semaphore-utils🔗ℹ

A small library for working with semaphores.

This is useful when you two functions that share a semaphore and one of them needs to call the other – it lets you tell the inner one not to use the semaphore since the outer one is already doing so.

; Note: The following code is intended for simplicity of example. In real use it would be

; better to, e.g., not share mutable state between threads and, at a minimum, to expose

; separate versions of the get-* functions that do not allow passing a semaphore in. Also,

; handle the case where a user is not already in the hashes.

;

> (require racket/splicing semaphore-utils)

> (splicing-let ([sema (make-semaphore 1)]                  [name->dept-id  (make-hash (list (cons 'alice 1) (cons 'bill 1) (cons 'charlie 2)))]                  [dept-id->names (make-hash (list (cons 1 (set 'alice 'bob))  (cons 2 (set 'charlie))))])        (define (get-dept-id name [sema sema])       (call/sema sema (thunk (hash-ref name->dept-id name))))        (define (get-users dept-id [sema sema])       (call/sema sema (thunk (hash-ref dept-id->names dept-id))))        (define (add-user! name dept-id)       (call/sema sema                  (thunk                   (hash-set! name->dept-id name dept-id)                   (hash-set! dept-id->names                              dept-id                              ; (set-add (get-users dept-id) name))))) ; WRONG! This will deadlock! `sema` is already in use!                              ; Pass #f as the semaphore so that we don't deadlock                              (set-add (get-users dept-id #f) name))))))

;

; If the following functions were running in different threads, the call/sema code would ensure that the 'get-*' functions

; did not interleave with a call to 'add-user!' and thereby see inconsistent state

> (define alice-dept-id (get-dept-id 'alice))

> (get-users alice-dept-id)

#<set: alice bob>

> (add-user! 'evan alice-dept-id)

> (get-users alice-dept-id)

#<set: evan alice bob>