• Replace all places that use old async node (@asyncio.coroutine and yield from) and use a new Python syntax (Only for Python v3.5+)
• Save the old code to allow users with Python v2+ to run our tests.
• requirements.txt:
  • Install futures package only for Python v2
  • Use the latest sex package version

Python3 ships with concurrent.futures so it shouldn't be pulled in but with latest concurrent.futures release on pypi they broke py3 compatibility:

https://github.com/agronholm/pythonfutures/issues/41

Superset of https://github.com/datastax/python-driver/pull/241

Updated to map native types to cql types.

The integration test will probably not be merged until we find out what versions of C* will have these types to test against.

I had some general difficulty testing against this branch due to some apparently unrelated server-side errors. https://github.com/josh-mckenzie/cassandra/compare/7523_squashed

This PR addresses the issue to build arm64 wheels. To address this issue, I updated travis.yml to include arm64.

Also there were two additional errors (Master Build Error) I encountered when I triggered master build w/o any changes.

1. On python: pypy2.7-6.0: Ref: https://travis-ci.com/github/tbbharaj/python-driver/jobs/515470491

Command "python setup.py egg_info" failed with error code 1 in /tmp/pip-install-9Ah7r6/lz4/
371The command "if [[ $TRAVIS_PYTHON_VERSION != pypy3.5 ]]; then pip install lz4; fi" failed and exited with 1 during .

2. On Python: 3.7 Ref: https://travis-ci.com/github/tbbharaj/python-driver/jobs/515470490

ERROR: InvocationError for command /home/travis/build/tbbharaj/python-driver/.tox/gevent_loop/bin/nosetests --verbosity=2 --no-path-adjustment /home/travis/build/tbbharaj/python-driver/tests/unit/io/test_geventreactor.py (exited with code -11 (SIGSEGV)) (exited with code -11)
939___________________________________ summary ____________________________________
940ERROR:   gevent_loop: commands failed
941The command "tox -e gevent_loop" exited with 1.

To fix above two errors, I made following changes:

1. Update setup tools in travis.yml to fix build error in python setup.py egg_info,
2. Update greenlet version in tox.ini to fix build error in tox -e gevent_loop

Successful builds Build

If there are any questions/comments - I'd be happy fix things as needed. I opened the following JIRA issue as well Issue

Thank you Tanveen

Description

In the concurrent environment, if the global lazy_connect_args is accessed by different threads or coroutine, it will cause exception: It is required to setup() cqlengine before executing queries.

Fix

set lazy_connect_args to be threading.local object.

According to Cassandra:

A Range is responsible for the tokens between (left, right]. **

However, TokenMap.get_replicas() appears to implement [left, right).

This pull request aims to resolve this disagreement.

** Source: https://github.com/apache/cassandra/blob/trunk/src/java/org/apache/cassandra/dht/Range.java

I need to write more documentation before we merge, but I wanted to get discussion going on API choices.

I chose to model the API here almost exactly off the Java driver implementation. The two relevant choices here:

• the predicate is immutable, and
• host-filtering takes place at policy-population time, not at query-planning time.

Does anyone have strong feelings either way about those? Changing either of these is totally doable, but would add more complexity and make the class harder to test.

This PR addresses a bug in NumpyProtocolHandler to properly handle empty strings:

https://datastax-oss.atlassian.net/browse/PYTHON-550

It doesn't support NULL string values however, which requires more invasive changes (e.g. making use of NumPy's missing values support).

Add comparison operator to allow Timers with the same end time to be pushed into heaps. It's poorly documented, but the warning can be found here: https://docs.python.org/2/library/heapq.html#priority-queue-implementation-notes "In the future with Python 3, tuple comparison breaks for (priority, task) pairs if the priorities are equal and the tasks do not have a default comparison order."

This PR adds accelerated versions of ProtocolHandler for message deserialization using Cython. By default the CythonProtocolHandler will be used if python-driver is compiled with Cython.

This work has been generously supported by SurveyMonkey.

This PR:

• Add full unit-tests of CQLengine's Ascii columns.
• Extend Text columns unit-tests for minimal and maximal length.
• Thoroughly cover edge-cases on empty / blank / None value strings handling, especially in relation to the required column parameter.
• Align Ascii column behaviour to Text, by adding min_length and max_length parameter support.

If log is somehow gone and file exception due to the race mention in PYTHON-1266 it will also inevitably fail for the same reason after the loop so we need to catch the exception there as well.

I have hit it in some Cassandra dtest:

Exception in thread asyncore_cassandra_driver_event_loop (most likely raised during interpreter shutdown):
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 801, in __bootstrap_inner
  File "/usr/lib/python2.7/threading.py", line 754, in run
  File "/dse/dse-db/bin/../lib/cassandra-driver-internal-only-3.25.0.zip/cassandra-driver-3.25.0/cassandra/io/asyncorereactor.py", line 262, in _run_loop
<type 'exceptions.AttributeError'>: 'NoneType' object has no attribute 'debug'

There are small typos in:

• cassandra/cqlengine/models.py
• tests/integration/advanced/init.py
• tests/integration/advanced/graph/test_graph_query.py
• tests/integration/advanced/test_spark.py
• tests/integration/cqlengine/connections/test_connection.py
• tests/integration/cqlengine/test_connections.py
• tests/integration/standard/test_metadata.py
• tests/integration/standard/test_metrics.py
• tests/integration/standard/test_prepared_statements.py
• tests/unit/test_policies.py
• tests/unit/test_resultset.py

Fixes:

• Should read updated rather than udpated.
• Should read segregation rather than segration.
• Should read reusable rather than resusable.
• Should read metadata rather than metatadata.
• Should read rethrow rather than retrhow.
• Should read multiple rather than multipltle.
• Should read graph rather than graoh.
• Should read geometric rather than geomteric.
• Should read default rather than defaul.
• Should read assignment rather than assignement.

Semi-automated pull request generated by https://github.com/timgates42/meticulous/blob/master/docs/NOTE.md

This commit introduced some changes around the handling of orphaned stream IDs. Unfortunately we had a utility class used by at least one of our integration tests which needed to be updated for some of the cases introduced by the improved stream ID handling.

PytestCollectionWarning: cannot collect test class 'TestCluster' because it has a __new__ constructor (from: tests/integration/standard/test_client_warnings.py)
  class TestCluster(object):