Thanks for the feedback. I will put together a proof of concept which we can
further evaluate/refine/merge upstream.
-Andy
On 7/13/17, 11:30 AM, "Kevin Doran" <kdoran.apa...@gmail.com> wrote:
Great idea, Andy! Additional types of automated tests would help the
minifi-cpp project significantly, and I think your proposal is an appropriate
way to add integration tests for the minifi agent. This sounds like a great way
to verify expected behavior of processors and the system of components in flow
combinations.
I like the idea of declarative tests that are interpreted / run by a
harness or framework as a way to allow others can contribute test cases.
I've never used the Bats framework before, but it seems like a reasonable
option for what you describe. It might require writing a fair amount of bash
code under-the-hood to get the functionality you want (helper functions and
such), but it looks like it would keep the test cases themselves and the output
clean and light. Perhaps others can offer suggestions here.
One comment, which you've probably already considered, is that we should
keep the dependencies (if any) that get added for integration tests that
leverage the docker target optional so they are not required for folks that
just want to build libminifi or the agent. It would be more of a
developer/contributor option but users could skip these tests.
/docker/test/integration seems like a reasonable place to add test cases.
Others would probably know better. I think the README.md would be a reasonable
place to document how to run the tests with a reference to another document
that describes how to add / contribute new test cases. I'm not sure where the
best location for the documentation should live.
Thanks,
Kevin
On 7/13/17, 10:34, "Andy Christianson" <achristian...@hortonworks.com>
wrote:
Yes, I envision having a directory of declarative test cases. Each
would include a flow yaml, one or more input files, and expected outputs.
I’d like to document the convention before writing the implementation
because if the conventions are solid, we can change out the actual test driver
implementation later on if needed.
Would it be best to document this in a section within /README.md, or
should I add a new file such as /docs/Testing.md, or /TESTING.md?
As for where the test cases would be added, I was thinking maybe
/docker/test/integration, keeping consistent with the existing convention (i.e.
/libminifi/test/integration).
-Andy
On 7/13/17, 10:14 AM, "Marc" <phroc...@apache.org> wrote:
Hi Andy,
I think this is a great idea to test integrating MiNiFi among
multiple
system components. Do you have a feel for how you will allow others
to
create test cases? Will you attempt to minimize the footprint of
contributed tests by creating a bats based framework? I ask because
it
would be cool if contributors could supply a flow ( input ) and
expected
output and we automatically run the necessary
containers/components. Is
this along the lines of your vision?
Thanks,
Marc
On Wed, Jul 12, 2017 at 12:26 PM, Andy Christianson <
achristian...@hortonworks.com> wrote:
> Hi All,
>
> I am looking at MINIFI-350 and would like to implement some
end-to-end
> integration tests for minifi cpp. Essentially, the tests would:
>
>
> 1. Stand up a new minifi cpp docker container
> 2. Send test data to HTTP input ports on the container
> 3. Run data through a minifi flow
> 4. Receive output data to a HTTP endpoint
> 5. Verify output data according to some constraints (headers
present,
> hash of the content, etc.)
>
> Most of this work, such as setting up a docker container and
sending data
> to it, can naturally be done with shell commands. As such, I’ve
taken a
> look at the bats [1] testing framework, which seems simple enough
and is
> very expressive.
>
> Any thoughts or suggestions on test frameworks to use are
appreciated.
>
> [1]: https://github.com/sstephenson/bats
>
>
