The Clojure source code gives the following definition of time
(defmacro time
"Evaluates expr and prints the time it took. Returns the value of
expr."
{:added "1.0"}
[expr]
`(let [start# (. System (nanoTime))
ret# ~expr]
(prn (str "Elapsed time: " (/ (double (- (. System (nanoTime)) start#)) 1000000.0) " msecs"))
ret#))
Why does this need to be a macro? What about it could not be done with a function?
Because function evaluates its arguments before it starts working with them. You can write time
as function and try to call it:
(defn my-time [expr]
(let [start (System/nanoTime)
ret expr]
(prn (str "Elapsed time: " (/ (double (- (System/nanoTime) start)) 1000000.0) " msecs"))
ret))
(my-time (Thread/sleep 1000))
"Elapsed time: 0.005199 msecs"
=> nil
Compare that with time
macro:
(time (Thread/sleep 1000))
"Elapsed time: 1000.1222 msecs"
=> nil
Function my-time
evaluated argument (Thread/sleep 1000)
, symbol expr
got value nil
and then body of that function happened- two (System/nanoTime)
were called shortly after each other, because value of expr
was already computed.
Macro time
didn't evaluate (Thread/sleep 1000)
, but expanded to:
(macroexpand `(time (Thread/sleep 1000)))
=>
(let*
[start__6136__auto__ (. java.lang.System (clojure.core/nanoTime)) ret__6137__auto__ (java.lang.Thread/sleep 1000)]
(clojure.core/prn
(clojure.core/str
"Elapsed time: "
(clojure.core//
(clojure.core/double (clojure.core/- (. java.lang.System (clojure.core/nanoTime)) start__6136__auto__))
1000000.0)
" msecs"))
ret__6137__auto__)
and functions were called in this order: (System/nanoTime)
, (Thread/sleep 1000)
and other (System/nanoTime)
, which returns correct time.