Layer Enforcer is a tool for linting your imports within the project.
The tool was designed to help maintain following order of things:
- Dependencies always flow towards root of the tree. Often this is
domainlayer. - Dependencies may skip layers, as long as they flow in same direction.
For instance, your
weblayer may import use cases fromdomainlayer. - Infer layer from imports or module names. Say, if you import
fastapianywhere on the chain, the module is assigned toweblayer, or maybe you wantmodelsto always bedblayer. - Flexible layer structure. Preferably defined via config file.
- Code is packaged by component. I.e. single component may contain code from the different layers.
Be warned, since this might not the best fit for you.
First pass: Match modules to layers. Report conflict if single module match more than one layer.
Second pass: For each module, iterate through all the imported modules. If current module has no assigned layer, assign first found layer within list of imported modules. If current module has an assigned layer, report conflict if import is not allowed.
pip install layer-enforcerpip install -e .[test,lint]layer-enforcer myproject myotherproject --layers layers.ymlInstead of explicitly specifying CLI arguments, it is possible to provide
configuration via pyproject.toml within the tool.layer_enforcer table:
[tool.layer_enforcer]
modules = ["myproject", "myotherproject"]
ignore = ["myproject.containers"]
layers = "layers.yml"An example of clean-architecture-ish layer layout for typical web app:
name: domain
submodules: ["entities", "use_cases"]
layers:
- name: service
submodules: ["services"]
layers:
- name: infrastructure
imports: ["stripe", "requests", "passlib"]
layers:
- name: db
imports: ["sqlalchemy", "psycopg2", "alembic"]
submodules: ["models", "memory", "database"]
- name: web
imports: ["fastapi", "jose"]
submodules: ["views", "schemas"]
- name: tasks
imports: ["celery"]
submodules: ["tasks"]