Commit 5d603cb4 authored by Roque's avatar Roque

erp5_wendelin: fix script query states and tests

See merge request nexedi/wendelin!92
parents 07c034b1 bc99f630
...@@ -10,7 +10,7 @@ if data_ingestion_line_list: ...@@ -10,7 +10,7 @@ if data_ingestion_line_list:
data_ingestion_uid_list = [x.uid for x in data_ingestion_line_list] data_ingestion_uid_list = [x.uid for x in data_ingestion_line_list]
catalog_kw = {'portal_type': 'Data Stream', catalog_kw = {'portal_type': 'Data Stream',
'aggregate__related__uid': data_ingestion_uid_list, 'aggregate__related__uid': data_ingestion_uid_list,
'validation_state':'validated', 'validation_state':['validated','published'],
'select_list': ['reference', 'relative_url', 'versioning.size', 'versioning.version'], 'select_list': ['reference', 'relative_url', 'versioning.size', 'versioning.version'],
} }
return context.getPortalObject().portal_catalog(**catalog_kw) return context.getPortalObject().portal_catalog(**catalog_kw)
......
...@@ -227,7 +227,7 @@ class TestDataIngestion(SecurityTestCase): ...@@ -227,7 +227,7 @@ class TestDataIngestion(SecurityTestCase):
self.tic() self.tic()
# check data relation between Data Set and Data Streams work # check data relation between Data Set and Data Streams work
self.assertSameSet(data_stream_list, data_set.DataSet_getDataStreamList()) self.assertSameSet(data_stream_list, [x.getObject() for x in data_set.DataSet_getDataStreamList()])
# check data set and all Data Streams states # check data set and all Data Streams states
self.assertEqual('validated', data_set.getValidationState()) self.assertEqual('validated', data_set.getValidationState())
...@@ -340,15 +340,18 @@ class TestDataIngestion(SecurityTestCase): ...@@ -340,15 +340,18 @@ class TestDataIngestion(SecurityTestCase):
""" """
# ingest a couple of files # ingest a couple of files
reference = self.getRandomReference() reference = self.getRandomReference()
self.ingest("some-data-1", reference, self.CSV, self.SINGLE_INGESTION_END) data_set_reference = "test-relation-dataset"
self.ingest("some-data-1", reference, self.CSV, self.SINGLE_INGESTION_END,
randomize_ingestion_reference=False, data_set_reference=data_set_reference)
time.sleep(1) time.sleep(1)
self.tic() self.tic()
reference += "-2" reference += "-2"
self.ingest("some-data-2", reference, self.CSV, self.SINGLE_INGESTION_END) self.ingest("some-data-2", reference, self.CSV, self.SINGLE_INGESTION_END,
randomize_ingestion_reference=False, data_set_reference=data_set_reference)
time.sleep(1) time.sleep(1)
self.tic() self.tic()
# get corresponding Data Streams by searching via Data Ingestion Lines of the Data Set # get corresponding Data Streams by searching via Data Ingestion Lines of the Data Set
data_set = self.portal.data_set_module.get(self.REF_DATASET) data_set = self.portal.data_set_module.get(data_set_reference)
data_ingestion_line_list = self.portal.portal_catalog( data_ingestion_line_list = self.portal.portal_catalog(
portal_type = 'Data Ingestion Line', portal_type = 'Data Ingestion Line',
aggregate_uid = data_set.getUid()) aggregate_uid = data_set.getUid())
...@@ -356,9 +359,10 @@ class TestDataIngestion(SecurityTestCase): ...@@ -356,9 +359,10 @@ class TestDataIngestion(SecurityTestCase):
data_stream_list = self.portal.portal_catalog( data_stream_list = self.portal.portal_catalog(
portal_type = 'Data Stream', portal_type = 'Data Stream',
aggregate__related__uid = data_ingestion_uid_list, aggregate__related__uid = data_ingestion_uid_list,
validation_state = ['validated','published'],
select_list = ['reference', 'relative_url', 'versioning.size', 'versioning.version']) select_list = ['reference', 'relative_url', 'versioning.size', 'versioning.version'])
data_stream_list = [x.getObject() for x in data_stream_list] data_stream_list = [x.getObject() for x in data_stream_list]
# assert that the list from the search is the same as DataSet_getDataStreamList # assert that the list from the search is the same as DataSet_getDataStreamList
self.assertSameSet(data_stream_list, data_set.DataSet_getDataStreamList()) self.assertSameSet(data_stream_list, [x.getObject() for x in data_set.DataSet_getDataStreamList()])
# XXX: new test which simulates download / upload of Data Set and increase DS version # XXX: new test which simulates download / upload of Data Set and increase DS version
\ No newline at end of file
""" """
Publish all Data Streams for a Data Set. Invalidate all Data Streams for a Data Set.
""" """
data_set = state_change['object'] data_set = state_change['object']
for data_stream in data_set.DataSet_getDataStreamList(): for data_stream in data_set.DataSet_getDataStreamList():
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment