[ 
https://issues.apache.org/jira/browse/CONTINUUM-2756?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Brent N Atkinson updated CONTINUUM-2756:
----------------------------------------
    Description: 
Due to the unfortunate naming convention of the spring and plexus configuration 
files, spring-context.xml and components.xml respectively, it is not possible 
to load a subset of the configurations on the classpath. This makes is more 
difficult to limit the scope of the loaded context, which makes test setup and 
teardown much more expensive when not required.

Unfortunately, this convention is common for plexus and not limited to the 
Continuum code base. It may not be possible to do this with external 
components. In these cases, it may make sense to mock them or at the very least 
used cached application contexts to minimize setup/teardown time.

The ultimate issue is the time it takes to verify. It significantly reduces the 
productivity of a developer when making changes and the longer feedback loop 
has the usual effects (see the reasoning for CI). 

The state of the tests in the code base today is that there are very few unit 
tests and a limited number of integration tests. The difference in speed with 
tests disabled/enabled is 1 minute versus 11+ minutes on a fast system with an 
SSD drive. The code would benefit from refocusing on unit testing over 
integration tests, but this requires very necessary architectural improvements 
to be realistic.

  was:
Due to the unfortunate naming convention of the spring and plexus configuration 
files, spring-context.xml and components.xml respectively, it is not possible 
to load a subset of the configurations on the classpath. This makes is more 
difficult to limit the scope of the loaded context, which makes test setup and 
teardown much more expensive when not required.

Unfortunately, this convention is common for plexus and not limited to the 
Continuum code base. It may not be possible to do this with external 
components. In these cases, it may make sense to mock them or at the very least 
used cached application contexts to minimize setup/teardown time.

The ultimate issue is the time it takes to verify. It significantly reduces the 
productivity of a developer when making changes and the longer feedback loop 
has the usual effects (see the reasoning for CI). 

The state of the tests in the code base today is that there are very few unit 
tests and a limited number of integration tests. The difference in speed with 
tests disabled/enabled is 1 minute versus 11+ minutes on a fast system with an 
SSD drive. The code would benefit from refocusing on unit testing over 
integration tests, but this requires very necessary architectural improvements 
be realistic.


> Reduce the time necessary for verification
> ------------------------------------------
>
>                 Key: CONTINUUM-2756
>                 URL: https://issues.apache.org/jira/browse/CONTINUUM-2756
>             Project: Continuum
>          Issue Type: Task
>            Reporter: Brent N Atkinson
>
> Due to the unfortunate naming convention of the spring and plexus 
> configuration files, spring-context.xml and components.xml respectively, it 
> is not possible to load a subset of the configurations on the classpath. This 
> makes is more difficult to limit the scope of the loaded context, which makes 
> test setup and teardown much more expensive when not required.
> Unfortunately, this convention is common for plexus and not limited to the 
> Continuum code base. It may not be possible to do this with external 
> components. In these cases, it may make sense to mock them or at the very 
> least used cached application contexts to minimize setup/teardown time.
> The ultimate issue is the time it takes to verify. It significantly reduces 
> the productivity of a developer when making changes and the longer feedback 
> loop has the usual effects (see the reasoning for CI). 
> The state of the tests in the code base today is that there are very few unit 
> tests and a limited number of integration tests. The difference in speed with 
> tests disabled/enabled is 1 minute versus 11+ minutes on a fast system with 
> an SSD drive. The code would benefit from refocusing on unit testing over 
> integration tests, but this requires very necessary architectural 
> improvements to be realistic.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to