際際滷

際際滷Share a Scribd company logo

Branching by abstraction

P
Piergiorgio Niero

We branch code to develop new features without affecting production, and to minimize merge conflicts. Branching by abstraction avoids directly porting code between branches. Instead, feature flags or configuration options are used to selectively enable features, allowing previous implementations and tests to be preserved when features are added or removed. This approach keeps code cleaner and makes it easier to promote changes or revert them without risk of losing features.

1 of 13
Download to read offline
why we need to branch write new code without affecting old code build and test new code in staging environments without affecting production
why we branch by abstraction to avoid porting code from one branch to the other to minimise merge operations performed by our SCM
we branch by abstraction not to be changing our code but to configure our code
when do we branch by abstraction we have to develop a new feature we have to change an existing feature
how do we branch by abstraction feature flagged feature flagged NOT feature flagged
how do we test feature flags
how do we test feature flags
what happens when we release remove feature flag from command remove feature flag from model remove feature flag from the command test remove feature flag from the model test run the tests again to see that everything is still working ...and we LOST the previous implementation!
we branch by abstraction not to be changing our code but not to configure our code
a possible solution: NO FLAGS!
a possible solution: CONFIGURATION Robotlegs Parsley
Solution ProsCons Pros: focusing only on the needed behavior without touching the previous class rely on previous implementations and tests cleaner tests against cleaner classes only one place for feature flags easy to revert/promote code from/to release no feature loss Cons: not a silver bullet for everything for sure any suggestion?
we branch by abstraction not to be changing our code but to configure our code

More Related Content

Branching by abstraction

  • 1. why we need to branch write new code without affecting old code build and test new code in staging environments without affecting production
  • 2. why we branch by abstraction to avoid porting code from one branch to the other to minimise merge operations performed by our SCM
  • 3. we branch by abstraction not to be changing our code but to configure our code
  • 4. when do we branch by abstraction we have to develop a new feature we have to change an existing feature
  • 5. how do we branch by abstraction feature flagged feature flagged NOT feature flagged
  • 6. how do we test feature flags
  • 7. how do we test feature flags
  • 8. what happens when we release remove feature flag from command remove feature flag from model remove feature flag from the command test remove feature flag from the model test run the tests again to see that everything is still working ...and we LOST the previous implementation!
  • 9. we branch by abstraction not to be changing our code but not to configure our code
  • 10. a possible solution: NO FLAGS!
  • 11. a possible solution: CONFIGURATION Robotlegs Parsley
  • 12. Solution ProsCons Pros: focusing only on the needed behavior without touching the previous class rely on previous implementations and tests cleaner tests against cleaner classes only one place for feature flags easy to revert/promote code from/to release no feature loss Cons: not a silver bullet for everything for sure any suggestion?
  • 13. we branch by abstraction not to be changing our code but to configure our code