Normalize Job instances in tests

tests/client/test_items.py

@@ -1,6 +1,8 @@
 import pytest
 from six.moves import range
 
+from .utils import normalize_job_for_tests
+
 
 def _add_test_items(job, size=3):
     for i in range(size):
@@ -28,6 +30,7 @@ def test_items_iter(spider, json_and_msgpack):
 
 def test_items_list(spider, json_and_msgpack):
     job = spider.jobs.run(meta={'state': 'running'})
+    job = normalize_job_for_tests(job)
     _add_test_items(job)
 
     o = job.items.list()
@@ -40,6 +43,7 @@ def test_items_list(spider, json_and_msgpack):
 
 def test_items_list_iter(spider, json_and_msgpack):
     job = spider.jobs.run(meta={'state': 'running'})
+    job = normalize_job_for_tests(job)
     _add_test_items(job)
     job.finish()
 

tests/client/utils.py

@@ -13,3 +13,30 @@ def validate_default_meta(meta, state='pending', units=1,
     assert meta.get('units') == units
     assert meta.get('api_url') == TEST_DASH_ENDPOINT
     assert meta.get('portia_url')
+
+
+def normalize_job_for_tests(job):
+    """A temporary workaround to deal with VCR.py cassettes(snapshots).
+
+    The existing tests highly rely on VCR.py which creates snapshots of real
+    HTTP requests and responses, and during the test process tries to match
+    requests with the snapshots. Sometimes it's hard to run an appropriate test
+    environment locally, so we allow to use our servers to create snapshots
+    for new tests, by "normalizing" the snapshots via patching hosts/credentials
+    on-the-fly before saving it (see #112).
+
+    The problem here is that we patch only requests data and not responses data,
+    which is pretty difficult to unify over the whole client. It means that if
+    some test gets data from API (say, a new job ID) and uses it to form another
+    requests (get the job data), it will form the HTTP requests differently,
+    thus it won't match with the snapshots during the test process and the tests
+    will fail.
+
+    As a temporary workaround, the helper gets a Job instance, extracts its key,
+    replaces the project ID part with TEST_PROJECT_ID, and returns a new Job.
+    So, the other requests done via the new job instance (updating job items,
+    accessing job logs, etc) will be done using proper URLs matching with
+    existing snapshots.
+    """
+    normalized_key = '{}/{}'.format(TEST_PROJECT_ID, job.key.split('/', 1)[1])
+    return job._client.get_job(normalized_key)