15.3 Wills and Executors
A will executor manages a collection of values and associated will procedures (a.k.a. finalizers). The will procedure for each value is ready to be executed when the value has been proven (by the garbage collector) to be unreachable, except through weak references (see Weak Boxes) or as the registrant for other will executors. A will is useful for triggering clean-up actions on data associated with an unreachable value, such as closing a port embedded in an object when the object is no longer used.
Calling the will-execute or will-try-execute procedure executes a will that is ready in the specified will executor. Wills are not executed automatically, because certain programs need control to avoid race conditions. However, a program can create a thread whose sole job is to execute wills for a particular executor.
If a value is registered with multiple wills (in one or multiple executors), the wills are readied in the reverse order of registration. Since readying a will procedure makes the value reachable again, the will must be executed and the value must be proven again unreachable through only weak references before another of the wills is readied or executed. However, wills for distinct unreachable values are readied at the same time, regardless of whether the values are reachable from each other.
A will executor’s register is held non-weakly until after the corresponding will procedure is executed. Thus, if the content value of a weak box (see Weak Boxes) is registered with a will executor, the weak box’s content is not changed to #f until all wills have been executed for the value and the value has been proven again reachable through only weak references.
(will-executor? v) → boolean? |
v : any/c |
(will-register executor v proc) → void? |
executor : will-executor? |
v : any/c |
proc : (any/c . -> . any) |
(will-execute executor) → any |
executor : will-executor? |
(will-try-execute executor) → any |
executor : any/c |