A consequence of the last point is that a client should never specifically “catch” a contract failure (the ContractException type, in fact, is internal to help enforce that principle). Instead, the client should call the target properly; if it fails, this indicates a bug that should be handled via your general exception backstop (which may include terminating the application). In other words, if you decide control-flow or do other things based on a precondition failure, it’s not really a contract because you can continue executing if it fails.
This leads to the following advice when choosing between preconditions and throwing ordinary exceptions:
• If failure always indicates a bug in the client, favor a precondition.
• If failure indicates an abnormal condition, which may mean a bug in the client,
throw a (catchable) exception instead.引自第538页
表示其中内容是对原文的摘抄