Chaining Decorators

Decorator becomes especially useful when multiple wrappers can be stacked. It also becomes easier to misuse.

    flowchart LR
	    A["Client"] --> B["Caching decorator"]
	    B --> C["Metrics decorator"]
	    C --> D["Authorization decorator"]
	    D --> E["Base service"]

Order Changes Meaning

The order of decorators is not cosmetic.

If authorization wraps metrics, denied requests may not be counted the same way. If caching wraps authorization, cached results may be reachable through the wrong path. If retry wraps logging, the logs may show attempts differently than expected.

That means the chain is part of the design, not an incidental assembly detail.

Keep Each Wrapper Honest

Each decorator should answer one question clearly:

• What concern does it add?
• Does it act before delegation, after delegation, or both?
• What happens on failure?

If those answers are fuzzy, chain behavior becomes hard to predict.

Test The Chain, Not Just The Pieces

Unit tests for individual decorators are useful, but chain-level tests matter too because order is often where the bugs appear.