remove use of private attributes

Author: tswast

google/cloud/bigquery/_job_helpers.py

@@ -147,12 +147,16 @@ def do_query():
                 # have pretty high confidence that by retrying the 404, we'll
                 # (hopefully) eventually recover the job.
                 # https://github.com/googleapis/python-bigquery/issues/2134
+                #
+                # Allow users who want to completely disable retries to
+                # continue to do so by setting retry to None.
                 get_job_retry = retry
                 if retry is not None:
-                    get_job_retry = retry.with_predicate(
-                        lambda exc: isinstance(exc, core_exceptions.NotFound)
-                        # Reference the original retry to avoid recursion.
-                        or retry._predicate(exc)
+                    # TODO(tswast): Amend the user's retry object with allowing
+                    # 404 to retry when there's a public way to do so.
+                    # https://github.com/googleapis/python-api-core/issues/796
+                    get_job_retry = (
+                        google.cloud.bigquery.retry._DEFAULT_GET_JOB_CONFLICT_RETRY
                     )
 
                 query_job = client.get_job(
@@ -169,28 +173,10 @@ def do_query():
         else:
             return query_job
 
+    # Allow users who want to completely disable retries to
+    # continue to do so by setting job_retry to None.
     if job_retry is not None:
-
-        def do_query_predicate(exc) -> bool:
-            if isinstance(exc, core_exceptions.RetryError):
-                exc = exc.cause
-
-            # Per https://github.com/googleapis/python-bigquery/issues/2134, sometimes
-            # we get a 404 error. In this case, if we get this far, assume that the job
-            # doesn't actually exist and try again. We can't add 404 to the default
-            # job_retry because that happens for errors like "this table does not
-            # exist", which probably won't resolve with a retry.
-            if isinstance(exc, core_exceptions.NotFound):
-                message = exc.message
-                # Don't try to retry table/dataset not found, just job not found.
-                # The URL contains jobs, so use whitespace to disambiguate.
-                return message is not None and " job" in message.lower()
-
-            # Reference the original job_retry to avoid recursion.
-            return job_retry._predicate(exc)
-
-        do_query_retry = job_retry.with_predicate(do_query_predicate)
-        do_query = do_query_retry(do_query)
+        do_query = google.cloud.bigquery.retry._DEFAULT_QUERY_JOB_INSERT_RETRY(do_query)
 
     future = do_query()
 

google/cloud/bigquery/retry.py

@@ -82,6 +82,28 @@ def _should_retry(exc):
 pass ``retry=bigquery.DEFAULT_RETRY.with_deadline(30)``.
 """
 
+
+def _should_retry_get_job_conflict(exc):
+    """Predicate for determining when to retry a jobs.get call after a conflict error.
+
+    Sometimes we get a 404 after a Conflict. In this case, we
+    have pretty high confidence that by retrying the 404, we'll
+    (hopefully) eventually recover the job.
+    https://github.com/googleapis/python-bigquery/issues/2134
+
+    Note: we may be able to extend this to user-specified predicates
+    after https://github.com/googleapis/python-api-core/issues/796
+    to tweak existing Retry object predicates.
+    """
+    return isinstance(exc, exceptions.NotFound) or _should_retry(exc)
+
+
+_DEFAULT_GET_JOB_CONFLICT_RETRY = retry.Retry(
+    predicate=_should_retry_get_job_conflict, deadline=_DEFAULT_RETRY_DEADLINE
+)
+"""Private, may be removed in future."""
+
+
 # Note: Take care when updating DEFAULT_TIMEOUT to anything but None. We
 # briefly had a default timeout, but even setting it at more than twice the
 # theoretical server-side default timeout of 2 minutes was not enough for
@@ -142,6 +164,35 @@ def _job_should_retry(exc):
 The default job retry object.
 """
 
+
+def _query_job_insert_should_retry(exc):
+    # Per https://github.com/googleapis/python-bigquery/issues/2134, sometimes
+    # we get a 404 error. In this case, if we get this far, assume that the job
+    # doesn't actually exist and try again. We can't add 404 to the default
+    # job_retry because that happens for errors like "this table does not
+    # exist", which probably won't resolve with a retry.
+
+    if isinstance(exc, exceptions.RetryError):
+        exc = exc.cause
+
+    if isinstance(exc, exceptions.NotFound):
+        message = exc.message
+        # Don't try to retry table/dataset not found, just job not found.
+        # The URL contains jobs, so use whitespace to disambiguate.
+        return message is not None and " job" in message.lower()
+
+    return _job_should_retry(exc)
+
+
+_DEFAULT_QUERY_JOB_INSERT_RETRY = retry.Retry(
+    predicate=_query_job_insert_should_retry,
+    # jobs.insert doesn't wait for the job to complete, so we don't need the
+    # long _DEFAULT_JOB_DEADLINE for this part.
+    deadline=_DEFAULT_RETRY_DEADLINE,
+)
+"""Private, may be removed in future."""
+
+
 DEFAULT_GET_JOB_TIMEOUT = 128
 """
 Default timeout for Client.get_job().