Index map for rec.identifier required in protocolMap: %s'
% self.db.get_path(session, 'protocolMap').id)
if not len(rs) or len(rs) > 1:
raise IdDoesNotExistError(
'%s records exist for identifier: %s' %
(len(rs), identifier))
# all records should be available in the same formats in a Cheshire3 database
mfs = []
for prefix, ns in self.protocolMap.recordNamespaces.iteritems():
mfs.append((prefix, self.protocolMap.schemaLocations[ns], ns))
if not len(mfs):
raise NoMetadataFormatsError()
return mfs
def listMetadataFormats(self, identifier=None):
"""Return a list of (metadataPrefix, schema, metadataNamespace) tuples (tuple items are strings).
identifier - identify record for which we want to know all
supported metadata formats. if absent, list all metadata
formats supported by repository. (optional)
Should raise error.IdDoesNotExistError if record with identifier does not exist.
Should raise error.NoMetadataFormatsError if no formats are available for the indicated record.
N.B.: Cheshire3 should supply same formats to all records in a database
"""
if identifier is not None:
q = cqlparse('rec.identifier exact "%s"' % (identifier))
try:
rs = self.db.search(session, q)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException('Index map for rec.identifier required in protocolMap: %s' % self.db.get_path(session, 'protocolMap').id)
if not len(rs) or len(rs) > 1:
raise IdDoesNotExistError('%s records exist for identifier: %s' % (len(rs), identifier))
# all records should be available in the same formats in a Cheshire3 database
mfs = []
for prefix, ns in self.protocolMap.recordNamespaces.iteritems():
mfs.append((prefix, self.protocolMap.schemaLocations[ns], ns))
if not len(mfs):
raise NoMetadataFormatsError()
return mfs
def __init__(self, dbName):
global configs, dbs, session
self.protocolMap = configs[dbName]
self.db = dbs[dbName]
session.database = self.db.id
# get some generally useful stuff now
self.baseURL = self.protocolMap.baseURL
# get earliest datestamp in database
q = cqlparse('rec.lastModificationDate > "%s"' % (str(datetime.datetime.utcfromtimestamp(0)))) # get UTC of the epoch as query term
try:
tl = self.db.scan(session, q, 1)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException('Index map for rec.lastModificationDate required in protocolMap: %s' % self.db.get_path(session, 'protocolMap').id)
else:
try:
datestamp = tl[0][0]
except IndexError:
#something went wrong :( - use the epoch
self.earliestDatestamp = datetime.datetime.utcfromtimestamp(0)
else:
try:
self.earliestDatestamp = datetime.datetime.strptime(datestamp, '%Y-%m-%dT%H:%M:%S')
except ValueError:
self.earliestDatestamp = datetime.datetime.strptime(datestamp, '%Y-%m-%d %H:%M:%S')
self.repositoryName = self.protocolMap.title
self.protocolVersion = self.protocolMap.version
self.adminEmails = self.protocolMap.contacts
self.deletedRecord = "no" # Cheshire3 does not support deletions at this time
self.granularity = "YYYY-MM-DDThh:mm:ssZ" # finest level of granularity
self.compression = [] # Cheshire3 does not support compressions at this time
self.metadataRegistry = OaiMetadataRegistry()
def testCori(self):
"Test combining with CORI relevance ranking."
# A clause / boolean is required to combine ResultSets
clause = cqlparse('my.index all/rel.algorithm=cori "foo bar"')
clause.addPrefix('rel', "info:srw/cql-context-set/2/relevance-1.2")
# A Database is required for relevance ranking
db = FakeDatabase(self.session, None)
# A RecordStore is required for CORI score calculation
recStore = FakeRecordStore(self.session, None)
# Test self.a
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
rs = rs.combine(self.session, [self.a], clause, db)
self.assertEqual(len(rs), 2)
for rsi in rs:
# Check that each ResultSetItem has a score (weight)
self.assertTrue(hasattr(rsi, 'weight'))
# Check that each ResultSetItem has a scaled score less than 1
self.assertLessEqual(rsi.scaledWeight, 1.0)
# Check scores are correct and in order
matches = len(self.a)
# I is used in calculating score for each item
I = (math.log((db.totalItems + 0.5) / matches) /
math.log(db.totalItems + 1.0))
expectedScores = []
for rsi in [self.rsi1, self.rsi3]:
size = recStore.fetch_recordMetadata(self.session,
rsi.id,
'wordCount')
T = (rsi.occurences /
(rsi.occurences + 50.0 + (( 150.0 * size) / db.meanWordCount))
)
expectedScores.append(0.4 + (0.6 * T * I))
self.assertListEqual([rsi.weight for rsi in rs], expectedScores)
# Test self.b
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
rs = rs.combine(self.session, [self.b], clause, db)
self.assertEqual(len(rs), 2)
for rsi in rs:
# Check that each ResultSetItem has a score (weight)
self.assertTrue(hasattr(rsi, 'weight'))
# Check that each ResultSetItem has a scaled score less than 1
self.assertLessEqual(rsi.scaledWeight, 1.0)
# Check scores are correct and in order
matches = len(self.b)
# I is used in calculating score for each item
I = (math.log((db.totalItems + 0.5) / matches) /
math.log(db.totalItems + 1.0))
expectedScores = []
for rsi in [self.rsi2, self.rsi4]:
size = recStore.fetch_recordMetadata(self.session,
rsi.id,
'wordCount')
T = (rsi.occurences /
(rsi.occurences + 50.0 + (( 150.0 * size) / db.meanWordCount))
)
expectedScores.append(0.4 + (0.6 * T * I))
self.assertListEqual([rsi.weight for rsi in rs], expectedScores)
def getRecord(self, metadataPrefix, identifier):
"""Return a (header, metadata, about) tuple for the the record.
metadataPrefix - identifies metadata set to retrieve the record in
identifier - repository-unique identifier of record
Should raise error.CannotDisseminateFormatError if metadataPrefix is unknown or not supported by identifier.
Should raise error.IdDoesNotExistError if identifier is unknown or illegal.
"""
if metadataPrefix and not (metadataPrefix
in self.protocolMap.recordNamespaces):
raise CannotDisseminateFormatError()
if not self.metadataRegistry.hasWriter(metadataPrefix):
# need to create a 'MetadataWriter' for this schema for oaipmh to use, and put in self.metadataRegister
schemaId = self.protocolMap.recordNamespaces[metadataPrefix]
txr = self.protocolMap.transformerHash.get(schemaId, None)
mdw = Cheshire3OaiMetadataWriter(txr)
self.metadataRegistry.registerWriter(metadataPrefix, mdw)
q = cqlparse('rec.identifier exact "%s"' % (identifier))
try:
rs = self.db.search(session, q)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException(
'Index map for rec.identifier required in protocolMap: %s' %
self.db.get_path(session, 'protocolMap').id)
if not len(rs) or len(rs) > 1:
raise IdDoesNotExistError('%s records exist for this identifier' %
(len(rs)))
r = rs[0]
rec = r.fetch_record(session)
# now reverse lookup lastModificationDate
q = cqlparse('rec.lastModificationDate < "%s"' %
(datetime.datetime.utcnow()))
pm = self.db.get_path(session, 'protocolMap') # get CQL ProtocolMap
idx = pm.resolveIndex(session, q)
vector = idx.fetch_vector(session, rec)
term = idx.fetch_termById(session, vector[2][0][0])
try:
datestamp = datetime.datetime.strptime(term, '%Y-%m-%dT%H:%M:%S')
except ValueError:
datestamp = datetime.datetime.strptime(term, '%Y-%m-%d %H:%M:%S')
return (Header(str(r.id), datestamp, [], None), rec, None)
def test_search(self):
"""Test a simple search of the Index."""
# Initialize IndexStore with some data
indexData = [0, None, None, 0, 0, 3, 1, 0, 2]
# Calculate total Records, total Occurences
totalRecs = len(indexData[3:]) / 3 # length of records part / 3
indexData[1] = totalRecs
self.assertTrue(indexData[1] is not None and indexData[1] >= 0,
"Incorrect definition of test data: totalRecs")
totalOccs = sum(indexData[3:][2::3]) # sum of record occurences
indexData[2] = totalOccs
self.assertTrue(indexData[2] is not None and indexData[2] >= 0,
"Incorrect definition of test data: totalRecs")
self.testObj.indexStore.indexes[self.testObj.id]['bar'] = indexData
# Parse a query
query = cqlparse('c3.foo = bar')
# Fetch a Database object
db = self.server.get_object(self.session, self.session.database)
# Carry out the search
rs = self.testObj.search(self.session, query, db)
# Check return value
self.assertIsInstance(rs, SimpleResultSet)
# Test ResultSet summary data
self.assertEqual(rs.totalRecs,
totalRecs,
"ResultSet.totalRecs not as expected: {0} != {1}"
"".format(rs.totalRecs, totalRecs)
)
# Test len(ResultSet)
self.assertEqual(len(rs),
totalRecs,
"ResultSet length not as expected ({0})"
"".format(totalRecs)
)
self.assertEqual(rs.termid,
indexData[0],
"ResultSet.termid not as expected: {0} != {1}"
"".format(rs.termid, indexData[0]))
self.assertEqual(rs.totalOccs,
totalOccs,
"ResultSet.totalOccs not as expected: {0} != {1}"
"".format(rs.totalOccs, totalOccs)
)
# Check items
for rsi in rs:
self.assertIsInstance(rsi, SimpleResultSetItem)
# Check identifiers
self.assertEqual(rs[0].id, 0)
self.assertEqual(rs[1].id, 1)
# Check occurences (sic)
self.assertEqual(rs[0].occurences, 3)
self.assertEqual(rs[1].occurences, 2)
# Check resultSet raise appropriate error when outside bounds
with self.assertRaises(IndexError):
rs[2]
def testCombineAll(self):
"Test combining ResultSets with 'all'"
# A clause / boolean is required to combine ResultSets
clause = cqlparse('my.index all "foo"')
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
rs = rs.combine(self.session, [self.a, self.b], clause)
# Check return value is a Resultset
self.assertIsInstance(rs, SimpleResultSet)
# Check merged ResultSet has 1 item
self.assertEqual(len(rs), 1)
# Check that merged ResultSet contains the correct item
self.assertIn(self.rsi1, rs)
def _listResults(self, metadataPrefix, set=None, from_=None, until=None):
"""Return a list of (datestamp, resultSet) tuples.
Suitable for use by:
- listIdentifiers
- listRecords
"""
session = self.session
if until and until < self.earliestDatestamp:
raise BadArgumentError('until argument value is earlier than '
'earliestDatestamp.')
if not from_:
from_ = self.earliestDatestamp
if not until:
until = datetime.datetime.now()
#(from_ < self.earliestDatestamp)
if (until < from_):
raise BadArgumentError('until argument value is earlier than from '
'argument value.')
q = cqlparse('rec.lastModificationDate > "%s" and '
'rec.lastModificationDate < "%s"' % (from_, until))
# Actually need datestamp values as well as results - interact with
# indexes directly for efficiency
# Get CQL ProtocolMap
pm = self.db.get_path(session, 'protocolMap')
idx = pm.resolveIndex(session, q.leftOperand)
q.config = pm
res = {}
for src in idx.sources[u'data']:
res.update(src[1].process(session, [[str(from_)]]))
res.update(src[1].process(session, [[str(until)]]))
from_ = min(res.keys())
until = max(res.keys())
# Tweak until value to make it inclusive
until = until[:-1] + chr(ord(until[-1])+1)
termList = idx.fetch_termList(session, from_, 0, '>=', end=until)
# Create list of datestamp, resultSet tuples
tuples = []
for t in termList:
try:
tuples.append((
datetime.datetime.strptime(t[0], u'%Y-%m-%dT%H:%M:%S'),
idx.construct_resultSet(session, t[1])
))
except ValueError:
tuples.append((
datetime.datetime.strptime(t[0], u'%Y-%m-%d %H:%M:%S'),
idx.construct_resultSet(session, t[1])
))
return tuples
def __init__(self, session, configs, dbs, dbName):
self.session = session
self.protocolMap = configs[dbName]
self.db = dbs[dbName]
session.database = self.db.id
# Get some generally useful stuff now
self.baseURL = self.protocolMap.baseURL
# Get earliest datestamp in database - UTC of the epoch as query term
q = cqlparse('rec.lastModificationDate > "%s"'
'' % (str(datetime.datetime.utcfromtimestamp(0))))
try:
tl = self.db.scan(session, q, 1)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException('Index map for '
'rec.lastModificationDate required '
'in protocolMap: %s'
'' % self.db.get_path(session,
'protocolMap').id
)
else:
try:
datestamp = tl[0][0]
except IndexError:
# Something went wrong :( - use the epoch
self.earliestDatestamp = datetime.datetime.utcfromtimestamp(0)
else:
try:
self.earliestDatestamp = datetime.datetime.strptime(
datestamp,
'%Y-%m-%dT%H:%M:%S'
)
except ValueError:
self.earliestDatestamp = datetime.datetime.strptime(
datestamp,
'%Y-%m-%d %H:%M:%S'
)
self.repositoryName = self.protocolMap.title
self.protocolVersion = self.protocolMap.version
self.adminEmails = self.protocolMap.contacts
# Check for deletion support
recordStore = self.db.get_path(session, 'recordStore')
deletions = recordStore.get_setting(session, 'storeDeletions')
# Cheshire3 cannot guarantee that deletions will persist
self.deletedRecord = "transient" if deletions else "no"
# Finest level of granularity
self.granularity = "YYYY-MM-DDThh:mm:ssZ"
# Cheshire3 does not support compressions at this time
self.compression = []
self.metadataRegistry = OaiMetadataRegistry()
def _listResults(self, metadataPrefix, set=None, from_=None, until=None):
"""Return a list of (datestamp, resultSet) tuples.
Suitable for use by:
- listIdentifiers
- listRecords
"""
session = self.session
if until and until < self.earliestDatestamp:
raise BadArgumentError('until argument value is earlier than '
'earliestDatestamp.')
if not from_:
from_ = self.earliestDatestamp
if not until:
until = datetime.datetime.now()
#(from_ < self.earliestDatestamp)
if (until < from_):
raise BadArgumentError('until argument value is earlier than from '
'argument value.')
q = cqlparse('rec.lastModificationDate > "%s" and '
'rec.lastModificationDate < "%s"' % (from_, until))
# Actually need datestamp values as well as results - interact with
# indexes directly for efficiency
# Get CQL ProtocolMap
pm = self.db.get_path(session, 'protocolMap')
idx = pm.resolveIndex(session, q.leftOperand)
q.config = pm
res = {}
for src in idx.sources[u'data']:
res.update(src[1].process(session, [[str(from_)]]))
res.update(src[1].process(session, [[str(until)]]))
from_ = min(res.keys())
until = max(res.keys())
# Tweak until value to make it inclusive
until = until[:-1] + chr(ord(until[-1]) + 1)
termList = idx.fetch_termList(session, from_, 0, '>=', end=until)
# Create list of datestamp, resultSet tuples
tuples = []
for t in termList:
try:
tuples.append(
(datetime.datetime.strptime(t[0], u'%Y-%m-%dT%H:%M:%S'),
idx.construct_resultSet(session, t[1])))
except ValueError:
tuples.append(
(datetime.datetime.strptime(t[0], u'%Y-%m-%d %H:%M:%S'),
idx.construct_resultSet(session, t[1])))
return tuples
def testCombineAny(self):
"Test combining ResultSets with 'any'"
# A clause / boolean is required to combine ResultSets
clause = cqlparse('my.index any "foo"')
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
rs = rs.combine(self.session, [self.a, self.b], clause)
# Check return value is a Resultset
self.assertIsInstance(rs, SimpleResultSet)
# Check merged ResultSet contains each ResultSetItem
self.assertIn(self.rsi1, rs)
self.assertIn(self.rsi2, rs)
self.assertIn(self.rsi3, rs)
self.assertIn(self.rsi4, rs)
# Check merged ResultSet has 3 items (as rsi1 and rsi2 are identical)
self.assertEqual(len(rs), 3)
def __init__(self, session, configs, dbs, dbName):
self.session = session
self.protocolMap = configs[dbName]
self.db = dbs[dbName]
session.database = self.db.id
# Get some generally useful stuff now
self.baseURL = self.protocolMap.baseURL
# Get earliest datestamp in database - UTC of the epoch as query term
q = cqlparse('rec.lastModificationDate > "%s"'
'' % (str(datetime.datetime.utcfromtimestamp(0))))
try:
tl = self.db.scan(session, q, 1)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException(
'Index map for '
'rec.lastModificationDate required '
'in protocolMap: %s'
'' % self.db.get_path(session, 'protocolMap').id)
else:
try:
datestamp = tl[0][0]
except IndexError:
# Something went wrong :( - use the epoch
self.earliestDatestamp = datetime.datetime.utcfromtimestamp(0)
else:
try:
self.earliestDatestamp = datetime.datetime.strptime(
datestamp, '%Y-%m-%dT%H:%M:%S')
except ValueError:
self.earliestDatestamp = datetime.datetime.strptime(
datestamp, '%Y-%m-%d %H:%M:%S')
self.repositoryName = self.protocolMap.title
self.protocolVersion = self.protocolMap.version
self.adminEmails = self.protocolMap.contacts
# Check for deletion support
recordStore = self.db.get_path(session, 'recordStore')
deletions = recordStore.get_setting(session, 'storeDeletions')
# Cheshire3 cannot guarantee that deletions will persist
self.deletedRecord = "transient" if deletions else "no"
# Finest level of granularity
self.granularity = "YYYY-MM-DDThh:mm:ssZ"
# Cheshire3 does not support compressions at this time
self.compression = []
self.metadataRegistry = OaiMetadataRegistry()
def __init__(self, dbName):
global configs, dbs, session
self.protocolMap = configs[dbName]
self.db = dbs[dbName]
session.database = self.db.id
# get some generally useful stuff now
self.baseURL = self.protocolMap.baseURL
# get earliest datestamp in database
q = cqlparse('rec.lastModificationDate > "%s"' %
(str(datetime.datetime.utcfromtimestamp(0)))
) # get UTC of the epoch as query term
try:
tl = self.db.scan(session, q, 1)
except SRWDiagnostics.Diagnostic16:
raise ConfigFileException(
'Index map for rec.lastModificationDate required in protocolMap: %s'
% self.db.get_path(session, 'protocolMap').id)
else:
try:
datestamp = tl[0][0]
except IndexError:
#something went wrong :( - use the epoch
self.earliestDatestamp = datetime.datetime.utcfromtimestamp(0)
else:
try:
self.earliestDatestamp = datetime.datetime.strptime(
datestamp, '%Y-%m-%dT%H:%M:%S')
except ValueError:
self.earliestDatestamp = datetime.datetime.strptime(
datestamp, '%Y-%m-%d %H:%M:%S')
self.repositoryName = self.protocolMap.title
self.protocolVersion = self.protocolMap.version
self.adminEmails = self.protocolMap.contacts
self.deletedRecord = "no" # Cheshire3 does not support deletions at this time
self.granularity = "YYYY-MM-DDThh:mm:ssZ" # finest level of granularity
self.compression = [
] # Cheshire3 does not support compressions at this time
self.metadataRegistry = OaiMetadataRegistry()
def testTfidf(self):
"Test combining with TF-IDF relevance ranking."
# A clause / boolean is required to combine ResultSets
clause = cqlparse('my.index all/rel.algorithm=tfidf "foo bar"')
clause.addPrefix('rel', "info:srw/cql-context-set/2/relevance-1.2")
# A Database is required for relevance ranking
db = FakeDatabase(self.session, None, parent=None)
# Test self.a
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
rs = rs.combine(self.session, [self.a], clause, db)
self.assertEqual(len(rs), 2)
for rsi in rs:
# Check that each ResultSetItem has a score (weight)
self.assertTrue(hasattr(rsi, 'weight'))
# Check that each ResultSetItem has a scaled score less than 1
self.assertLessEqual(rsi.scaledWeight, 1.0)
# Check scores are correct and in order
matches = len(self.a)
self.assertListEqual([rsi.weight for rsi in rs],
[5 * math.log(db.totalItems / matches),
1 * math.log(db.totalItems / matches)]
)
# Test self.b
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
rs = rs.combine(self.session, [self.b], clause, db)
self.assertEqual(len(rs), 2)
for rsi in rs:
# Check that each ResultSetItem has a score (weight)
self.assertTrue(hasattr(rsi, 'weight'))
# Check that each ResultSetItem has a scaled score less than 1
self.assertLessEqual(rsi.scaledWeight, 1.0)
# Check scores are correct and in order
matches = len(self.b)
self.assertListEqual([rsi.weight for rsi in rs],
[3 * math.log(db.totalItems / matches),
2 * math.log(db.totalItems / matches)]
)
def testOkapi(self):
"Test combining with OKAPI BM-25 relevance ranking."
# A clause / boolean is required to combine ResultSets
b, k1, k3 = [0.75, 1.5, 1.5]
clause = cqlparse('my.index all/rel.algorithm=okapi/'
'rel.const0={0}/'
'rel.const1={1}/'
'rel.const2={2}'
' "foo bar"'.format(b, k1, k3))
clause.addPrefix('rel', "info:srw/cql-context-set/2/relevance-1.2")
# A Database is required for relevance ranking
db = FakeDatabase(self.session, None)
# A RecordStore is required for CORI score calculation
recStore = FakeRecordStore(self.session, None)
# Test self.a
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
# Set ResultSet queryFrequency - required for OKAPI BM-25
self.a.queryFreq = 1
rs = rs.combine(self.session, [self.a], clause, db)
self.assertEqual(len(rs), 2)
for rsi in rs:
# Check that each ResultSetItem has a score (weight)
self.assertTrue(hasattr(rsi, 'weight'))
# self.assertTrue(rsi.weight)
# Check that each ResultSetItem has a scaled score less than 1
self.assertLessEqual(rsi.scaledWeight, 1.0)
# Check scores are correct and in order
matches = len(self.a)
idf = math.log(db.totalItems / matches)
qtw = ((k3 + 1) * 1) / (k3 + 1)
expectedScores = []
for rsi in [self.rsi1, self.rsi3]:
size = recStore.fetch_recordMetadata(self.session,
rsi.id,
'wordCount')
T = (((k1 + 1) * rsi.occurences) /
((k1 * ((1 - b) + b *
(size / db.meanWordCount)
)
) +
rsi.occurences)
)
expectedScores.append(idf * T * qtw)
self.assertListEqual([rsi.weight for rsi in rs], expectedScores)
# Test self.b
# Create a new ResultSet to combine into
rs = SimpleResultSet(self.session)
# Set ResultSet queryFrequency - required for OKAPI BM-25
self.b.queryFreq = 1
rs = rs.combine(self.session, [self.b], clause, db)
self.assertEqual(len(rs), 2)
for rsi in rs:
# Check that each ResultSetItem has a score (weight)
self.assertTrue(hasattr(rsi, 'weight'))
# self.assertTrue(rsi.weight)
# Check that each ResultSetItem has a scaled score less than 1
self.assertLessEqual(rsi.scaledWeight, 1.0)
# Check scores are correct and in order
matches = len(self.a)
idf = math.log(db.totalItems / matches)
qtw = ((k3 + 1) * 1) / (k3 + 1)
expectedScores = []
for rsi in [self.rsi2, self.rsi4]:
size = recStore.fetch_recordMetadata(self.session,
rsi.id,
'wordCount')
T = (((k1 + 1) * rsi.occurences) /
((k1 * ((1 - b) + b *
(size / db.meanWordCount)
)
) +
rsi.occurences)
)
expectedScores.append(idf * T * qtw)
self.assertListEqual([rsi.weight for rsi in rs], expectedScores)
def _listResults(self, metadataPrefix, set_=None, from_=None, until=None):
"""Return a list of (datestamp, resultSet) tuples.
Suitable for use by:
- listIdentifiers
- listRecords
"""
session = self.session
# Check set value
if set_ and not set_.startswith('contributor:'):
raise StopIteration
elif set_:
set_ = set_.split(':', 1)[-1]
if until and until < self.earliestDatestamp:
raise BadArgumentError('until argument value is earlier than '
'earliestDatestamp.')
if not from_:
from_ = self.earliestDatestamp
if not until:
until = datetime.datetime.now()
#(from_ < self.earliestDatestamp)
if (until < from_):
raise BadArgumentError('until argument value is earlier than from '
'argument value.')
q = cqlparse('rec.lastModificationDate > "%s" and '
'rec.lastModificationDate < "%s"' % (from_, until)
)
# Actually need datestamp values as well as results - interact with
# indexes directly for efficiency
# Get CQL ProtocolMap
pm = self.db.get_path(session, 'protocolMap')
idx = pm.resolveIndex(session, q.leftOperand)
q.config = pm
res = {}
for src in idx.sources[u'data']:
res.update(src[1].process(session, [[str(from_)]]))
res.update(src[1].process(session, [[str(until)]]))
from_ = min(res.keys())
until = max(res.keys())
# Tweak until value to make it inclusive
until = until[:-1] + chr(ord(until[-1]) + 1)
termList = idx.fetch_termList(session, from_, 0, '>=', end=until)
# Generate sequence of datestamp, resultSet tuples
for t in termList:
try:
datetime_obj = datetime.datetime.strptime(
t[0],
u'%Y-%m-%dT%H:%M:%S'
)
except ValueError:
datetime_obj = datetime.datetime.strptime(
t[0],
u'%Y-%m-%d %H:%M:%S'
)
datetime_rs = idx.construct_resultSet(session, t[1])
if not set_:
yield (datetime_obj, datetime_rs)
else:
# Filter by set
set_q = cqlparse('vdb.identifier = {0}'.format(set_))
set_rs = self.db.search(session, set_q)
full_rs = SimpleResultSet(session)
full_q = cqlparse('{0} and {1}'
''.format(q.toCQL(), set_q.toCQL()))
yield (datetime_obj, full_rs.combine(session,
[datetime_rs, set_rs],
full_q
)
)