You basically have two choices as step writer:

• CASE 1: The steps are bundled with your feature files in the steps directory
• CASE 2: The steps are provided in a Python package and are used in a step-library with some glue code in the steps directory to import the needed/used steps from the Python package

If you want to reuse your steps in multiple test projects, you basically have CASE 2 (which is the focus point in this issue).
But this requires that your step library is generic enough to be used in multiple test contexts.

GLUE CODE:

• The glue code is currently required (as user control point).
• It allows the user to select which steps of a Python package are used and which are not used.
• It avoids the ambiguous-step problem that may occur.

SEE ALSO:

• features/step.use_step_library.feature
• EXAMPLE: features/steps/use_steplib_behave4cmd.py

OTHER ASPECTS:

• You normally do not have a Python package that just contain steps.
• The steps often need some test infrastructure code (in other parts of this Python package) that makes writing the steps simpler and not overly complex.
• The test infrastructure parts help the to interact with the system that should be tested or perform tasks to simplify the steps.
• Therefore, the test infrastructure parts should be tested (as well as the steps) as normal for any Python package parts.