We plan to move to Python 3 in the near future. Where possible, new code should be Python 3-compatible. In particular:

  • favor format over % for formatting strings.
  • use the / on numbers in a Python 3-compatible way. In particular, if you want floor division (which is the behavior of / in Python 2), use // instead. If you want the result of integer division to be a float (e.g. 1 / 2 == 0.5), add from __future__ import division to the top of the imports and use /. For more information, see the official Python 3 porting docs.
  • use absolute_import, division, and unicode_literals in new test files.

Contributions will be evaluated by PEP8. We now strictly enforce compliance, via a linter run with Travis CI against all new pull requests. We do not enforce the default limits on line length, but have established a maximum length of 200 chars as a sanity check. You can conform to PEP8 by running autopep8 which can be installed via pip. pip install autopep8 && autopep8 --in-place -a --ignore E501

Another way to make sure that your code will pass compliance checks is to run flake8 from a commit hook:

flake8 --install-hook
git config flake8.strict true
git config flake8.ignore E501,F811,F812,F821,F822,F823,F831,F841,N8,C9,W503

We do not enforce import sorting, but if you choose to organize imports by some convention, use the isort tool (pip install isort).

Please use session, and not cursor when naming your connection variables, to match the style preferred by the DataStax Python Driver, which is how these tests connect to C*.

All version objects being passed from CCM are now LooseVersion objects, instead of strings. Those can still be safely compared to strings, so there is no need to do version < LooseVersion('3.10').


Please view the following list of tags, as well as an example test. While this full list of tags is available for use, there is no need to use every tag for a given test. The description and jira_ticket fields should be included, but most others should only be used when appropriate. The test plan will live in a comment block below the test method declaration.

Test nameName of the test
DescriptionBrief description of the test
@paramDescription of Parameter 1 (Usually used for helper methods)
@returnDescription of expected return value (Usually used for helper methods)
@expected_errorsWhat exceptions this test is expected to throw on normal behavior (should be caught and expected in the test)
@throwsWhat exceptions this test would throw upon failure (if expecting a specific regression)
@jira_ticketAssociated JIRA ticket identifier, including the project name (e.g. CASSANDRA-42, not 42).
@expected_resultBrief summary of what the expected results of this test are
@test_assumptionsTest requirements (auth, hints disabled , etc)
@note(future improvments, todo, etc)
@test_categoryWhat categories this test falls under (deprecated)
def test_example(self):
Demonstrates the expected syntax for a test plan.
@jira_ticket CASSANDRA-0000
@note Test should not be implemented, it is an example.


In some cases, we organize our test files by putting them in directories. If you do so, please export a module from that directory by placing an file in the directory with the test files. This makes the modules visible to our test infrastructure scripts that divide tests into buckets for CI.


Byteman Files

Any and all byteman (.btm) files should be saved in the cassandra-dtest/byteman/ directory.

Summary: Review Checklist

  • Correctness
    • Does the test pass? If not, is the failure expected?
    • If the test shells out to other processes,
      • does it validate the output to ensure there were no failures?
      • is it Windows-compatibile?
    • Does the test exercise a new feature? If so, is it tagged with @since to skip older versions?
  • Style and Python 3 Compatibility:
    • Does the code use .format() over % for format strings?
    • If there are new test files, do they use the requested imports for Python 3 compatibility?
    • Have the changes caused any style regressions? In particular, did Travis find any?
    • Are cassandra.cluster.Session objects named session (and not cursor)?
  • Documentation and Metadata:
    • Are new tests and test classes documented with docstrings?
    • Are changed tests' documentation updated?
    • Is Cassandra‘s desired behavior described in the documentation if it’s not immediately readable in the test?
    • Does the documentation include all appropriate Doxygen annotations, in particular @jira_ticket?
  • Readability and Reusability
    • Are any data structures built by looping that could be succinctly created in a comprehension
    • Is there repeated logic that could be factored out and given a descriptive name?
      • Does that repeated logic belong somewhere other than this particular test? Possible appropriate locations include the modules in tools/, or ccm.
    • If there is no assertion in the test, should there be? If not, is the statement that would fail under a regression commented to indicate that?
    • Is it possible for an uninitiated reader to understand what Casssandra behavior is being tested for?
      • If not, could the code be rewritten so it is?
      • If not, are there comments and documentation describing the desired behavior and how it's tested?