Scope rules to domains and contexts¶

Real systems are not one flat vocabulary. A term that is canonical in one bounded context can be an alias in another. dddlint models this with three layers of scope.

The three layers¶

Domain and context rules are added to the global rules for matching files; they never remove a global rule. When a domain and a context disagree on an alias mapping, the context wins because it is applied last.

Scope a rule to a directory¶

# applies everywhere
forbidden: [util, helper, manager]

domains:
  - name: commerce
    include: ["**/commerce/**"]
    synonyms:
      - canonical: order
        aliases: [purchase]

Now PurchaseService under src/commerce/ is flagged as an alias of order, while the same name elsewhere is left alone.

Let a context override a domain¶

domains:
  - name: commerce
    include: ["**/commerce/**"]
    synonyms:
      - canonical: order
        aliases: [invoice]

contexts:
  - name: billing
    include: ["**/commerce/billing/**"]
    synonyms:
      - canonical: invoice
        aliases: [bill, statement]

Under commerce/, invoice redirects to order. But under commerce/billing/, the context makes invoice canonical and redirects bill and statement to it instead. Context beats domain.

Verify your scoping in Python¶

You can assert this behaviour directly against the checker:

from pathlib import Path

from dddlint.check import check
from dddlint.config import Config, Context
from dddlint.extract import Definition

config = Config(
    contexts=[Context(name="billing", include=["**/billing/**"], forbidden=["discount"])]
)


def rules_for(name: str, path: str) -> set[str]:
    defs = [Definition(name, "Class", Path(path), 1)]
    return {f.rule for f in check(defs, config)}


# 'discount' is banned inside the billing context only
assert "forbidden" in rules_for("DiscountPolicy", "src/billing/rules.py")
assert "forbidden" not in rules_for("DiscountPolicy", "src/catalog/rules.py")

See the configuration reference for every key, and how checking works for the precedence model in depth.