def test_index_not_dropped_by_init(self):
with self.admin_access.repo_cnx() as cnx:
store = MassiveObjectStore(cnx) # noqa
cnx.commit()
indexes = all_indexes(cnx)
self.assertIn('entities_pkey', indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from', 'eid_to'], 'key_'),
indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from'], 'idx_'),
indexes)
def test_consider_metagen(self):
"""Ensure index on owned_by is not deleted if we don't consider this metadata."""
with self.admin_access.repo_cnx() as cnx:
metagen = stores.MetadataGenerator(cnx, meta_skipped=('owned_by',))
store = MassiveObjectStore(cnx, metagen=metagen)
store._drop_metadata_constraints()
indexes = all_indexes(cnx)
self.assertIn(build_index_name('owned_by_relation', ['eid_from', 'eid_to'], 'key_'),
indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from'], 'idx_'),
indexes)
def expected_indexes(cnx):
"""Return a dictionary describing indexes expected by the schema {index name: (table, column)}.
This doesn't include primary key indexes.
"""
source = cnx.repo.system_source
dbh = source.dbhelper
schema = cnx.repo.schema
schema_indexes = SYSTEM_INDEXES.copy()
if source.dbdriver == 'postgres':
schema_indexes.update({'appears_words_idx': ('appears', 'words')})
else:
schema_indexes.update({'appears_uid': ('appears', 'uid'),
'appears_word_id': ('appears', 'word_id')})
for rschema in schema.relations():
if rschema.rule or rschema in PURE_VIRTUAL_RTYPES:
continue # computed relation
if rschema.final or rschema.inlined:
for rdef in rschema.rdefs.values():
table = 'cw_{0}'.format(rdef.subject)
column = 'cw_{0}'.format(rdef.rtype)
if any(isinstance(cstr, UniqueConstraint) for cstr in rdef.constraints):
schema_indexes[dbh._index_name(table, column, unique=True)] = (
table, [column])
if rschema.inlined or rdef.indexed:
schema_indexes[dbh._index_name(table, column)] = (table, [column])
else:
table = '{0}_relation'.format(rschema)
if source.dbdriver == 'postgres':
# index built after the primary key constraint
schema_indexes[build_index_name(table, ['eid_from', 'eid_to'], 'key_')] = (
table, ['eid_from', 'eid_to'])
schema_indexes[build_index_name(table, ['eid_from'], 'idx_')] = (
table, ['eid_from'])
schema_indexes[build_index_name(table, ['eid_to'], 'idx_')] = (
table, ['eid_to'])
for eschema in schema.entities():
if eschema.final:
continue
table = 'cw_{0}'.format(eschema)
for columns, index_name in iter_unique_index_names(eschema):
schema_indexes[index_name] = (table, columns)
return schema_indexes
def test_drop_index_recreation(self):
with self.admin_access.repo_cnx() as cnx:
store = MassiveObjectStore(cnx)
store._drop_metadata_constraints()
indexes = all_indexes(cnx)
self.assertIn('entities_pkey', indexes)
self.assertNotIn(build_index_name('owned_by_relation', ['eid_from', 'eid_to'], 'key_'),
indexes)
self.assertNotIn(build_index_name('owned_by_relation', ['eid_from'], 'idx_'),
indexes)
store.finish()
indexes = all_indexes(cnx)
self.assertIn('entities_pkey', indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from', 'eid_to'], 'key_'),
indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from'], 'idx_'),
indexes)
def test_index_building(self):
with self.admin_access.repo_cnx() as cnx:
store = MassiveObjectStore(cnx)
self.push_geonames_data(self.datapath('geonames.csv'), store)
store.flush()
# Check index
indexes = all_indexes(cnx)
self.assertIn('entities_pkey', indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from', 'eid_to'], 'key_'),
indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from'], 'idx_'),
indexes)
# Cleanup -> index
store.finish()
# Check index again
indexes = all_indexes(cnx)
self.assertIn('entities_pkey', indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from', 'eid_to'], 'key_'),
indexes)
self.assertIn(build_index_name('owned_by_relation', ['eid_from'], 'idx_'),
indexes)
for rschema in schema.relations():
if rschema.rule or rschema in PURE_VIRTUAL_RTYPES:
continue
if rschema.final or rschema.inlined:
for rdef in rschema.rdefs.values():
table = 'cw_{0}'.format(rdef.subject)
column = 'cw_{0}'.format(rdef.rtype)
if rschema.inlined or rdef.indexed:
old_name = '%s_%s_idx' % (table.lower(), column.lower())
sql('DROP INDEX IF EXISTS %s' % old_name)
source.create_index(cnx, table, column)
else:
table = '{0}_relation'.format(rschema)
sql('ALTER TABLE %s ADD CONSTRAINT %s PRIMARY KEY(eid_from, eid_to)'
% (table, build_index_name(table, ['eid_from', 'eid_to'], 'key_')))
for column in ('from', 'to'):
sql('DROP INDEX IF EXISTS %s_%s_idx' % (table, column))
sql('CREATE INDEX %s ON %s(eid_%s);'
% (build_index_name(table, ['eid_' + column], 'idx_'), table, column))
# we changed constraint serialization, which also changes their name
for table, cstr in sql("""
SELECT DISTINCT table_name, constraint_name FROM information_schema.constraint_column_usage
WHERE constraint_name LIKE 'cstr%'"""):
sql("ALTER TABLE %(table)s DROP CONSTRAINT %(cstr)s" % locals())
for cwconstraint in rql('Any C WHERE R constrained_by C').entities():
cwrdef = cwconstraint.reverse_constrained_by[0]
rdef = cwrdef.yams_schema()
for rschema in schema.relations():
if rschema.rule or rschema in PURE_VIRTUAL_RTYPES:
continue
if rschema.final or rschema.inlined:
for rdef in rschema.rdefs.values():
table = 'cw_{0}'.format(rdef.subject)
column = 'cw_{0}'.format(rdef.rtype)
if rschema.inlined or rdef.indexed:
old_name = '%s_%s_idx' % (table.lower(), column.lower())
sql('DROP INDEX IF EXISTS %s' % old_name)
source.create_index(cnx, table, column)
else:
table = '{0}_relation'.format(rschema)
sql('ALTER TABLE %s ADD CONSTRAINT %s PRIMARY KEY(eid_from, eid_to)' %
(table, build_index_name(table, ['eid_from', 'eid_to'], 'key_')))
for column in ('from', 'to'):
sql('DROP INDEX IF EXISTS %s_%s_idx' % (table, column))
sql('CREATE INDEX %s ON %s(eid_%s);' % (build_index_name(
table, ['eid_' + column], 'idx_'), table, column))
# we changed constraint serialization, which also changes their name
for table, cstr in sql("""
SELECT DISTINCT table_name, constraint_name FROM information_schema.constraint_column_usage
WHERE constraint_name LIKE 'cstr%'"""):
sql("ALTER TABLE %(table)s DROP CONSTRAINT %(cstr)s" % locals())
for cwconstraint in rql('Any C WHERE R constrained_by C').entities():
cwrdef = cwconstraint.reverse_constrained_by[0]
rdef = cwrdef.yams_schema()