Active object
You are encouraged to solve this task according to the task description, using any language you may know.
In object-oriented programming an object is active when its state depends on clock. Usually an active object encapsulates a task that updates the object's state. To the outer world the object looks like a normal object with methods that can be called from outside. Implementation of such methods must have a certain synchronization mechanism with the encapsulated task in order to prevent object's state corruption.
A typical instance of an active object is an animation widget. The widget state changes with the time, while as an object it has all properties of a normal widget.
The task
Implement an active integrator object. The object has an input and output. The input can be set using the method Input. The input is a function of time. The output can be queried using the method Output. The object integrates its input over the time and the result becomes the object's output. So if the input is K(t) and the output is S, the object state S is changed to S + (K(t1) + K(t0)) * (t1 - t0) / 2, i.e. it integrates K using the trapeze method. Initially K is constant 0 and S is 0.
In order to test the object:
- set its input to sin (2π f t), where the frequency f=0.5Hz. The phase is irrelevant.
- wait 2s
- set the input to constant 0
- wait 0.5s
Verify that now the object's output is approximately 0 (the sine has the period of 2s). The accuracy of the result will depend on the OS scheduler time slicing and the accuracy of the clock.
Ada
with Ada.Calendar; use Ada.Calendar;
with Ada.Numerics; use Ada.Numerics;
with Ada.Numerics.Elementary_Functions; use Ada.Numerics.Elementary_Functions;
with Ada.Text_IO; use Ada.Text_IO;
procedure Test_Integrator is
type Func is access function (T : Time) return Float;
function Zero (T : Time) return Float is
begin
return 0.0;
end Zero;
Epoch : constant Time := Clock;
function Sine (T : Time) return Float is
begin
return Sin (Pi * Float (T - Epoch));
end Sine;
task type Integrator is
entry Input (Value : Func);
entry Output (Value : out Float);
entry Shut_Down;
end Integrator;
task body Integrator is
K : Func := Zero'Access;
S : Float := 0.0;
F0 : Float := 0.0;
F1 : Float;
T0 : Time := Clock;
T1 : Time;
begin
loop
select
accept Input (Value : Func) do
K := Value;
end Input;
or accept Output (Value : out Float) do
Value := S;
end Output;
or accept Shut_Down;
exit;
else
T1 := Clock;
F1 := K (T1);
S := S + 0.5 * (F1 + F0) * Float (T1 - T0);
T0 := T1;
F0 := F1;
end select;
end loop;
end Integrator;
I : Integrator;
S : Float;
begin
I.Input (Sine'Access);
delay 2.0;
I.Input (Zero'Access);
delay 0.5;
I.Output (S);
Put_Line ("Integrated" & Float'Image (S) & "s");
I.Shut_Down;
end Test_Integrator;
Sample output:
Integrated-5.34100E-05s
ATS
No memory management is needed. Everything is done in registers or on the stack, including the closure that is used to connect signal source and integrator.
The core code is templates for any floating point type, but, in the end, one must choose a type. Our choice is C float type.
(*------------------------------------------------------------------*)
(* I will not bother with threads. All we need is the ability to get
the time from the operating system. This is available as
clock(3). *)
#define ATS_PACKNAME "rosettacode.activeobject"
#define ATS_EXTERN_PREFIX "rosettacode_activeobject_"
#include "share/atspre_staload.hats"
(*------------------------------------------------------------------*)
(* Some math functionality, for all the standard floating point
types. The ats2-xprelude package includes this, and more, but one
may wish to avoid the dependency. And there is support for math
functions in libats/libc, but not with typekinds. *)
%{^
#include <math.h>
// sinpi(3) would be better than sin(3), but I do not yet have
// sinpi(3).
#define rosettacode_activeobject_pi \
3.14159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798214L
#define rosettacode_activeobject_sinpi_float(x) \
(sinf (((atstype_float) rosettacode_activeobject_pi) * (x)))
#define rosettacode_activeobject_sinpi_double \
(sin (((atstype_double) rosettacode_activeobject_pi) * (x)))
#define rosettacode_activeobject_sinpi_ldouble \
(sinl (((atstype_ldouble) rosettacode_activeobject_pi) * (x)))
%}
extern fn sinpi_float : float -<> float = "mac#%"
extern fn sinpi_double : double -<> double = "mac#%"
extern fn sinpi_ldouble : ldouble -<> ldouble = "mac#%"
extern fn {tk : tkind} g0float_sinpi : g0float tk -<> g0float tk
implement g0float_sinpi<fltknd> x = sinpi_float x
implement g0float_sinpi<dblknd> x = sinpi_double x
implement g0float_sinpi<ldblknd> x = sinpi_ldouble x
overload sinpi with g0float_sinpi
(*------------------------------------------------------------------*)
(* Some clock(3) functionality for the three standard floating point
types. *)
%{^
#include <time.h>
typedef clock_t rosettacode_activeobject_clock_t;
ATSinline() rosettacode_activeobject_clock_t
rosettacode_activeobject_clock () // C23 drops the need for "void".
{
return clock ();
}
ATSinline() rosettacode_activeobject_clock_t
rosettacode_activeobject_clock_difference
(rosettacode_activeobject_clock_t t,
rosettacode_activeobject_clock_t t0)
{
return (t - t0);
}
ATSinline() atstype_float
rosettacode_activeobject_clock_scaled2float
(rosettacode_activeobject_clock_t t)
{
return ((atstype_float) t) / CLOCKS_PER_SEC;
}
ATSinline() atstype_double
rosettacode_activeobject_clock_scaled2double
(rosettacode_activeobject_clock_t t)
{
return ((atstype_double) t) / CLOCKS_PER_SEC;
}
ATSinline() atstype_ldouble
rosettacode_activeobject_clock_scaled2ldouble
(rosettacode_activeobject_clock_t t)
{
return ((atstype_ldouble) t) / CLOCKS_PER_SEC;
}
%}
typedef clock_t = $extype"clock_t"
extern fn clock : () -<> clock_t = "mac#%"
extern fn clock_difference : (clock_t, clock_t) -<> clock_t = "mac#%"
extern fn clock_scaled2float : clock_t -<> float = "mac#%"
extern fn clock_scaled2double : clock_t -<> double = "mac#%"
extern fn clock_scaled2ldouble : clock_t -<> ldouble = "mac#%"
extern fn {tk : tkind} clock_scaled2g0float : clock_t -<> g0float tk
implement clock_scaled2g0float<fltknd> t = clock_scaled2float t
implement clock_scaled2g0float<dblknd> t = clock_scaled2double t
implement clock_scaled2g0float<ldblknd> t = clock_scaled2ldouble t
overload - with clock_difference
overload clock2f with clock_scaled2g0float
(*------------------------------------------------------------------*)
%{^
#if defined __GNUC__ && (defined __i386__ || defined __x86_64__)
// A small, machine-dependent pause, for improved performance of spin
// loops.
#define rosettacode_activeobject_pause() __builtin_ia32_pause ()
#else
// Failure to insert a small, machine-dependent pause may overwork
// your hardware, but the task can be done anyway.
#define rosettacode_activeobject_pause() do{}while(0)
#endif
%}
extern fn pause : () -<> void = "mac#%"
(*------------------------------------------------------------------*)
(* Types such as this can have their internals hidden, but here I will
not bother with such details. *)
vtypedef sinusoidal_generator (tk : tkind) =
@{
phase = g0float tk,
afreq = g0float tk, (* angular frequency IN UNITS OF 2*pi. *)
clock0 = clock_t,
stopped = bool
}
fn {tk : tkind}
sinusoidal_generator_Initize
(gen : &sinusoidal_generator tk?
>> sinusoidal_generator tk,
phase : g0float tk,
afreq : g0float tk) : void =
gen := @{phase = phase,
afreq = afreq,
clock0 = clock (),
stopped = true}
fn {tk : tkind}
sinusoidal_generator_Start
(gen : &sinusoidal_generator tk) : void =
gen.stopped := false
(* IMO changing the integrator's input is bad OO design: akin to
unplugging one generator and plugging in another. What we REALLY
want is to have the generator produce a different signal. So
gen.Stop() will connect the output to a constant
zero. (Alternatively, the channel between the signal source and the
integrator could effect the shutoff.) *)
fn {tk : tkind}
sinusoidal_generator_Stop
(gen : &sinusoidal_generator tk) : void =
gen.stopped := true
fn {tk : tkind}
sinusoidal_generator_Sample
(gen : !sinusoidal_generator tk) : g0float tk =
let
val @{phase = phase,
afreq = afreq,
clock0 = clock0,
stopped = stopped} = gen
in
if stopped then
g0i2f 0
else
let
val t = (clock2f (clock () - clock0)) : g0float tk
in
sinpi ((afreq * t) + phase)
end
end
overload .Initize with sinusoidal_generator_Initize
overload .Start with sinusoidal_generator_Start
overload .Stop with sinusoidal_generator_Stop
overload .Sample with sinusoidal_generator_Sample
(*------------------------------------------------------------------*)
vtypedef inputter (tk : tkind, p : addr) =
(* This is a closure type that can reside either in the heap or on
the stack. *)
@((() -<clo1> g0float tk) @ p | ptr p)
vtypedef active_integrator (tk : tkind, p : addr) =
@{
inputter = inputter (tk, p),
t_last = clock_t,
sample_last = g0float tk,
integral = g0float tk
}
vtypedef active_integrator (tk : tkind) =
[p : addr] active_integrator (tk, p)
fn {tk : tkind}
active_integrator_Input
{p : addr}
(igrator : &active_integrator tk?
>> active_integrator (tk, p),
inputter : inputter (tk, p)) : void =
let
val now = clock ()
in
igrator := @{inputter = inputter,
t_last = now,
sample_last = g0i2f 0,
integral = g0i2f 0}
end
fn {tk : tkind}
active_integrator_Output
{p : addr}
(igrator : !active_integrator (tk, p)) : g0float tk =
igrator.integral
fn {tk : tkind}
active_integrator_Integrate
{p : addr}
(igrator : &active_integrator (tk, p)) : void =
let
val @{inputter = @(pf | p),
t_last = t_last,
sample_last = sample_last,
integral = integral} = igrator
macdef inputter_closure = !p
val t_now = clock ()
val sample_now = inputter_closure ()
val integral = integral + ((sample_last + sample_now)
* clock2f (t_last - t_now)
* g0f2f 0.5)
val sample_last = sample_now
val t_last = t_now
val () = igrator := @{inputter = @(pf | p),
t_last = t_last,
sample_last = sample_last,
integral = integral}
in
end
overload .Input with active_integrator_Input
overload .Output with active_integrator_Output
overload .Integrate with active_integrator_Integrate
(*------------------------------------------------------------------*)
implement
main () =
let
(* We put on the stack all objects that are not in registers. Thus
we avoid the need for malloc/free. *)
vtypedef gen_vt = sinusoidal_generator float_kind
vtypedef igrator_vt = active_integrator float_kind
var gen : gen_vt
var igrator : igrator_vt
val phase = 0.0f
and afreq = 1.0f (* Frequency of 0.5 Hz. *)
val () = gen.Initize (phase, afreq)
val () = gen.Start ()
(* Create a thunk on the stack. This thunk acts as a channel
between the sinusoidal generator and the active integrator. We
could probably work this step into the OO style of most of the
code, but doing that is left as an exercise. The mechanics of
creating a closure on the stack are already enough for a person
to absorb. (Of course, rather than use a closure, we could have
set up a type hierarchy. However, IMO a type hierarchy is
needlessly clumsy. Joining the objects with a closure lets any
thunk of the correct type serve as input.) *)
val p_gen = addr@ gen
var gen_clo_on_stack =
lam@ () : float =<clo1>
let
(* A little unsafeness is needed here. AFAIK there is no way
to SAFELY enclose the stack variable "gen" in the
closure. A negative effect is that (at least without some
elaborate scheme) it becomes POSSIBLE to use this
closure, even after "gen" has been destroyed. But we will
be careful not to do that. *)
extern praxi p2view :
{p : addr} ptr p -<prf>
(gen_vt @ p, gen_vt @ p -<lin,prf> void)
prval @(pf, fpf) = p2view p_gen
macdef gen = !p_gen
val sample = gen.Sample ()
prval () = fpf pf
in
sample
end
val sinusoidal_inputter =
@(view@ gen_clo_on_stack | addr@ gen_clo_on_stack)
val () = igrator.Input (sinusoidal_inputter)
fn {}
integrate_for_seconds
(igrator : &igrator_vt,
seconds : float) : void =
let
val t0 = clock2f (clock ())
fun
loop (igrator : &igrator_vt) : void =
if clock2f (clock ()) - t0 < seconds then
begin
igrator.Integrate ();
pause ();
loop igrator
end
in
loop igrator
end
(* Start the sinusoid and then integrate for 2.0 seconds. *)
val () = gen.Start ()
val () = integrate_for_seconds (igrator, 2.0f)
(* Stop the sinusoid and then integrate for 0.5 seconds. *)
val () = gen.Stop ()
val () = integrate_for_seconds (igrator, 0.5f)
val () = println! ("integrator output = ", igrator.Output ());
(* The following "prval" lines are necessary for type-safety, and
produce no executable code. *)
prval @{inputter = @(pf | _),
t_last = _,
sample_last = _,
integral = _} = igrator
prval () = view@ gen_clo_on_stack := pf
in
0
end
(*------------------------------------------------------------------*)
- Output:
One will get different results on different runs.
$ patscc -std=gnu2x -Ofast active_object_task.dats -lm && ./a.out integrator output = -0.000002
BASIC
BBC BASIC
INSTALL @lib$+"CLASSLIB"
INSTALL @lib$+"TIMERLIB"
INSTALL @lib$+"NOWAIT"
REM Integrator class:
DIM integ{f$, t#, v#, tid%, @init, @@exit, input, output, tick}
PROC_class(integ{})
REM Methods:
DEF integ.@init integ.f$ = "0" : integ.tid% = FN_ontimer(10, PROC(integ.tick), 1) : ENDPROC
DEF integ.@@exit PROC_killtimer(integ.tid%) : ENDPROC
DEF integ.input (f$) integ.f$ = f$ : ENDPROC
DEF integ.output = integ.v#
DEF integ.tick integ.t# += 0.01 : integ.v# += EVAL(integ.f$) : ENDPROC
REM Test:
PROC_new(myinteg{}, integ{})
PROC(myinteg.input) ("SIN(2*PI*0.5*myinteg.t#)")
PROCwait(200)
PROC(myinteg.input) ("0")
PROCwait(50)
PRINT "Final value = " FN(myinteg.output)
PROC_discard(myinteg{})
Output:
Final value = -1.43349462E-6
C
Uses POSIX threads.
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <math.h>
#include <sys/time.h>
#include <pthread.h>
/* no need to lock the object: at worst the readout would be 1 tick off,
which is no worse than integrator's inate inaccuracy */
typedef struct {
double (*func)(double);
struct timeval start;
double v, last_v, last_t;
pthread_t id;
} integ_t, *integ;
void update(integ x)
{
struct timeval tv;
double t, v, (*f)(double);
f = x->func;
gettimeofday(&tv, 0);
t = ((tv.tv_sec - x->start.tv_sec) * 1000000
+ tv.tv_usec - x->start.tv_usec) * 1e-6;
v = f ? f(t) : 0;
x->v += (x->last_v + v) * (t - x->last_t) / 2;
x->last_t = t;
}
void* tick(void *a)
{
integ x = a;
while (1) {
usleep(100000); /* update every .1 sec */
update(x);
}
}
void set_input(integ x, double (*func)(double))
{
update(x);
x->func = func;
x->last_t = 0;
x->last_v = func ? func(0) : 0;
}
integ new_integ(double (*func)(double))
{
integ x = malloc(sizeof(integ_t));
x->v = x->last_v = 0;
x->func = 0;
gettimeofday(&x->start, 0);
set_input(x, func);
pthread_create(&x->id, 0, tick, x);
return x;
}
double sine(double t) { return sin(4 * atan2(1, 1) * t); }
int main()
{
integ x = new_integ(sine);
sleep(2);
set_input(x, 0);
usleep(500000);
printf("%g\n", x->v);
return 0;
}
output
-9.99348e-05
C#
using System;
using System.Threading.Tasks;
using static System.Diagnostics.Stopwatch;
using static System.Math;
using static System.Threading.Thread;
class ActiveObject
{
static double timeScale = 1.0 / Frequency;
Func<double, double> func;
Task updateTask;
double integral;
double value;
long timestamp0, timestamp;
public ActiveObject(Func<double, double> input)
{
timestamp0 = timestamp = GetTimestamp();
func = input;
value = func(0);
updateTask = Integrate();
}
public void ChangeInput(Func<double, double> input)
{
lock (updateTask)
{
func = input;
}
}
public double Value
{
get
{
lock (updateTask)
{
return integral;
}
}
}
async Task Integrate()
{
while (true)
{
await Task.Yield();
var newTime = GetTimestamp();
double newValue;
lock (updateTask)
{
newValue = func((newTime - timestamp0) * timeScale);
integral += (newValue + value) * (newTime - timestamp) * timeScale / 2;
}
timestamp = newTime;
value = newValue;
}
}
}
class Program
{
static Func<double, double> Sine(double frequency) =>
t => Sin(2 * PI * frequency * t);
static void Main(string[] args)
{
var ao = new ActiveObject(Sine(0.5));
Sleep(TimeSpan.FromSeconds(2));
ao.ChangeInput(t => 0);
Sleep(TimeSpan.FromSeconds(0.5));
Console.WriteLine(ao.Value);
}
}
Output:
8.62230019255E-5
C++
#include <atomic>
#include <chrono>
#include <cmath>
#include <iostream>
#include <mutex>
#include <thread>
using namespace std::chrono_literals;
class Integrator
{
public:
using clock_type = std::chrono::high_resolution_clock;
using dur_t = std::chrono::duration<double>;
using func_t = double(*)(double);
explicit Integrator(func_t f = nullptr);
~Integrator();
void input(func_t new_input);
double output() { return integrate(); }
private:
std::atomic_flag continue_;
std::mutex mutex;
std::thread worker;
func_t func;
double state = 0;
//Improves precision by reducing sin result error on large values
clock_type::time_point const beginning = clock_type::now();
clock_type::time_point t_prev = beginning;
void do_work();
double integrate();
};
Integrator::Integrator(func_t f) : func(f)
{
continue_.test_and_set();
worker = std::thread(&Integrator::do_work, this);
}
Integrator::~Integrator()
{
continue_.clear();
worker.join();
}
void Integrator::input(func_t new_input)
{
integrate();
std::lock_guard<std::mutex> lock(mutex);
func = new_input;
}
void Integrator::do_work()
{
while (continue_.test_and_set()) {
integrate();
std::this_thread::sleep_for(1ms);
}
}
double Integrator::integrate()
{
std::lock_guard<std::mutex> lock(mutex);
auto now = clock_type::now();
dur_t start = t_prev - beginning;
dur_t fin = now - beginning;
if (func)
state += (func(start.count()) + func(fin.count())) * (fin - start).count() / 2;
t_prev = now;
return state;
}
double sine(double time)
{
constexpr double PI = 3.1415926535897932;
return std::sin(2 * PI * 0.5 * time);
}
int main()
{
Integrator foo(sine);
std::this_thread::sleep_for(2s);
foo.input(nullptr);
std::this_thread::sleep_for(500ms);
std::cout << foo.output();
}
output
1.23136e-011
Clojure
(ns active-object
(:import (java.util Timer TimerTask)))
(defn input [integrator k]
(send integrator assoc :k k))
(defn output [integrator]
(:s @integrator))
(defn tick [integrator t1]
(send integrator
(fn [{:keys [k s t0] :as m}]
(assoc m :s (+ s (/ (* (+ (k t1) (k t0)) (- t1 t0)) 2.0)) :t0 t1))))
(defn start-timer [integrator interval]
(let [timer (Timer. true)
start (System/currentTimeMillis)]
(.scheduleAtFixedRate timer
(proxy [TimerTask] []
(run [] (tick integrator (double (/ (- (System/currentTimeMillis) start) 1000)))))
(long 0)
(long interval))
#(.cancel timer)))
(defn test-integrator []
(let [integrator (agent {:k (constantly 0.0) :s 0.0 :t0 0.0})
stop-timer (start-timer integrator 10)]
(input integrator #(Math/sin (* 2.0 Math/PI 0.5 %)))
(Thread/sleep 2000)
(input integrator (constantly 0.0))
(Thread/sleep 500)
(println (output integrator))
(stop-timer)))
user> (test-integrator)
1.414065859052494E-5
Common Lisp
(defclass integrator ()
((input :initarg :input :writer input :reader %input)
(lock :initform (bt:make-lock) :reader lock)
(start-time :initform (get-internal-real-time) :reader start-time)
(interval :initarg :interval :reader interval)
(thread :reader thread :writer %set-thread)
(area :reader area :initform 0 :accessor %area)))
(defmethod shared-initialize
((integrator integrator) slot-names &key (interval nil interval-s-p) &allow-other-keys)
(declare (ignore interval))
(cond
;; Restart the thread if any unsynchronized slots are
;; being initialized
((or
(eql slot-names t)
(member 'thread slot-names)
(member 'interval slot-names)
(member 'start-time slot-names)
(member 'lock slot-names)
interval-s-p)
;; If the instance already has a thread, stop it and wait for it
;; to stop before initializing any slots
(when (slot-boundp integrator 'thread)
(input nil integrator)
(bt:join-thread (thread integrator)))
(call-next-method)
(let* ((now (get-internal-real-time))
(current-value (funcall (%input integrator) (- (start-time integrator) now))))
(%set-thread
(bt:make-thread
(lambda ()
(loop
;; Sleep for the amount required to reach the next interval;
;; mitigates drift from theoretical interval times
(sleep
(mod
(/ (- (start-time integrator) (get-internal-real-time))
internal-time-units-per-second)
(interval integrator)))
(let* ((input
(bt:with-lock-held ((lock integrator))
;; If input is nil, exit the thread
(or (%input integrator) (return))))
(previous-time (shiftf now (get-internal-real-time)))
(previous-value
(shiftf
current-value
(funcall input (/ (- now (start-time integrator)) internal-time-units-per-second)))))
(bt:with-lock-held ((lock integrator))
(incf (%area integrator)
(*
(/ (- now previous-time)
internal-time-units-per-second)
(/ (+ previous-value current-value)
2)))))))
:name "integrator-thread")
integrator)))
(t
;; If lock is not in SLOT-NAMES, it must already be initialized,
;; so it can be taken while slots synchronized to it are set
(bt:with-lock-held ((lock integrator))
(call-next-method)))))
(defmethod input :around (new-value (integrator integrator))
(bt:with-lock-held ((lock integrator))
(call-next-method)))
(defmethod area :around ((integrator integrator))
(bt:with-lock-held ((lock integrator))
(call-next-method)))
(let ((integrator
(make-instance 'integrator
:input (lambda (time) (sin (* 2 pi 0.5 time)))
:interval 1/1000)))
(unwind-protect
(progn
(sleep 2)
(input (constantly 0) integrator)
(sleep 0.5)
(format t "~F~%" (area integrator)))
(input nil integrator)))
Crystal
Crystal currently runs all code in a single thread, so a trivial example wouldn't have any issues with thread safety. However, this behavior will likely change in the future. This example was written with that in mind, and is somewhat more complex to show better idioms and be future-proof.
require "math"
require "time"
# this enum allows us to specify what type of message the proc_chan received.
# this trivial example only has one action, but more enum members can be added
# to update the proc, or take other actions
enum Action
Finished # we've waited long enough, and are asking for our result
# Update # potential member representing an update to the integrator function
end
class Integrator
property interval : Float64
getter s : Float64 = 0f64
# initialize our k function as a proc that takes a float and just returns 0
getter k : Proc(Float64, Float64) = ->(t : Float64) { 0f64 }
# channels used for communicating with the main fiber
@proc_chan : Channel(Tuple(Action, Proc(Float64, Float64)|Nil))
@result_chan : Channel(Float64)
def initialize(@k, @proc_chan, @result_chan, @interval = 1e-4)
# use a monotonic clock for accuracy
start = Time.monotonic.total_seconds
t0, k0 = 0f64, @k.call(0f64)
loop do
# this sleep returns control to the main fiber. if the main fiber hasn't finished sleeping,
# control will be returned to this loop
sleep interval.seconds
# check the channel to see if the function has changed
self.check_channel()
t1 = Time.monotonic.total_seconds - start
k1 = @k.call(t1)
@s += (k1 + k0) * (t1 - t0) / 2.0
t0, k0 = t1, k1
end
end
# check the proc_chan for messages, update the integrator function or send the result as needed
def check_channel
select
when message = @proc_chan.receive
action, new_k = message
case action
when Action::Finished
@result_chan.send @s
@k = new_k unless new_k.nil?
end
else
nil
end
end
end
# this channel allows us to update the integrator function,
# and inform the integrator to send the result over the result channel
proc_chan = Channel(Tuple(Action, Proc(Float64, Float64)|Nil)).new
# channel used to return the result from the integrator
result_chan = Channel(Float64).new
# run everything in a new top-level fiber to avoid shared memory issues.
# since the fiber immediately sleeps, control is returned to the main code.
# the main code then sleeps for two seconds, returning control to our state_clock fiber.
# when two seconds is up, this state_clock fiber will return control
# to the main code on the next `sleep interval.seconds`
spawn name: "state_clock" do
ai = Integrator.new ->(t : Float64) { Math.sin(Math::PI * t) }, proc_chan, result_chan
end
sleep 2.seconds
proc_chan.send({Action::Finished, ->(t : Float64) { 0f64 }})
sleep 0.5.seconds
puts result_chan.receive
Output:
-2.5475883655389925e-10
D
import core.thread;
import std.datetime;
import std.math;
import std.stdio;
void main() {
auto func = (double t) => sin(cast(double) PI * t);
Integrator integrator = new Integrator(func);
Thread.sleep(2000.msecs);
integrator.setFunc(t => 0.0);
Thread.sleep(500.msecs);
integrator.stop();
writeln(integrator.getOutput());
}
/**
* Integrates input function K over time
* S + (t1 - t0) * (K(t1) + K(t0)) / 2
*/
public class Integrator {
public alias Function = double function (double);
private SysTime start;
private shared bool running;
private Function func;
private shared double t0;
private shared double v0;
private shared double sum = 0.0;
public this(Function func) {
this.start = Clock.currTime();
setFunc(func);
new Thread({
integrate();
}).start();
}
public void setFunc(Function func) {
this.func = func;
v0 = func(0.0);
t0 = 0.0;
}
public double getOutput() {
return sum;
}
public void stop() {
running = false;
}
private void integrate() {
running = true;
while (running) {
Thread.sleep(1.msecs);
update();
}
}
private void update() {
import core.atomic;
Duration t1 = (Clock.currTime() - start);
double v1 = func(t1.total!"msecs");
double rect = (t1.total!"msecs" - t0) * (v0 + v1) / 2;
atomicOp!"+="(this.sum, rect);
t0 = t1.total!"msecs";
v0 = v1;
}
}
- Output:
-3.07837e-13
Delphi
program Active_object;
{$APPTYPE CONSOLE}
uses
System.SysUtils,
System.Classes;
type
TIntegrator = class(TThread)
private
{ Private declarations }
interval, s: double;
IsRunning: Boolean;
protected
procedure Execute; override;
public
k: Tfunc<Double, Double>;
constructor Create(k: Tfunc<Double, Double>; inteval: double = 1e-4); overload;
procedure Join;
end;
{ TIntegrator }
constructor TIntegrator.Create(k: Tfunc<Double, Double>; inteval: double = 1e-4);
begin
self.interval := Interval;
self.K := k;
self.S := 0.0;
IsRunning := True;
FreeOnTerminate := True;
inherited Create(false);
end;
procedure TIntegrator.Execute;
var
interval, t0, k0, t1, k1: double;
start: Cardinal;
begin
inherited;
interval := self.interval;
start := GetTickCount;
t0 := 0;
k0 := self.K(0);
while IsRunning do
begin
t1 := (GetTickCount - start) / 1000;
k1 := self.K(t1);
self.S := self.S + ((k1 + k0) * (t1 - t0) / 2.0);
t0 := t1;
k0 := k1;
end;
end;
procedure TIntegrator.Join;
begin
IsRunning := false;
end;
var
Integrator: TIntegrator;
begin
Integrator := TIntegrator.create(
function(t: double): double
begin
Result := sin(pi * t);
end);
sleep(2000);
Writeln(Integrator.s);
Integrator.k :=
function(t: double): double
begin
Result := 0;
end;
sleep(500);
Writeln(Integrator.s);
Integrator.Join;
Readln;
end.
- Output:
-1.51242391413465E-0016 -1.51242391413465E-0016
E
def makeIntegrator() {
var value := 0.0
var input := fn { 0.0 }
var input1 := input()
var t1 := timer.now()
def update() {
def t2 := timer.now()
def input2 :float64 := input()
def dt := (t2 - t1) / 1000
value += (input1 + input2) * dt / 2
t1 := t2
input1 := input2
}
var task() {
update <- ()
task <- ()
}
task()
def integrator {
to input(new) :void { input := new }
to output() :float64 { return value }
to shutdown() { task := fn {} }
}
return integrator
}
def test() {
def result
def pi := (-1.0).acos()
def freq := pi / 1000
def base := timer.now()
def i := makeIntegrator()
i.input(fn { (freq * timer.now()).sin() })
timer.whenPast(base + 2000, fn {
i.input(fn {0})
})
timer.whenPast(base + 2500, fn {
bind result := i.output()
i.shutdown()
})
return result
}
EchoLisp
We use the functions (at ..) : scheduling, (wait ...), and (every ...) ot the timer.lib. The accuracy will be function of the browser's functions setTimeout and setInterval ...
(require 'timer)
;; returns an 'object' : (&lamdba; message [values])
;; messages : input, output, sample, inspect
(define (make-active)
(let [
(t0 #f) (dt 0)
(t 0) (Kt 0) ; K(t)
(S 0) (K 0)]
(lambda (message . args)
(case message
((output) (// S 2))
((input ) (set! K (car args)) (set! t0 #f))
((inspect) (printf " Active obj : t0 %v t %v S %v " t0 t Kt (// S 2 )))
((sample)
(when (procedure? K)
;; recved new K : init
(unless t0
(set! t0 (first args))
(set! t 0)
(set! Kt (K 0)))
;; integrate K(t) every time 'sample message is received
(set! dt (- (first args) t t0)) ;; compute once K(t)
(set! S (+ S (* dt Kt)))
(set! t (+ t dt))
(set! Kt (K t))
(set! S (+ S (* dt Kt)))))
(else (error "active:bad message" message))))))
- Output:
(define (experiment)
(define (K t) (sin (* PI t )))
(define A (make-active))
(define (stop) (A 'input 0))
(define (sample t) (A 'sample (// t 1000)))
(define (result) (writeln 'result (A 'output)))
(at 2.5 'seconds 'result)
(every 10 'sample) ;; integrate every 10 ms
(A 'input K)
(wait 2000 'stop))
(experiment) →
3/7/2015 20:34:18 : result
result 0.0002266920372221955
(experiment) →
3/7/2015 20:34:28 : result
result 0.00026510586971023164
Erlang
I could not see what time to use between each integration so it is the argument to task().
-module( active_object ).
-export( [delete/1, input/2, new/0, output/1, task/1] ).
-compile({no_auto_import,[time/0]}).
delete( Object ) ->
Object ! stop.
input( Object, Fun ) ->
Object ! {input, Fun}.
new( ) ->
K = fun zero/1,
S = 0,
T0 = seconds_with_decimals(),
erlang:spawn( fun() -> loop(K, S, T0) end ).
output( Object ) ->
Object ! {output, erlang:self()},
receive
{output, Object, Output} -> Output
end.
task( Integrate_millisec ) ->
Object = new(),
{ok, _Ref} = timer:send_interval( Integrate_millisec, Object, integrate ),
io:fwrite( "New ~p~n", [output(Object)] ),
input( Object, fun sine/1 ),
timer:sleep( 2000 ),
io:fwrite( "Sine ~p~n", [output(Object)] ),
input( Object, fun zero/1 ),
timer:sleep( 500 ),
io:fwrite( "Approx ~p~n", [output(Object)] ),
delete( Object ).
loop( Fun, Sum, T0 ) ->
receive
integrate ->
T1 = seconds_with_decimals(),
New_sum = trapeze( Sum, Fun, T0, T1 ),
loop( Fun, New_sum, T1 );
stop ->
ok;
{input, New_fun} ->
loop( New_fun, Sum, T0 );
{output, Pid} ->
Pid ! {output, erlang:self(), Sum},
loop( Fun, Sum, T0 )
end.
sine( T ) ->
math:sin( 2 * math:pi() * 0.5 * T ).
seconds_with_decimals() ->
{Megaseconds, Seconds, Microseconds} = os:timestamp(),
(Megaseconds * 1000000) + Seconds + (Microseconds / 1000000).
trapeze( Sum, Fun, T0, T1 ) ->
Sum + (Fun(T1) + Fun(T0)) * (T1 - T0) / 2.
zero( _ ) -> 0.
F#
open System
open System.Threading
// current time in seconds
let now() = float( DateTime.Now.Ticks / 10000L ) / 1000.0
type Integrator( intervalMs ) as x =
let mutable k = fun _ -> 0.0 // function to integrate
let mutable s = 0.0 // current value
let mutable t0 = now() // last time s was updated
let mutable running = true // still running?
do x.ScheduleNextUpdate()
member x.Input(f) = k <- f
member x.Output() = s
member x.Stop() = running <- false
member private x.Update() =
let t1 = now()
s <- s + (k t0 + k t1) * (t1 - t0) / 2.0
t0 <- t1
x.ScheduleNextUpdate()
member private x.ScheduleNextUpdate() =
if running then
async { do! Async.Sleep( intervalMs )
x.Update()
}
|> Async.Start
let i = new Integrator(10)
i.Input( fun t -> Math.Sin (2.0 * Math.PI * 0.5 * t) )
Thread.Sleep(2000)
i.Input( fun _ -> 0.0 )
Thread.Sleep(500)
printfn "%f" (i.Output())
i.Stop()
Factor
Working with dynamic quotations requires the stack effect to be known in advance. The apply-stack-effect serves this purpose.
USING: accessors alarms calendar combinators kernel locals math
math.constants math.functions prettyprint system threads ;
IN: rosettacode.active
TUPLE: active-object alarm function state previous-time ;
: apply-stack-effect ( quot -- quot' )
[ call( x -- x ) ] curry ; inline
: nano-to-seconds ( -- seconds ) nano-count 9 10^ / ;
: object-times ( active-object -- t1 t2 )
[ previous-time>> ]
[ nano-to-seconds [ >>previous-time drop ] keep ] bi ;
:: adding-function ( t1 t2 active-object -- function )
t2 t1 active-object function>> apply-stack-effect bi@ +
t2 t1 - * 2 / [ + ] curry ;
: integrate ( active-object -- )
[ object-times ]
[ adding-function ]
[ swap apply-stack-effect change-state drop ] tri ;
: <active-object> ( -- object )
active-object new
0 >>state
nano-to-seconds >>previous-time
[ drop 0 ] >>function
dup [ integrate ] curry 1 nanoseconds every >>alarm ;
: destroy ( active-object -- ) alarm>> cancel-alarm ;
: input ( object quot -- object ) >>function ;
: output ( object -- val ) state>> ;
: active-test ( -- )
<active-object>
[ 2 pi 0.5 * * * sin ] input
2 seconds sleep
[ drop 0 ] input
0.5 seconds sleep
[ output . ] [ destroy ] bi ;
MAIN: active-test
( scratchpad ) "rosettacode.active" run -5.294207647335787e-05
FBSL
The Dynamic Assembler and Dynamic C JIT compilers integrated in FBSL v3.5 handle multithreading perfectly well. However, pure FBSL infrastructure has never been designed especially to support own multithreading nor can it handle long long integers natively. Yet a number of tasks with careful design and planning are quite feasible in pure FBSL too:
#APPTYPE CONSOLE
#INCLUDE <Include\Windows.inc>
DIM Entity AS NEW Integrator(): Sleep(2000) ' respawn and do the job
Entity.Relax(): Sleep(500) ' get some rest
PRINT ">>> ", Entity.Yield(): DELETE Entity ' report and die
PAUSE
' ------------- End Program Code -------------
#DEFINE SpawnMutex CreateMutex(NULL, FALSE, "mutex")
#DEFINE LockMutex WaitForSingleObject(mutex, INFINITE)
#DEFINE UnlockMutex ReleaseMutex(mutex)
#DEFINE KillMutex CloseHandle(mutex)
CLASS Integrator
PRIVATE:
TYPE LARGE_INTEGER
lowPart AS INTEGER
highPart AS INTEGER
END TYPE
DIM dfreq AS DOUBLE, dlast AS DOUBLE, dnow AS DOUBLE, llint AS LARGE_INTEGER
DIM dret0 AS DOUBLE, dret1 AS DOUBLE, mutex AS INTEGER, sum AS DOUBLE, thread AS INTEGER
' --------------------------------------------
SUB INITIALIZE()
mutex = SpawnMutex
QueryPerformanceFrequency(@llint)
dfreq = LargeInt2Double(llint)
QueryPerformanceCounter(@llint)
dlast = LargeInt2Double(llint) / dfreq
thread = FBSLTHREAD(ADDRESSOF Sampler)
FBSLTHREADRESUME(thread)
END SUB
SUB TERMINATE()
' nothing special
END SUB
' --------------------------------------------
SUB Sampler()
DO
LockMutex
Sleep(5)
QueryPerformanceCounter(@llint)
dnow = LargeInt2Double(llint) / dfreq
dret0 = Task(dlast): dret1 = Task(dnow)
sum = sum + (dret1 + dret0) * (dnow - dlast) / 2
dlast = dnow
UnlockMutex
LOOP
END SUB
FUNCTION LargeInt2Double(obj AS VARIANT) AS DOUBLE
STATIC ret
ret = obj.highPart
IF obj.highPart < 0 THEN ret = ret + (2 ^ 32)
ret = ret * 2 ^ 32
ret = ret + obj.lowPart
IF obj.lowPart < 0 THEN ret = ret + (2 ^ 32)
RETURN ret
END FUNCTION
PUBLIC:
METHOD Relax()
LockMutex
ADDRESSOF Task = ADDRESSOF Idle
UnlockMutex
END METHOD
METHOD Yield() AS DOUBLE
LockMutex
Yield = sum
FBSLTHREADKILL(thread)
UnlockMutex
KillMutex
END METHOD
END CLASS
FUNCTION Idle(BYVAL t AS DOUBLE) AS DOUBLE
RETURN 0.0
END FUNCTION
FUNCTION Task(BYVAL t AS DOUBLE) AS DOUBLE
RETURN SIN(2 * PI * 0.5 * t)
END FUNCTION
Typical console output:
>>> -0.000769965989580346 Press any key to continue...
FreeBASIC
#define twopi 6.2831853071795864769252867665590057684
dim shared as double S = 0 'set up the state as a global variable
dim shared as double t0, t1, ta
function sine( x as double, f as double ) as double
return sin(twopi*f*x)
end function
function zero( x as double, f as double ) as double
return 0
end function
sub integrate( K as function(as double, as double) as double, f as double )
'represent input as pointers to functions
t1 = timer
s += (K(t1,f) + K(t0,f))*(t1-t0)/2.0
t0 = t1
end sub
t0 = timer
ta = timer
while timer-ta <= 2.5
if timer-ta <= 2 then integrate( @sine, 0.5 ) else integrate( @zero, 0 )
wend
print S
- Output:
8.926050531860172e-07
Go
Using time.Tick to sample K at a constant frequency. Three goroutines are involved, main, aif, and tk. Aif controls access to the accumulator s and the integration function K. Tk and main must talk to aif through channels to access s and K.
package main
import (
"fmt"
"math"
"time"
)
// type for input function, k.
// input is duration since an arbitrary start time t0.
type tFunc func(time.Duration) float64
// active integrator object. state variables are not here, but in
// function aif, started as a goroutine in the constructor.
type aio struct {
iCh chan tFunc // channel for setting input function
oCh chan chan float64 // channel for requesting output
}
// constructor
func newAio() *aio {
var a aio
a.iCh = make(chan tFunc)
a.oCh = make(chan chan float64)
go aif(&a)
return &a
}
// input method required by task description. in practice, this method is
// unnecessary; you would just put that single channel send statement in
// your code wherever you wanted to set the input function.
func (a aio) input(f tFunc) {
a.iCh <- f
}
// output method required by task description. in practice, this method too
// would not likely be best. instead any client interested in the value would
// likely make a return channel sCh once, and then reuse it as needed.
func (a aio) output() float64 {
sCh := make(chan float64)
a.oCh <- sCh
return <-sCh
}
// integration function that returns constant 0
func zeroFunc(time.Duration) float64 { return 0 }
// goroutine serializes access to integrated function k and state variable s
func aif(a *aio) {
var k tFunc = zeroFunc // integration function
s := 0. // "object state" initialized to 0
t0 := time.Now() // initial time
k0 := k(0) // initial sample value
t1 := t0 // t1, k1 used for trapezoid formula
k1 := k0
tk := time.Tick(10 * time.Millisecond) // 10 ms -> 100 Hz
for {
select {
case t2 := <-tk: // timer tick event
k2 := k(t2.Sub(t0)) // new sample value
s += (k1 + k2) * .5 * t2.Sub(t1).Seconds() // trapezoid formula
t1, k1 = t2, k2 // save time and value
case k = <-a.iCh: // input method event: function change
case sCh := <-a.oCh: // output method event: sample object state
sCh <- s
}
}
}
func main() {
a := newAio() // create object
a.input(func(t time.Duration) float64 { // 1. set input to sin function
return math.Sin(t.Seconds() * math.Pi)
})
time.Sleep(2 * time.Second) // 2. sleep 2 sec
a.input(zeroFunc) // 3. set input to zero function
time.Sleep(time.Second / 2) // 4. sleep .5 sec
fmt.Println(a.output()) // output should be near zero
}
Output:
2.4517135756807704e-05
Groovy
/**
* Integrates input function K over time
* S + (t1 - t0) * (K(t1) + K(t0)) / 2
*/
class Integrator {
interface Function {
double apply(double timeSinceStartInSeconds)
}
private final long start
private volatile boolean running
private Function func
private double t0
private double v0
private double sum
Integrator(Function func) {
this.start = System.nanoTime()
setFunc(func)
new Thread({ this.&integrate() }).start()
}
void setFunc(Function func) {
this.func = func
def temp = func.apply(0.0.toDouble())
v0 = temp
t0 = 0.0.doubleValue()
}
double getOutput() {
return sum
}
void stop() {
running = false
}
private void integrate() {
running = true
while (running) {
try {
Thread.sleep(1)
update()
} catch (InterruptedException ignored) {
return
}
}
}
private void update() {
double t1 = (System.nanoTime() - start) / 1.0e9
double v1 = func.apply(t1)
double rect = (t1 - t0) * (v0 + v1) / 2.0
this.sum += rect
t0 = t1
v0 = v1
}
static void main(String[] args) {
Integrator integrator = new Integrator({ t -> Math.sin(Math.PI * t) })
Thread.sleep(2000)
integrator.setFunc({ t -> 0.0.toDouble() })
Thread.sleep(500)
integrator.stop()
System.out.println(integrator.getOutput())
}
}
- Output:
0.0039642136156300455
Haskell
module Integrator (
newIntegrator, input, output, stop,
Time, timeInterval
) where
import Control.Concurrent (forkIO, threadDelay)
import Control.Concurrent.MVar (MVar, newMVar, modifyMVar_, modifyMVar, readMVar)
import Control.Exception (evaluate)
import Data.Time (UTCTime)
import Data.Time.Clock (getCurrentTime, diffUTCTime)
-- RC task
main = do let f = 0.5 {- Hz -}
t0 <- getCurrentTime
i <- newIntegrator
input i (\t -> sin(2*pi * f * timeInterval t0 t)) -- task step 1
threadDelay 2000000 {- µs -} -- task step 2
input i (const 0) -- task step 3
threadDelay 500000 {- µs -} -- task step 4
result <- output i
stop i
print result
---- Implementation ------------------------------------------------------
-- Utilities for working with the time type
type Time = UTCTime
type Func a = Time -> a
timeInterval t0 t1 = realToFrac $ diffUTCTime t1 t0
-- Type signatures of the module's interface
newIntegrator :: Fractional a => IO (Integrator a) -- Create an integrator
input :: Integrator a -> Func a -> IO () -- Set the input function
output :: Integrator a -> IO a -- Get the current value
stop :: Integrator a -> IO () -- Stop integration, don't waste CPU
-- Data structures
data Integrator a = Integrator (MVar (IntState a)) -- MVar is a thread-safe mutable cell
deriving Eq
data IntState a = IntState { func :: Func a, -- The current function
run :: Bool, -- Whether to keep going
value :: a, -- The current accumulated value
time :: Time } -- The time of the previous update
newIntegrator = do
now <- getCurrentTime
state <- newMVar $ IntState { func = const 0,
run = True,
value = 0,
time = now }
thread <- forkIO (intThread state) -- The state variable is shared between the thread
return (Integrator state) -- and the client interface object.
input (Integrator stv) f = modifyMVar_ stv (\st -> return st { func = f })
output (Integrator stv) = fmap value $ readMVar stv
stop (Integrator stv) = modifyMVar_ stv (\st -> return st { run = False })
-- modifyMVar_ takes an MVar and replaces its contents according to the provided function.
-- a { b = c } is record-update syntax: "the record a, except with field b changed to c"
-- Integration thread
intThread :: Fractional a => MVar (IntState a) -> IO ()
intThread stv = whileM $ modifyMVar stv updateAndCheckRun
-- modifyMVar is like modifyMVar_ but the function returns a tuple of the new value
-- and an arbitrary extra value, which in this case ends up telling whileM whether
-- to keep looping.
where updateAndCheckRun st = do
now <- getCurrentTime
let value' = integrate (func st) (value st) (time st) now
evaluate value' -- avoid undesired laziness
return (st { value = value', time = now }, -- updated state
run st) -- whether to continue
integrate :: Fractional a => Func a -> a -> Time -> Time -> a
integrate f value t0 t1 = value + (f t0 + f t1)/2 * dt
where dt = timeInterval t0 t1
-- Execute 'action' until it returns false.
whileM action = do b <- action; if b then whileM action else return ()
J
Implementation:
coclass 'activeobject'
require'dates'
create=:setinput NB. constructor
T=:3 :0
if. nc<'T0' do. T0=:tsrep 6!:0'' end.
0.001*(tsrep 6!:0'')-T0
)
F=:G=:0:
Zero=:0
setinput=:3 :0
zero=. getoutput''
'`F ignore'=: y,_:`''
G=: F f.d._1
Zero=: zero-G T ''
getoutput''
)
getoutput=:3 :0
Zero+G T''
)
Task example (code):
cocurrent 'testrig'
delay=: 6!:3
object=: conew 'activeobject'
setinput__object 1&o.@o.`''
smoutput (T__object,getoutput__object) ''
delay 2
smoutput (T__object,getoutput__object) ''
setinput__object 0:`''
smoutput (T__object,getoutput__object) ''
delay 0.5
smoutput (T__object,getoutput__object) ''
Task example (output):
0.001 0 2.002 4.71237e_6 2.004 1.25663e_5 2.504 1.25663e_5
First column is time relative to start of processing, second column is object's output at that time.
Using a task thread
Variant using an independent task thread:
delay=: 6!:3
task=: {{
obj=. '' conew 'integra'
F__obj=: 1 o. o.
delay 2
F__obj=: 0:
delay 0.5
s=. S__obj
destroy__obj''
s
}}
coclass'integra'
reqthreads=: {{ 0&T.@''^:(0>.y-1 T.'')0 }}
time=: 6!:1
F=: 0:
K=: S=: SHUTDOWN=: 0
create=: {{
reqthreads cores=. {.8 T. ''
integrator t. '' T=: time''
}}
destroy=: {{ codestroy '' [ SHUTDOWN=: 1 }}
integrator=: {{
while. -.SHUTDOWN do.
t=. time''
k=. F t
S=: S + (k+K)*t-T
T=: t
K=: k
end.
}}
This exhibits more timing variance because of the loose coupling of scheduling between threads:
task''
0.0194745
task''
_4.40316e_15
task''
0.00874017
task''
_0.0159841
Java
/**
* Integrates input function K over time
* S + (t1 - t0) * (K(t1) + K(t0)) / 2
*/
public class Integrator {
public interface Function {
double apply(double timeSinceStartInSeconds);
}
private final long start;
private volatile boolean running;
private Function func;
private double t0;
private double v0;
private double sum;
public Integrator(Function func) {
this.start = System.nanoTime();
setFunc(func);
new Thread(this::integrate).start();
}
public void setFunc(Function func) {
this.func = func;
v0 = func.apply(0.0);
t0 = 0;
}
public double getOutput() {
return sum;
}
public void stop() {
running = false;
}
private void integrate() {
running = true;
while (running) {
try {
Thread.sleep(1);
update();
} catch (InterruptedException e) {
return;
}
}
}
private void update() {
double t1 = (System.nanoTime() - start) / 1.0e9;
double v1 = func.apply(t1);
double rect = (t1 - t0) * (v0 + v1) / 2;
this.sum += rect;
t0 = t1;
v0 = v1;
}
public static void main(String[] args) throws InterruptedException {
Integrator integrator = new Integrator(t -> Math.sin(Math.PI * t));
Thread.sleep(2000);
integrator.setFunc(t -> 0.0);
Thread.sleep(500);
integrator.stop();
System.out.println(integrator.getOutput());
}
}
Output:
4.783602720556498E-13
JavaScript
function Integrator(sampleIntervalMS) {
var inputF = function () { return 0.0 };
var sum = 0.0;
var t1 = new Date().getTime();
var input1 = inputF(t1 / 1000);
function update() {
var t2 = new Date().getTime();
var input2 = inputF(t2 / 1000);
var dt = (t2 - t1) / 1000;
sum += (input1 + input2) * dt / 2;
t1 = t2;
input1 = input2;
}
var updater = setInterval(update, sampleIntervalMS);
return ({
input: function (newF) { inputF = newF },
output: function () { return sum },
shutdown: function () { clearInterval(updater) },
});
}
Test program as a HTML fragment:
<p><span id="a">Test running...</span> <code id="b">-</code></p>
<script type="text/javascript">
var f = 0.5;
var i = new Integrator(1);
var displayer = setInterval(function () { document.getElementById("b").firstChild.data = i.output() }, 100)
setTimeout(function () {
i.input(function (t) { return Math.sin(2*Math.PI*f*t) }); // test step 1
setTimeout(function () { // test step 2
i.input(function (t) { return 0 }); // test step 3
setTimeout(function () { // test step 3
i.shutdown();
clearInterval(displayer);
document.getElementById("a").firstChild.data = "Done, should be about 0: "
}, 500);
}, 2000);
}, 1)
</script>
Julia
Julia has inheritance of data structures and first-class types, but structures do not have methods. Instead, methods are functions with multiple dispatch based on argument type.
mutable struct Integrator
func::Function
runningsum::Float64
dt::Float64
running::Bool
function Integrator(f::Function, dt::Float64)
this = new()
this.func = f
this.runningsum = 0.0
this.dt = dt
this.running = false
return this
end
end
function run(integ::Integrator, lastval::Float64 = 0.0)
lasttime = time()
while integ.running
sleep(integ.dt)
newtime = time()
measuredinterval = newtime - lasttime
newval = integ.func(measuredinterval)
integ.runningsum += (lastval + newval) * measuredinterval / 2.0
lasttime = newtime
lastval = newval
end
end
start!(integ::Integrator) = (integ.running = true; @async run(integ))
stop!(integ) = (integ.running = false)
f1(t) = sin(2π * t)
f2(t) = 0.0
it = Integrator(f1, 0.00001)
start!(it)
sleep(2.0)
it.func = f2
sleep(0.5)
v2 = it.runningsum
println("After 2.5 seconds, integrator value was $v2")
Kotlin
Athough this is a faithful translation of the Java entry, on my machine the output of the latter is typically an order of magnitude smaller than this version. I have no idea why.
// version 1.2.0
import kotlin.math.*
typealias Function = (Double) -> Double
/**
* Integrates input function K over time
* S + (t1 - t0) * (K(t1) + K(t0)) / 2
*/
class Integrator {
private val start: Long
private @Volatile var running = false
private lateinit var func: Function
private var t0 = 0.0
private var v0 = 0.0
private var sum = 0.0
constructor(func: Function) {
start = System.nanoTime()
setFunc(func)
Thread(this::integrate).start()
}
fun setFunc(func: Function) {
this.func = func
v0 = func(0.0)
t0 = 0.0
}
fun getOutput() = sum
fun stop() {
running = false
}
private fun integrate() {
running = true
while (running) {
try {
Thread.sleep(1)
update()
}
catch(e: InterruptedException) {
return
}
}
}
private fun update() {
val t1 = (System.nanoTime() - start) / 1.0e9
val v1 = func(t1)
val rect = (t1 - t0) * (v0 + v1) / 2.0
sum += rect
t0 = t1
v0 = v1
}
}
fun main(args: Array<String>) {
val integrator = Integrator( { sin(PI * it) } )
Thread.sleep(2000)
integrator.setFunc( { 0.0 } )
Thread.sleep(500)
integrator.stop()
println(integrator.getOutput())
}
Sample output:
2.884266305153741E-4
Lingo
Parent script "Integrator":
property _sum
property _func
property _timeLast
property _valueLast
property _ms0
property _updateTimer
on new (me, func)
if voidP(func) then func = "0.0"
me._sum = 0.0
-- update frequency: 100/sec (arbitrary)
me._updateTimer = timeout().new("update", 10, #_update, me)
me.input(func)
return me
end
on stop (me)
me._updateTimer.period = 0 -- deactivates timer
end
-- func is a term (as string) that might contain "t" and is evaluated at runtime
on input (me, func)
me._func = func
me._ms0 = _system.milliseconds
me._timeLast = 0.0
t = 0.0
me._valueLast = value(me._func)
end
on output (me)
return me._sum
end
on _update (me)
now = _system.milliseconds - me._ms0
t = now/1000.0
val = value(me._func)
me._sum = me._sum + (me._valueLast+val)*(t - me._timeLast)/2
me._timeLast = t
me._valueLast = val
end
In some movie script:
global gIntegrator
-- entry point
on startMovie
gIntegrator = script("Integrator").new("sin(PI * t)")
timeout().new("timer", 2000, #step1)
end
on step1 (_, timer)
gIntegrator.input("0.0")
timer.timeoutHandler = #step2
timer.period = 500
end
on step2 (_, timer)
gIntegrator.stop()
put gIntegrator.output()
timer.forget()
end
- Output:
-- 0.0004
Lua
Pure/native Lua is not multithreaded, so this task should perhaps be marked "omit from|Lua" if following the implicit intent of the task. However, the explicit wording of the task does not seem to require a multithreaded solution. Perhaps this is cheating, but I thought it might interest the reader to see the integrator portion nonetheless, so it is demonstrated using a mock sampling method at various intervals (to simulate multithreaded updates).
local seconds = os.clock
local integrator = {
new = function(self, fn)
return setmetatable({fn=fn,t0=seconds(),v0=0,sum=0,nup=0},self)
end,
update = function(self)
self.t1 = seconds()
self.v1 = self.fn(self.t1)
self.sum = self.sum + (self.v0 + self.v1) * (self.t1 - self.t0) / 2
self.t0, self.v0, self.nup = self.t1, self.v1, self.nup+1
end,
input = function(self, fn) self.fn = fn end,
output = function(self) return self.sum end,
}
integrator.__index = integrator
-- "fake multithreaded sleep()"
-- waits for "duration" seconds calling "f" at every "interval" seconds
local function sample(duration, interval, f)
local now = seconds()
local untilwhen, nextinterval = now+duration, now+interval
f()
repeat
if seconds() >= nextinterval then f() nextinterval=nextinterval+interval end
until seconds() >= untilwhen
end
local pi, sin = math.pi, math.sin
local ks = function(t) return sin(2.0*pi*0.5*t) end
local kz = function(t) return 0 end
local intervals = { 0.5, 0.25, 0.1, 0.05, 0.025, 0.01, 0.005, 0.0025, 0.001 }
for _,interval in ipairs(intervals) do
local i = integrator:new(ks)
sample(2.0, interval, function() i:update() end)
i:input(kz)
sample(0.5, interval, function() i:update() end)
print(string.format("sampling interval: %f, %5d updates over 2.5s total = %.15f", interval, i.nup, i:output()))
end
- Output:
sampling interval: 0.500000, 6 updates over 2.5s total = -0.003628054395752 sampling interval: 0.250000, 11 updates over 2.5s total = 0.003994231784540 sampling interval: 0.100000, 25 updates over 2.5s total = 0.001891527454886 sampling interval: 0.050000, 51 updates over 2.5s total = 0.023521980508657 sampling interval: 0.025000, 101 updates over 2.5s total = -0.000573259909112 sampling interval: 0.010000, 250 updates over 2.5s total = 0.003001745575344 sampling interval: 0.005000, 501 updates over 2.5s total = -0.000415541052666 sampling interval: 0.002500, 999 updates over 2.5s total = -0.001480800340644 sampling interval: 0.001000, 2493 updates over 2.5s total = 0.000362576907805
Mathematica /Wolfram Language
Block[{start = SessionTime[], K, t0 = 0, t1, kt0, S = 0},
K[t_] = Sin[2 Pi f t] /. f -> 0.5; kt0 = K[t0];
While[True, t1 = SessionTime[] - start;
S += (kt0 + (kt0 = K[t1])) (t1 - t0)/2; t0 = t1;
If[t1 > 2, K[t_] = 0; If[t1 > 2.5, Break[]]]]; S]
1.1309*10^-6
Curiously, this value never changes; it is always exactly the same (at 1.1309E-6). Note that closer answers could be achieved by using Mathematica's better interpolation methods, but it would require collecting the data (in a list), which would have a speed penalty large enough to negate the improved estimation.
Nim
In Nim, objects managed by the garbage collector are allocated in one heap per thread. In order to share an object, the active object, one solution consists to manage it manually and to create it in a shared heap.
Of course, it is necessary to take some precautions when accessing or updating the shared object. We use a lock for this purpose.
# Active object.
# Compile with "nim c --threads:on".
import locks
import os
import std/monotimes
type
# Function to use for integration.
TimeFunction = proc (t: float): float {.gcsafe.}
# Integrator object.
Integrator = ptr TIntegrator
TIntegrator = object
k: TimeFunction # The function to integrate.
dt: int # Time interval in milliseconds.
thread: Thread[Integrator] # Thread which does the computation.
s: float # Computed value.
lock: Lock # Lock to manage concurrent accesses.
isRunning: bool # True if integrator is running.
#---------------------------------------------------------------------------------------------------
proc newIntegrator(f: TimeFunction; dt: int): Integrator =
## Create an integrator.
result = cast[Integrator](allocShared(sizeof(TIntegrator)))
result.k = f
result.dt = dt
result.s = 0
result.lock.initLock()
result.isRunning = false
#---------------------------------------------------------------------------------------------------
proc process(integrator: Integrator) {.thread, gcsafe.} =
## Do the integration.
integrator.isRunning = true
let start = getMonotime().ticks
var t0: float = 0
var k0 = integrator.k(0)
while true:
sleep(integrator.dt)
withLock integrator.lock:
if not integrator.isRunning:
break
let t1 = float(getMonoTime().ticks - start) / 1e9
let k1 = integrator.k(t1)
integrator.s += (k1 + k0) * (t1 - t0) / 2
t0 = t1
k0 = k1
#---------------------------------------------------------------------------------------------------
proc start(integrator: Integrator) =
## Start the integrator by launching a thread to do the computation.
integrator.thread.createThread(process, integrator)
#---------------------------------------------------------------------------------------------------
proc stop(integrator: Integrator) =
## Stop the integrator.
withLock integrator.lock:
integrator.isRunning = false
integrator.thread.joinThread()
#---------------------------------------------------------------------------------------------------
proc setInput(integrator: Integrator; f: TimeFunction) =
## Set the function.
withLock integrator.lock:
integrator.k = f
#---------------------------------------------------------------------------------------------------
proc output(integrator: Integrator): float =
## Return the current output.
withLock integrator.lock:
result = integrator.s
#---------------------------------------------------------------------------------------------------
proc destroy(integrator: Integrator) =
## Destroy an integrator, freing the resources.
if integrator.isRunning:
integrator.stop()
integrator.lock.deinitLock()
integrator.deallocShared()
#---------------------------------------------------------------------------------------------------
from math import PI, sin
# Create the integrator and start it.
let integrator = newIntegrator(proc (t: float): float {.gcsafe.} = sin(PI * t), 1)
integrator.start()
echo "Integrator started."
sleep(2000)
echo "Value after 2 seconds: ", integrator.output()
# Change the function to use.
integrator.setInput(proc (t: float): float {.gcsafe.} = 0)
echo "K function changed."
sleep(500)
# Stop the integrator and display the computed value.
integrator.stop()
echo "Value after 0.5 more second: ", integrator.output()
integrator.destroy()
- Output:
Integrator started. Value after 2 seconds: 2.058071586661761e-06 K function changed. Value after 0.5 more second: -3.007318220146679e-09
ooRexx
Not totally certain this is a correct implementation since the value coming out is not close to zero. It does show all of the basics of multithreading and object synchronization though.
integrater = .integrater~new(.routines~sine) -- start the integrater function
call syssleep 2
integrater~input = .routines~zero -- update the integrater function
call syssleep .5
say integrater~output
integrater~stop -- terminate the updater thread
::class integrater
::method init
expose stopped start v last_v last_t k
use strict arg k
stopped = .false
start = .datetime~new -- initial time stamp
v = 0
last_v = 0
last_t = 0
self~input = k
self~start
-- spin off a new thread and start updating. Note, this method is unguarded
-- to allow other threads to make calls
::method start unguarded
expose stopped
reply -- this spins this method invocation off onto a new thread
do while \stopped
call sysSleep .1
self~update -- perform the update operation
end
-- turn off the thread. Since this is unguarded,
-- it can be called any time, any where
::method stop unguarded
expose stopped
stopped = .true
-- perform the update. Since this is a guarded method, the object
-- start is protected.
::method update
expose start v last_v t last_t k
numeric digits 20 -- give a lot of precision
current = .datetime~new
t = (current - start)~microseconds
new_v = k~call(t) -- call the input function
v += (last_v + new_v) * (t - last_t) / 2
last_t = t
last_v = new_v
say new value is v
-- a write-only attribute setter (this is GUARDED)
::attribute input SET
expose k last_t last_v
self~update -- update current values
use strict arg k -- update the call function to the provided value
last_t = 0
last_v = k~call(0) -- and update to the zero value
-- the output function...returns current calculated value
::attribute output GET
expose v
return v
::routine zero
return 0
::routine sine
use arg t
return rxcalcsin(rxcalcpi() * t)
::requires rxmath library
OxygenBasic
Built from scratch. The ringmaster orchestrates all the active-objects, keeping a list of each individual and its method call.
With a high precision timer the result is around -.0002
double MainTime
'===============
class RingMaster
'===============
'
indexbase 1
sys List[512] 'limit of 512 objects per ringmaster
sys max,acts
'
method Register(sys meth,obj) as sys
sys i
for i=1 to max step 2
if list[i]=0 then exit for 'vacant slot
next
if i>=max then max+=2
List[i]<=meth,obj
return i 'token for deregistration etc
end method
'
method Deregister(sys *i)
if i then List[i]<=0,0 : i=0
end method
'
method Clear()
max=0
end method
'
method Act() 'called by the timer
sys i,q
for i=1 to max step 2
q=List[i]
if q then
call q List[i+1] 'anon object
end if
next
acts++
end method
'
end class
'=================
class ActiveObject
'=================
'
double s,freq,t1,t2,v1,v2
sys nfun,acts,RingToken
RingMaster *Master
'
method fun0() as double
end method
'
method fun1() as double
return sin(2*pi()*freq*MainTime)
end method
'
method func() as double
select case nfun
case 0 : return fun0()
case 1 : return fun1()
end select
'error?
end method
'
method TimeBasedDuties()
t1=t2
v1=v2
t2=MainTime
v2=func
s=s+(v2+v1)*(t2-t1)*0.5 'add slice to integral
acts++
end method
'
method RegisterWith(RingMaster*r)
@Master=@r
if @Master then
RingToken=Master.register @TimeBasedDuties,@this
end if
end method
'
method Deregister()
if @Master then
Master.Deregister RingToken 'this is set to null
end if
end method
'
method Output() as double
return s
end method
'
method Input(double fr=0,fun=0)
if fr then freq=fr
nfun=fun
end method
method ClearIntegral()
s=0
end method
'
end class
'SETUP TIMING SYSTEM
'===================
extern library "kernel32.dll"
declare QueryPerformanceCounter (quad*c)
declare QueryPerformanceFrequency(quad*f)
declare Sleep(sys milliseconds)
end extern
'
quad scount,tcount,freq
QueryPerformanceFrequency freq
double tscale=1/freq
double t1,t2
QueryPerformanceCounter scount
macro PrecisionTime(time)
QueryPerformanceCounter tcount
time=(tcount-scount)*tscale
end macro
'====
'TEST
'====
double integral
double tevent1,tevent2
RingMaster Rudolpho
ActiveObject A
'
A.RegisterWith Rudolpho
A.input (fr=0.5, fun=1) 'start with the freqency function (1)
'
'SET EVENT TIMES
'===============
tEvent1=2.0 'seconds
tEvent2=2.5 'seconds
'
PrecisionTime t1 'mark initial time
MainTime=t1
'
'
'EVENT LOOP
'==========
'
do
PrecisionTime t2
MainTime=t2
if t2-t1>=0.020 'seconds interval
Rudolpho.Act 'service all active objects
t1=t2
end if
'
if tEvent1>=0 and MainTime>=tEvent1
A.input (fun=0) 'switch to null function (0)
tEvent1=-1 'disable this event from happening again
end if
if MainTime>=tEvent2
integral=A.output()
exit do 'end of session
end if
'
sleep 5 'hand control to OS for a while
end do
print str(integral,4)
Rudolpho.clear
Oz
declare
fun {Const X}
fun {$ _} X end
end
fun {Now}
{Int.toFloat {Property.get 'time.total'}} / 1000.0
end
class Integrator from Time.repeat
attr
k:{Const 0.0}
s:0.0
t1 k_t1
t2 k_t2
meth init(SampleIntervalMS)
t1 := {Now}
k_t1 := {@k @t1}
{self setRepAll(action:Update
delay:SampleIntervalMS)}
thread
{self go}
end
end
meth input(K)
k := K
end
meth output($)
@s
end
meth Update
t2 := {Now}
k_t2 := {@k @t2}
s := @s + (@k_t1 + @k_t2) * (@t2 - @t1) / 2.0
t1 := @t2
k_t1 := @k_t2
end
end
Pi = 3.14159265
F = 0.5
I = {New Integrator init(10)}
in
{I input(fun {$ T}
{Sin 2.0 * Pi * F * T}
end)}
{Delay 2000} %% ms
{I input({Const 0.0})}
{Delay 500} %% ms
{Show {I output($)}}
{I stop}
Perl
#!/usr/bin/perl
use strict;
use 5.10.0;
package Integrator;
use threads;
use threads::shared;
sub new {
my $cls = shift;
my $obj = bless { t => 0,
sum => 0,
ref $cls ? %$cls : (),
stop => 0,
tid => 0,
func => shift,
}, ref $cls || $cls;
share($obj->{sum});
share($obj->{stop});
$obj->{tid} = async {
my $upd = 0.1; # update every 0.1 second
while (!$obj->{stop}) {
{
my $f = $obj->{func};
my $t = $obj->{t};
$obj->{sum} += ($f->($t) + $f->($t + $upd))* $upd/ 2;
$obj->{t} += $upd;
}
select(undef, undef, undef, $upd);
}
# say "stopping $obj";
};
$obj
}
sub output { shift->{sum} }
sub delete {
my $obj = shift;
$obj->{stop} = 1;
$obj->{tid}->join;
}
sub setinput {
# This is surprisingly difficult because of the perl sharing model.
# Func refs can't be shared, thus can't be replaced by another thread.
# Have to create a whole new object... there must be a better way.
my $obj = shift;
$obj->delete;
$obj->new(shift);
}
package main;
my $x = Integrator->new(sub { sin(atan2(1, 1) * 8 * .5 * shift) });
sleep(2);
say "sin after 2 seconds: ", $x->output;
$x = $x->setinput(sub {0});
select(undef, undef, undef, .5);
say "0 after .5 seconds: ", $x->output;
$x->delete;
Phix
classes
requires("0.8.2") integer xlock = init_cs() class integrator -- -- Integrates input function f over time -- v + (t1 - t0) * (f(t1) + f(t0)) / 2 -- integer f -- function f(atom t); (see note) atom interval, t0, k0 = 0, v = 0 bool running public integer id procedure set_func(integer rid) this.f = rid end procedure procedure update() enter_cs(xlock) integer f = this.f -- (nb: no "this") atom t1 = time(), k1 = f(t1) v += (t1 - t0) * (k1 + k0) / 2 t0 = t1 k0 = k1 leave_cs(xlock) end procedure procedure tick() running = true while running do sleep(interval) update() end while end procedure procedure stop() running = false wait_thread(id) end procedure function get_output() return v end function end class function new_integrator(integer rid, atom interval) integrator i = new({rid,interval,time()}) i.update() i.id = create_thread(i.tick,{i}) return i end function function zero(atom /*t*/) return 0 end function function sine(atom t) return sin(2*PI*0.5*t) end function integrator i = new_integrator(sine,0.01); sleep(2) ?i.get_output() i.set_func(zero) sleep(0.5) i.stop() ?i.get_output()
Note that were f a regular member function of the class, it would get a "this" parameter/argument, which we avoid by stashing it in a local integer prior to the call. Alternatively you could of course use zero/sine functions with an ignored parameter and the usual this.f() syntax [along with the usual "this." being optional inside the class definition].
- Output:
0.0003532983803 4.049495114e-17
pre-classes
Note that in Phix you cannot pass a variable to another procedure and have it "change under your feet". [erm, now you can, see classes above]
The copy-on-write semantics mean it would not have any effect, in that the original would be preserved
(deemed in phix to be a "very good thing") while the value passed along, a shared reference until it gets
modified and a copy made, would most likely simply be discarded, unless explicitly returned and stored,
which obviously cannot be done from a separate thread.
Instead we pass around an index (dx) as a way of emulating the "pointer references" of other languages.
If anything phix requires more locking that other languages due to the hidden shared reference counts.
Just lock everything, it is not that hard, and you should never need much more than the stuff below.
sequence x = {} enum TERMINATE, INTERVAL, KFUN, VALUE, T0, K0, ID, ISIZE=$ integer xlock = init_cs() function zero(atom /*t*/) return 0 end function function sine(atom t) return sin(2*PI*0.5*t) end function procedure update(integer dx) enter_cs(xlock) atom t1 = time(), k1 = call_func(x[dx][KFUN],{t1}) x[dx][VALUE] += (k1 + x[dx][K0]) * (t1 - x[dx][T0]) / 2 x[dx][T0] = t1 x[dx][K0] = k1 leave_cs(xlock) end procedure procedure tick(integer dx) while not x[dx][TERMINATE] do sleep(x[dx][INTERVAL]) update(dx) end while end procedure function new_integrator(integer rid, atom interval) x = append(x,repeat(0,ISIZE)) integer dx = length(x) x[dx][TERMINATE] = false x[dx][INTERVAL] = interval x[dx][KFUN] = rid x[dx][T0] = time() update(dx) x[dx][ID] = create_thread(tick,{dx}) return dx end function procedure set_input(integer dx, rid) enter_cs(xlock) x[dx][KFUN] = rid x[dx][K0] = 0 leave_cs(xlock) end procedure function get_output(integer dx) enter_cs(xlock) atom v = x[dx][VALUE] leave_cs(xlock) return v end function procedure stop_integrator(integer dx) x[dx][TERMINATE] = true wait_thread(x[dx][ID]) end procedure puts(1,"") integer dx = new_integrator(sine,0.01) sleep(2) printf(1,"%f\n",get_output(dx)) set_input(dx,zero) sleep(0.5) printf(1,"%f\n",get_output(dx)) stop_integrator(dx)
- Output:
-0.00326521 0.00196980
PicoLisp
(load "@lib/math.l")
(class +Active)
# inp val sum usec
(dm T ()
(unless (assoc -100 *Run) # Install timer task
(task -100 100 # Update objects every 0.1 sec
(mapc 'update> *Actives) ) )
(=: inp '((U) 0)) # Set zero input function
(=: val 0) # Initialize last value
(=: sum 0) # Initialize sum
(=: usec (usec)) # and time
(push '*Actives This) ) # Install in notification list
(dm input> (Fun)
(=: inp Fun) )
(dm update> ()
(let (U (usec) V ((: inp) U)) # Get current time, calculate value
(inc (:: sum)
(*/
(+ V (: val)) # (K(t[1]) + K(t[0])) *
(- U (: usec)) # (t[1] - t[0]) /
2.0 ) ) # 2.0
(=: val V)
(=: usec U) ) )
(dm output> ()
(format (: sum) *Scl) ) # Get result
(dm stop> ()
(unless (del This '*Actives) # Removing the last active object?
(task -100) ) ) # Yes: Uninstall timer task
(de integrate () # Test it
(let Obj (new '(+Active)) # Create an active object
(input> Obj # Set input function
'((U) (sin (*/ pi U 1.0))) ) # to sin(π * t)
(wait 2000) # Wait 2 sec
(input> Obj '((U) 0)) # Reset input function
(wait 500) # Wait 0.5 sec
(prinl "Output: " (output> Obj)) # Print return value
(stop> Obj) ) ) # Stop active object
PureBasic
Using the open-source precompiler SimpleOOP.
Prototype.d ValueFunction(f.d, t.d)
Class IntegralClass
Time0.i
Mutex.i
S.d
Freq.d
Thread.i
Quit.i
*func.ValueFunction
Protect Method Sampler()
Repeat
Delay(1)
If This\func And This\Mutex
LockMutex(This\Mutex)
This\S + This\func(This\Freq, ElapsedMilliseconds()-This\Time0)
UnlockMutex(This\Mutex)
EndIf
Until This\Quit
EndMethod
BeginPublic
Method Input(*func.ValueFunction)
LockMutex(This\Mutex)
This\func = *func
UnlockMutex(This\Mutex)
EndMethod
Method.d Output()
Protected Result.d
LockMutex(This\Mutex)
Result = This\S
UnlockMutex(This\Mutex)
MethodReturn Result
EndMethod
Method Init(F.d, *f)
This\Freq = F
This\func = *f
This\Mutex = CreateMutex()
This\Time0 = ElapsedMilliseconds()
This\Thread = CreateThread(This\Sampler, This)
ThreadPriority(This\Thread, 10)
EndMethod
Method Release()
This\Quit = #True
WaitThread(This\Thread)
EndMethod
EndPublic
EndClass
;- Procedures for generating values
Procedure.d n(f.d, t.d)
; Returns nothing
EndProcedure
Procedure.d f(f.d, t.d)
; Returns the function of this task
ProcedureReturn Sin(2*#PI*f*t)
EndProcedure
;- Test Code
*a.IntegralClass = NewObject.IntegralClass(0.5, @n()) ; Create the AO
*a\Input(@f()) ; Start sampling function f()
Delay(2000) ; Delay 2 sec
*a\Input(@n()) ; Change to sampling 'nothing'
Delay( 500) ; Wait 1/2 sec
MessageRequester("Info", StrD(*a\Output())) ; Present the result
*a= FreeObject
Python
Assignment is thread-safe in Python, so no extra locks are needed in this case.
from time import time, sleep
from threading import Thread
class Integrator(Thread):
'continuously integrate a function `K`, at each `interval` seconds'
def __init__(self, K=lambda t:0, interval=1e-4):
Thread.__init__(self)
self.interval = interval
self.K = K
self.S = 0.0
self.__run = True
self.start()
def run(self):
"entry point for the thread"
interval = self.interval
start = time()
t0, k0 = 0, self.K(0)
while self.__run:
sleep(interval)
t1 = time() - start
k1 = self.K(t1)
self.S += (k1 + k0)*(t1 - t0)/2.0
t0, k0 = t1, k1
def join(self):
self.__run = False
Thread.join(self)
if __name__ == "__main__":
from math import sin, pi
ai = Integrator(lambda t: sin(pi*t))
sleep(2)
print(ai.S)
ai.K = lambda t: 0
sleep(0.5)
print(ai.S)
Racket
#lang racket
(require (only-in racket/gui sleep/yield timer%))
(define active%
(class object%
(super-new)
(init-field k) ; input function
(field [s 0]) ; state
(define t_0 0)
(define/public (input new-k) (set! k new-k))
(define/public (output) s)
(define (callback)
(define t_1 (/ (- (current-inexact-milliseconds) start) 1000))
(set! s (+ s (* (+ (k t_0) (k t_1))
(/ (- t_1 t_0) 2))))
(set! t_0 t_1))
(define start (current-inexact-milliseconds))
(new timer%
[interval 1000]
[notify-callback callback])))
(define active (new active% [k (λ (t) (sin (* 2 pi 0.5 t)))]))
(sleep/yield 2)
(send active input (λ _ 0))
(sleep/yield 0.5)
(displayln (send active output))
Raku
(formerly Perl 6)
There is some jitter in the timer, but it is typically accurate to within a few thousandths of a second.
class Integrator {
has $.f is rw = sub ($t) { 0 };
has $.now is rw;
has $.value is rw = 0;
has $.integrator is rw;
method init() {
self.value = &(self.f)(0);
self.integrator = Thread.new(
:code({
loop {
my $t1 = now;
self.value += (&(self.f)(self.now) + &(self.f)($t1)) * ($t1 - self.now) / 2;
self.now = $t1;
sleep .001;
}
}),
:app_lifetime(True)
).run
}
method Input (&f-of-t) {
self.f = &f-of-t;
self.now = now;
self.init;
}
method Output { self.value }
}
my $a = Integrator.new;
$a.Input( sub ($t) { sin(2 * π * .5 * $t) } );
say "Initial value: ", $a.Output;
sleep 2;
say "After 2 seconds: ", $a.Output;
$a.Input( sub ($t) { 0 } );
sleep .5;
say "f(0): ", $a.Output;
- Typical output:
Initial value: 0 After 2 seconds: -0.0005555887464620366 f(0): 0
Rust
#![feature(mpsc_select)]
extern crate num;
extern crate schedule_recv;
use num::traits::Zero;
use num::Float;
use schedule_recv::periodic_ms;
use std::f64::consts::PI;
use std::ops::Mul;
use std::sync::mpsc::{self, SendError, Sender};
use std::sync::{Arc, Mutex};
use std::thread;
use std::time::Duration;
pub type Actor<S> = Sender<Box<Fn(u32) -> S + Send>>;
pub type ActorResult<S> = Result<(), SendError<Box<Fn(u32) -> S + Send>>>;
/// Rust supports both shared-memory and actor models of concurrency, and the `Integrator` utilizes
/// both. We use an `Actor` to send the `Integrator` new functions, while we use a `Mutex`
/// (shared-memory concurrency) to hold the result of the integration.
///
/// Note that these are not the only options here--there are many, many ways you can deal with
/// concurrent access. But when in doubt, a plain old `Mutex` is often a good bet. For example,
/// this might look like a good situation for a `RwLock`--after all, there's no reason for a read
/// in the main task to block writes. Unfortunately, unless you have significantly more reads than
/// writes (which is certainly not the case here), a `Mutex` will usually outperform a `RwLock`.
pub struct Integrator<S: 'static, T: Send> {
input: Actor<S>,
output: Arc<Mutex<T>>,
}
/// In Rust, time durations are strongly typed. This is usually exactly what you want, but for a
/// problem like this--where the integrated value has unusual (unspecified?) units--it can actually
/// be a bit tricky. Right now, `Duration`s can only be multiplied or divided by `i32`s, so in
/// order to be able to actually do math with them we say that the type parameter `S` (the result
/// of the function being integrated) must yield `T` (the type of the integrated value) when
/// multiplied by `f64`. We could possibly replace `f64` with a generic as well, but it would make
/// things a bit more complex.
impl<S, T> Integrator<S, T>
where
S: Mul<f64, Output = T> + Float + Zero,
T: 'static + Clone + Send + Float,
{
pub fn new(frequency: u32) -> Integrator<S, T> {
// We create a pipe allowing functions to be sent from tx (the sending end) to input (the
// receiving end). In order to change the function we are integrating from the task in
// which the Integrator lives, we simply send the function through tx.
let (tx, input) = mpsc::channel();
// The easiest way to do shared-memory concurrency in Rust is to use atomic reference
// counting, or Arc, around a synchronized type (like Mutex<T>). Arc gives you a guarantee
// that memory will not be freed as long as there is at least one reference to it.
// It is similar to C++'s shared_ptr, but it is guaranteed to be safe and is never
// incremented unless explicitly cloned (by default, it is moved).
let s: Arc<Mutex<T>> = Arc::new(Mutex::new(Zero::zero()));
let integrator = Integrator {
input: tx,
// Here is the aforementioned clone. We have to do it before s enters the closure,
// because once that happens it is moved into the closure (and later, the new task) and
// becomes inaccessible to the outside world.
output: Arc::clone(&s),
};
thread::spawn(move || -> () {
// The frequency is how often we want to "tick" as we update our integrated total. In
// Rust, timers can yield Receivers that are periodically notified with an empty
// message (where the period is the frequency). This is useful because it lets us wait
// on either a tick or another type of message (in this case, a request to change the
// function we are integrating).
let periodic = periodic_ms(frequency);
let mut t = 0;
let mut k: Box<Fn(u32) -> S + Send> = Box::new(|_| Zero::zero());
let mut k_0: S = Zero::zero();
loop {
// Here's the selection we talked about above. Note that we are careful to call
// the *non*-failing function, recv(), here. The reason we do this is because
// recv() will return Err when the sending end of a channel is dropped. While
// this is unlikely to happen for the timer (so again, you could argue for failure
// there), it's normal behavior for the sending end of input to be dropped, since
// it just happens when the Integrator falls out of scope. So we handle it cleanly
// and break out of the loop, rather than failing.
select! {
res = periodic.recv() => match res {
Ok(_) => {
t += frequency;
let k_1: S = k(t);
// Rust Mutexes are a bit different from Mutexes in many other
// languages, in that the protected data is actually encapsulated by
// the Mutex. The reason for this is that Rust is actually capable of
// enforcing (via its borrow checker) the invariant that the contents
// of a Mutex may only be read when you have acquired its lock. This
// is enforced by way of a MutexGuard, the return value of lock(),
// which implements some special traits (Deref and DerefMut) that allow
// access to the inner element "through" the guard. The element so
// acquired has a lifetime bounded by that of the MutexGuard, the
// MutexGuard can only be acquired by taking a lock, and the only way
// to release the lock is by letting the MutexGuard fall out of scope,
// so it's impossible to access the data incorrectly. There are some
// additional subtleties around the actual implementation, but that's
// the basic idea.
let mut s = s.lock().unwrap();
*s = *s + (k_1 + k_0) * (f64::from(frequency) / 2.);
k_0 = k_1;
}
Err(_) => break,
},
res = input.recv() => match res {
Ok(k_new) => k = k_new,
Err(_) => break,
}
}
}
});
integrator
}
pub fn input(&self, k: Box<Fn(u32) -> S + Send>) -> ActorResult<S> {
// The meat of the work is done in the other thread, so to set the
// input we just send along the Sender we set earlier...
self.input.send(k)
}
pub fn output(&self) -> T {
// ...and to read the input, we simply acquire a lock on the output Mutex and return a
// copy. Why do we have to copy it? Because, as mentioned above, Rust won't let us
// retain access to the interior of the Mutex unless we have possession of its lock. There
// are ways and circumstances in which one can avoid this (e.g. by using atomic types) but
// a copy is a perfectly reasonable solution as well, and a lot easier to reason about :)
*self.output.lock().unwrap()
}
}
/// This function is fairly straightforward. We create the integrator, set its input function k(t)
/// to 2pi * f * t, and then wait as described in the Rosetta stone problem.
fn integrate() -> f64 {
let object = Integrator::new(10);
object
.input(Box::new(|t: u32| {
let two_seconds_ms = 2 * 1000;
let f = 1. / f64::from(two_seconds_ms);
(2. * PI * f * f64::from(t)).sin()
}))
.expect("Failed to set input");
thread::sleep(Duration::from_secs(2));
object.input(Box::new(|_| 0.)).expect("Failed to set input");
thread::sleep(Duration::from_millis(500));
object.output()
}
fn main() {
println!("{}", integrate());
}
/// Will fail on a heavily loaded machine
#[test]
#[ignore]
fn solution() {
// We should just be able to call integrate, but can't represent the closure properly due to
// rust-lang/rust issue #17060 if we make frequency or period a variable.
// FIXME(pythonesque): When unboxed closures are fixed, fix integrate() to take two arguments.
let object = Integrator::new(10);
object
.input(Box::new(|t: u32| {
let two_seconds_ms = 2 * 1000;
let f = 1. / (two_seconds_ms / 10) as f64;
(2. * PI * f * t as f64).sin()
}))
.expect("Failed to set input");
thread::sleep(Duration::from_millis(200));
object.input(Box::new(|_| 0.)).expect("Failed to set input");
thread::sleep(Duration::from_millis(100));
assert_eq!(object.output() as u32, 0)
}
Scala
object ActiveObject {
class Integrator {
import java.util._
import scala.actors.Actor._
case class Pulse(t: Double)
case class Input(k: Double => Double)
case object Output
case object Bye
val timer = new Timer(true)
var k: Double => Double = (_ => 0.0)
var s: Double = 0.0
var t0: Double = 0.0
val handler = actor {
loop {
react {
case Pulse(t1) => s += (k(t1) + k(t0)) * (t1 - t0) / 2.0; t0 = t1
case Input(k) => this.k = k
case Output => reply(s)
case Bye => timer.cancel; exit
}
}
}
timer.scheduleAtFixedRate(new TimerTask {
val start = System.currentTimeMillis
def run { handler ! Pulse((System.currentTimeMillis - start) / 1000.0) }
}, 0, 10) // send Pulse every 10 ms
def input(k: Double => Double) = handler ! Input(k)
def output = handler !? Output
def bye = handler ! Bye
}
def main(args: Array[String]) {
val integrator = new Integrator
integrator.input(t => Math.sin(2.0 * Math.Pi * 0.5 * t))
Thread.sleep(2000)
integrator.input(_ => 0.0)
Thread.sleep(500)
println(integrator.output)
integrator.bye
}
}
Smalltalk
Object subclass:#Integrator
instanceVariableNames:'tickRate input s thread'
classVariableNames:''
poolDictionaries:''
category:'Rosetta'
instance methods:
input:aFunctionOfT
input := aFunctionOfT.
startWithTickRate:r
"setup and start sampling"
tickRate := r.
s := 0.
thread := [ self integrateLoop ] fork.
stop
"stop and return the 'final' output"
thread terminate.
^ s
integrateLoop
"no need for any locks
- the assignment to s is atomic in Smallalk; its either done or not, when terminated, so who cares"
|tBegin tPrev tNow kPrev kNow deltaT delta|
tBegin := tPrev := Timestamp nowWithMilliseconds.
kPrev := input value:0.
[true] whileTrue:[
Delay waitForSeconds: tickRate.
tNow := Timestamp nowWithMilliseconds.
kNow := input value:(tNow millisecondDeltaFrom:tBegin) / 1000.
deltaT := (tNow millisecondDeltaFrom:tPrev) / 1000.
delta := (kPrev + kNow) * deltaT / 2.
s := s + delta.
tPrev := tNow. kPrev := kNow.
].
class methods:
example
#( 0.5 0.1 0.05 0.01 0.005 0.001 0.0005 ) do:[:sampleRate |
|i|
i := Integrator new.
i input:[:t | (2 * Float pi * 0.5 * t) sin].
i startWithTickRate:sampleRate.
Delay waitForSeconds:2.
i input:[:t | 0].
Delay waitForSeconds:0.5.
Transcript
show:'Sample rate: '; showCR:sampleRate;
showCR:(i stop).
].
running:
Integrator example
output:
Sample rate: 0.5 -0.0258202058271805 Sample rate: 0.1 -0.00519217893508676 Sample rate: 0.05 -0.000897807957672559 Sample rate: 0.01 -0.000650159409949159 Sample rate: 0.005 -0.00033633922519125 Sample rate: 0.001 0.000286557714782226 Sample rate: 0.0005 0.000253571129723327
for backward compatibility, the smalltalk used here returns only timestamps with second-precision from "Timestamp now". Therefore, the millisecond-precision variant was used here. An alternative would have been to ask the OS for its ticker, which is more precise.
SuperCollider
Instead of writing a class, here we just use an environment to encapsulate state.
(
a = TaskProxy { |envir|
envir.use {
~integral = 0;
~time = 0;
~prev = 0;
~running = true;
loop {
~val = ~input.(~time);
~integral = ~integral + (~val + ~prev * ~dt / 2);
~prev = ~val;
~time = ~time + ~dt;
~dt.wait;
}
}
};
)
// run the test
(
fork {
a.set(\dt, 0.0001);
a.set(\input, { |t| sin(2pi * 0.5 * t) });
a.play(quant: 0); // play immediately
2.wait;
a.set(\input, 0);
0.5.wait;
a.stop;
a.get(\integral).postln; // answers -7.0263424372343e-15
}
)
Swift
// For NSObject, NSTimeInterval and NSThread
import Foundation
// For PI and sin
import Darwin
class ActiveObject:NSObject {
let sampling = 0.1
var K: (t: NSTimeInterval) -> Double
var S: Double
var t0, t1: NSTimeInterval
var thread = NSThread()
func integrateK() {
t0 = t1
t1 += sampling
S += (K(t:t1) + K(t: t0)) * (t1 - t0) / 2
}
func updateObject() {
while true {
integrateK()
usleep(100000)
}
}
init(function: (NSTimeInterval) -> Double) {
S = 0
t0 = 0
t1 = 0
K = function
super.init()
thread = NSThread(target: self, selector: "updateObject", object: nil)
thread.start()
}
func Input(function: (NSTimeInterval) -> Double) {
K = function
}
func Output() -> Double {
return S
}
}
// main
func sine(t: NSTimeInterval) -> Double {
let f = 0.5
return sin(2 * M_PI * f * t)
}
var activeObject = ActiveObject(function: sine)
var date = NSDate()
sleep(2)
activeObject.Input({(t: NSTimeInterval) -> Double in return 0.0})
usleep(500000)
println(activeObject.Output())
Sample output:
1.35308431126191e-16
Tcl
or
This implementation Tcl 8.6 for object support (for the active integrator object) and coroutine support (for the controller task). It could be rewritten to only use 8.5 plus the TclOO library.
package require Tcl 8.6
oo::class create integrator {
variable e sum delay tBase t0 k0 aid
constructor {{interval 1}} {
set delay $interval
set tBase [clock microseconds]
set t0 0
set e { 0.0 }
set k0 0.0
set sum 0.0
set aid [after $delay [namespace code {my Step}]]
}
destructor {
after cancel $aid
}
method input expression {
set e $expression
}
method output {} {
return $sum
}
method Eval t {
expr $e
}
method Step {} {
set aid [after $delay [namespace code {my Step}]]
set t [expr {([clock microseconds] - $tBase) / 1e6}]
set k1 [my Eval $t]
set sum [expr {$sum + ($k1 + $k0) * ($t - $t0) / 2.}]
set t0 $t
set k0 $k1
}
}
set pi 3.14159265
proc pause {time} {
yield [after [expr {int($time * 1000)}] [info coroutine]]
}
proc task {script} {
coroutine task_ apply [list {} "$script;set ::done ok"]
vwait done
}
task {
integrator create i
i input {sin(2*$::pi * 0.5 * $t)}
pause 2
i input { 0.0 }
pause 0.5
puts [format %.15f [i output]]
}
Sample output:
-0.000000168952702
Visual Basic .NET
Since this object is CPU intensive, shutting it down when done is crucial. To facilitate this, the IDisposable pattern was used.
Module Module1
Sub Main()
Using active As New Integrator
active.Operation = Function(t As Double) Math.Sin(2 * Math.PI * 0.5 * t)
Threading.Thread.Sleep(TimeSpan.FromSeconds(2))
Console.WriteLine(active.Value)
active.Operation = Function(t As Double) 0
Threading.Thread.Sleep(TimeSpan.FromSeconds(0.5))
Console.WriteLine(active.Value)
End Using
Console.ReadLine()
End Sub
End Module
Class Integrator
Implements IDisposable
Private m_Operation As Func(Of Double, Double)
Private m_Disposed As Boolean
Private m_SyncRoot As New Object
Private m_Value As Double
Public Sub New()
m_Operation = Function(void) 0.0
Dim t As New Threading.Thread(AddressOf MainLoop)
t.Start()
End Sub
Private Sub MainLoop()
Dim epoch = Now
Dim t0 = 0.0
Do
SyncLock m_SyncRoot
Dim t1 = (Now - epoch).TotalSeconds
m_Value = m_Value + (Operation(t1) + Operation(t0)) * (t1 - t0) / 2
t0 = t1
End SyncLock
Threading.Thread.Sleep(10)
Loop Until m_Disposed
End Sub
Public Property Operation() As Func(Of Double, Double)
Get
SyncLock m_SyncRoot
Return m_Operation
End SyncLock
End Get
Set(ByVal value As Func(Of Double, Double))
SyncLock m_SyncRoot
m_Operation = value
End SyncLock
End Set
End Property
Public ReadOnly Property Value() As Double
Get
SyncLock m_SyncRoot
Return m_Value
End SyncLock
End Get
End Property
Protected Overridable Sub Dispose(ByVal disposing As Boolean)
m_Disposed = True
End Sub
Public Sub Dispose() Implements IDisposable.Dispose
Dispose(True)
GC.SuppressFinalize(Me)
End Sub
End Class
Output: 0.000241446762282308
Wren
Wren doesn't have threads but does have fibers which are cooperatively (rather than preemptively) scheduled. Only one fiber can run at a time.
However, it is possible to perform asynchronous operations in Wren-cli using a combination of the Scheduler and Timer classes which use the C library libuv under the hood. The problem is that, due to a bug, the Timer.sleep method doesn't just pause the current fiber, it also pauses the System.clock method! So we can't use the latter to measure time here.
What I've done instead is to pre-compute the number of updates performed on my machine for a given function and time period which is a fairly stable figure (though it will obviously be different on other machines). I've then used this figure to measure elapsed time for each update. On average this gives results of around 0.003 seconds which I consider acceptable in the circumstances.
import "scheduler" for Scheduler
import "timer" for Timer
var Interval = 0
class Integrator {
construct new() {
_sum = 0
}
input(k) {
_k = k
_v0 = k.call(0)
_t = 0
_running = true
integrate_()
}
output { _sum }
stop() {
_running = false
}
integrate_() {
while (_running) {
Timer.sleep(1)
update_()
}
}
update_() {
_t = _t + Interval
var v1 = _k.call(_t)
var trap = Interval * (_v0 + v1) / 2
_sum = _sum + trap
_v0 = v1
}
}
var integrator = Integrator.new()
Scheduler.add {
Interval = 2 / 1550 // machine specific value
integrator.input(Fn.new { |t| (Num.pi * t).sin })
}
Timer.sleep(2000)
Scheduler.add {
Interval = 0.5 / 775 // machine specific value
integrator.input(Fn.new { |t| 0 })
}
Timer.sleep(500)
integrator.stop()
System.print(integrator.output)
- Output:
0.0028437802254386
zkl
Uses cheese ball thread safety: since the integrator runs continuously and I don't want to queue the output, just sample it, strong references are used as they change atomically.
class Integrator{
// continuously integrate a function `K`, at each `interval` seconds'
fcn init(f,interval=1e-4){
var _interval=interval, K=Ref(f), S=Ref(0.0), run=True;
self.launch(); // start me as a thread
}
fcn liftoff{ // entry point for the thread
start:=Time.Clock.timef; // floating point seconds since Epoch
t0,k0,s:=0,K.value(0),S.value;
while(run){
Atomic.sleep(_interval);
t1,k1:=Time.Clock.timef - start, K.value(t1);
s+=(k1 + k0)*(t1 - t0)/2.0; S.set(s);
t0,k0=t1,k1;
}
}
fcn sample { S.value }
fcn setF(f) { K.set(f) }
}
ai:=Integrator(fcn(t){ ((0.0).pi*t).sin() });
Atomic.sleep(2);
ai.sample().println();
ai.setF(fcn{ 0 });
Atomic.sleep(0.5);
ai.sample().println();
- Output:
4.35857e-09 1.11571e-07
- Object oriented
- Programming Tasks
- Concurrency
- Ada
- ATS
- BASIC
- BBC BASIC
- C
- Pthread
- C sharp
- C++
- Clojure
- Common Lisp
- Bordeaux Threads
- Crystal
- D
- Delphi
- System.SysUtils
- System.Classes
- E
- EchoLisp
- Erlang
- F Sharp
- Factor
- FBSL
- FreeBASIC
- Go
- Groovy
- Haskell
- J
- Java
- JavaScript
- Julia
- Kotlin
- Lingo
- Lua
- Mathematica
- Wolfram Language
- Nim
- OoRexx
- OxygenBasic
- Oz
- Perl
- Phix
- Phix/Class
- PicoLisp
- PureBasic
- Python
- Racket
- Raku
- Rust
- Scala
- Smalltalk
- SuperCollider
- Swift
- Tcl
- TclOO
- Visual Basic .NET
- Wren
- Zkl
- 6502 Assembly/Omit
- 8080 Assembly/Omit
- 8086 Assembly/Omit
- 68000 Assembly/Omit
- ACL2/Omit
- ARM Assembly/Omit
- AWK/Omit
- Gnuplot/Omit
- GUISS/Omit
- LaTeX/Omit
- Locomotive Basic/Omit
- M4/Omit
- Make/Omit
- Maxima/Omit
- Metafont/Omit
- ML/I/Omit
- Octave/Omit
- PlainTeX/Omit
- Retro/Omit
- TI-89 BASIC/Omit
- UNIX Shell/Omit
- VBScript/Omit
- X86 Assembly/Omit
- Z80 Assembly/Omit
- ZX Spectrum Basic/Omit