Why would a function state "myVariable = myVariable;"? - variables

In the course of working with some as2 code from our dev team, I came across a baffling handful of statements where variables were set to themselves. Is there some reason for such a redundancy that I'm not thinking of?
I'm talking literally like this:
function timeLine(x,w){
x = x;
p = ((x) * 100) / w;
t = v.totalTime;
n = (t * p) / 100;
n = n;
What am I missing? (While we're at it, what's with (x)? I assume it used to be (x + z)...)

Why would a function state myVariable = myVariable;?
Because the programmer who wrote it was incompetent. You're not missing anything, whomever wrote that code originally didn't know what they were doing.
If that was AS3 code, it might be possible that they're setting a class variable from the function parameter, but that's best done explicitly using this to show that the variables are different:
function foo(bar, baz) {
this.bar = bar;
this.baz = baz;
}
Additionally AS3 has accessors and mutators that could cause side-effects. If this was the case, the programmer would not just be incompetent, but dangerous as well.

Related

Does MATLAB support "callable" (i.e. function-like) classes?

Is it possible to define a MATLAB class such that the objects from this class can be called like any other function?
IOW, I'm asking whether one can write in MATLAB the equivalent of something like the following Python class:
# define the class FxnClass
class FxnClass(object):
def __init__(self, template):
self.template = template
def __call__(self, x, y, z):
print self.template % locals()
# create an instance of FxnClass
f = FxnClass('x is %(x)r; y is %(y)r; z is %(z)r')
# call the instance of FxnClass
f(3, 'two', False)
...
[OUTPUT]
x is 3; y is 'two'; z is False
Thanks!
I do not know, whether MATLAB directly supports what you want, but MATLAB does support first-class functions; closures might therefore provide a useable substitute, for instance:
function f = count_call(msg)
calls = 0;
function current_count()
disp(strcat(msg, num2str(calls)));
calls = calls + 1;
end
f = #current_count;
end
In this case, current_count closes over calls (and msg). That way you can express functions that depend on some internal state. You would use it this way:
g = count_call('number of calls: ') % returns a new function ("__init__")
g() % "__call__"
I will be interested to see if this is possible without simply creating a java method in Matlab. I know you can do the following
classdef ExampleObject
properties
test;
end
methods
function exampleObject = ExampleObject(inputTest)
exampleObject.test=inputTest;
end
function f(exampleObject,funcInput)
disp(funcInput+exampleObject.test);
end
end
end
>> e=ExampleObject(5);
>> f(e,10)
15
But as far as my knowledge goes, if you tried to override the call function you'd run into a conflict with Matlab's parenthetical subscript reference subsref. You can find a reference here showing how to overwrite that, and you might be able to get it to do what you want...but it doesn't seem like good form to do so. Not sure how Matlab would handle a call to an object (as opposed to a function) without it getting confused with this.
One way is to override the feval function for your class:
classdef FxnClass < handle
properties
template
end
methods
function obj = FxnClass(t)
obj.template = t;
end
function out = feval(obj, varargin)
out = sprintf(obj.template, varargin{:});
end
end
end
This would be used as:
>> f = FxnClass('x = %f, y = %s, z = %d');
>> feval(f, 3,'two',false)
ans =
x = 3.000000, y = two, z = 0
Now if you want to provide additional syntactic sugar, you could redefine the subsref function for your class as #Salain suggested. Add the following to the previous class definition:
classdef FxnClass < handle
...
methods
function out = subsref(obj, S)
switch S(1).type
case '.'
% call builtin subsref, so we dont break the dot notation
out = builtin('subsref', obj, S);
case '()'
out = feval(obj, S.subs{:});
case '{}'
error('Not a supported subscripted reference');
end
end
end
end
Now you could simply write:
>> f = FxnClass('x = %f, y = %s, z = %d');
>> f(3,'two',false)
ans =
x = 3.000000, y = two, z = 0
Personally I don't particularly like overriding the subsref or subsasgn functions. They are used for too many cases, and its sometimes hard to get them write. For example all the following will eventually call the subsref method with different input:
f(..)
f.template
f.template(..)
f(..).template
f(..).template(..)
There is also the case of the end keyword which could appear in indexing, so you might have to also override it as well in some cases. Not to mention that objects can also be concatenated into arrays, which makes things even more complicated:
>> ff = [f,f];
>> ff(1) % not what you expect!
That said, I think #Frank's suggestion to use nested functions with closures is more elegant in this case:
function f = FxnClass(t)
f = #call;
function out = call(varargin)
out = sprintf(t, varargin{:});
end
end
which is called as before:
>> f = FxnClass('x = %f, y = %s, z = %d');
>> f(3, 'two', false)
If you mean that you want a class to hold a method which you use like a normal function (eg. defined in an m-file), then yes, Matlab does support static methods.
A static method runs independently of any instances of that class, in fact, you don't even need to instantiate a class to use its static methods.
Matlab does not support static fields, however, so you would have to instantiate such a class first, and then set its fields before using the functions (which presumably make use of these fields, since you are asking this question).
Given the limitation with static members, you might be better off with closures, as described by Frank.

Why can't you assign and call a block at the same time?

The following code compiles:
^{}();
And this compiles:
void (^x)();
(x = ^{})();
But this doesn't:
(void (^x)() = ^{})();
The error I get is Expected ')'. Is this a bug with llvm or something? It's totally holding me back from pretending Objective-C is JavaScript.
This wouldn't make sense in a C-like language. To see why, let's build the statement from the ground up.
First, we'll use your working declaration for x:
void (^x)();
Now let's initialize it in the same statement:
void (^x)() = ^{};
So far so good - x has been initialized with the correct block. So let's invoke x now. But where will the () go? Naturally, we need to place the () immediately after a block-valued expression. However, in C, declarations are statements, not expressions so
(void (^x)() = ^{})();
doesn't make sense. The only place the () can go is after the ^{}:
void (^x)() = ^{}();
But ^{}() has type void, not type void (^)().
To sum up: you can't declare a block variable and invoke it at the same time. you'll have to either declare and initialize the variable, and then call it
void (^x)() = ^{};
x();
or declare it and then assign and call it
void (^x)();
(x = ^{})();
or just separate all three:
void (^x)();
x = ^{};
x();
As a concluding thought, let's say it was desirable to declare and invoke blocks at the same time. If we decided to allow code like (void (^x)() = ^{})();, then for the sake of consistency, we would have to also allow code such as ++(void x = 4); or (void x = 1) + (void y = 2);. I hope you'll agree that these just look strange in C.
As an analogy, consider:
This compiles:
if (42) { }
And this compiles:
int x;
if (x = 42) { }
But this doesn't:
if (int x = 42) { }

The best way to implement piece-wise periodic function in Objective-C?

I need to implement kind of illustrated function ( ratio = somefunction(time)) in Objective-c. Language doesn't actually matters, because task seems purely algorithmic. Is there any common way to do things like this?
Next things should be easily adjustable in process of design:
1) Number of small intervals per period (now it 4, but it can be 3 or 20 for example).
2) f1 can be changed. It's simple function like f(x) = sin(x).
3) if we say
f_resulting =
f1 for a time t1
f2 for a time t2
then again
f1 for a time t1
etc..
ratio when f1 "works" vs f2 "works" (t1 to t2) should be adjustable.
Standard C itself has no closures so if you use C function pointers you need to build your own closure - you just use a class instance to store the two function pointers and duration and an invoke method to call the composed periodic function.
However Apple has introduced blocks, which are just inline functions implemented with closures, so in Apple C you can easily compose functions. This is more interesting, but "best" you will have to decide...
An example which composes double -> double functions. First the header, Periodic.h:
typedef double (^Monadic)(double);
Monadic makePeriodic(Monadic firstFunction, double firstPeriod,
Monadic secondFunction, double secondPeriod);
And the implementation, Periodic.c:
#include "Periodic.h"
Monadic makePeriodic(Monadic firstFunction, double firstPeriod,
Monadic secondFunction, double secondPeriod)
{
return Block_copy(^(double time)
{
return fmod(time, firstPeriod+secondPeriod) < firstPeriod
? firstFunction(time)
: secondFunction(time);
}
);
}
Blocks are constructed on the stack and can reference local variables and parameters in the enclosing scope. As such you cannot simply return a block from a function as the scope it refers to disappears on return. The function Block_copy() moves a block, and any blocks it refers to, onto the heap allowing them to outlive their creating scope. A heap block must be released with Block_release() when no longer needed.
And a simple demo:
Monadic queer = makePeriodic(^(double t) { return t * t; }, 5,
^(double t) { return sqrt(t); }, 3);
for(double ix = 0; ix <= 16; ix++)
printf("%f -> %f\n", ix, queer(ix));
Block_release(queer); // clean up
Now you can wrap all this up in a class if you wish so it fits Obj-C style. When you do this you can also send copy and release messages to the block, just as if it was an Obj-C object. In a garbage collected environment you still need the copy to get thge block onto the heap, but the release is not needed.
Periodic.h:
typedef double (^Monadic)(double);
#interface ComposeOne : NSObject
{
}
+ (Monadic) makePeriodicWithFunction:(Monadic)firstFunction
forPeriod:(double)firstPeriod
andFunction:(Monadic)secondFunction
forPeriod:(double)secondPeriod;
#end
Periodic.M:
#implementation ComposeOne
+ (Monadic) makePeriodicWithFunction:(Monadic)firstFunction
forPeriod:(double)firstPeriod
andFunction:(Monadic)secondFunction
forPeriod:(double)secondPeriod
{
Monadic result = (^(double time)
{
return fmod(time, firstPeriod+secondPeriod) < firstPeriod
? firstFunction(time)
: secondFunction(time);
}
);
return [result copy];
}
#end
And the simple demo (still using printf though for convenience):
Monadic queer = [ComposeOne makePeriodicWithFunction:^(double t) { return t * t; }
forPeriod:5
andFunction:^(double t) { return sqrt(t); }
forPeriod:3];
for(double ix = 0; ix <= 16; ix++)
printf("%f -> %f\n", ix, queer(ix));
[queer release];

Dynamic/Static scope with Deep/Shallow binding (exercises)

I'm studying dynamic/static scope with deep/shallow binding and running code manually to see how these different scopes/bindings actually work. I read the theory and googled some example exercises and the ones I found are very simple (like this one which was very helpful with dynamic scoping) But I'm having trouble understanding how static scope works.
Here I post an exercise I did to check if I got the right solution:
considering the following program written in pseudocode:
int u = 42;
int v = 69;
int w = 17;
proc add( z:int )
u := v + u + z
proc bar( fun:proc )
int u := w;
fun(v)
proc foo( x:int, w:int )
int v := x;
bar(add)
main
foo(u,13)
print(u)
end;
What is printed to screen
a) using static scope? answer=180
b) using dynamic scope and deep binding? answer=69 (sum for u = 126 but it's foo's local v, right?)
c) using dynamic scope and shallow binding? answer=69 (sum for u = 101 but it's foo's local v, right?)
PS: I'm trying to practice doing some exercises like this if you know where I can find these types of problems (preferable with solutions) please give the link, thanks!
Your answer for lexical (static) scope is correct. Your answers for dynamic scope are wrong, but if I'm reading your explanations right, it's because you got confused between u and v, rather than because of any real misunderstanding about how deep and shallow binding work. (I'm assuming that your u/v confusion was just accidental, and not due to a strange confusion about values vs. references in the call to foo.)
a) using static scope? answer=180
Correct.
b) using dynamic scope and deep binding? answer=69 (sum for u = 126 but it's foo's local v, right?)
Your parenthetical explanation is right, but your answer is wrong: u is indeed set to 126, and foo indeed localizes v, but since main prints u, not v, the answer is 126.
c) using dynamic scope and shallow binding? answer=69 (sum for u = 101 but it's foo's local v, right?)
The sum for u is actually 97 (42+13+42), but since bar localizes u, the answer is 42. (Your parenthetical explanation is wrong for this one — you seem to have used the global variable w, which is 17, in interpreting the statement int u := w in the definition of bar; but that statement actually refers to foo's local variable w, its second parameter, which is 13. But that doesn't actually affect the answer. Your answer is wrong for this one only because main prints u, not v.)
For lexical scope, it's pretty easy to check your answers by translating the pseudo-code into a language with lexical scope. Likewise dynamic scope with shallow binding. (In fact, if you use Perl, you can test both ways almost at once, since it supports both; just use my for lexical scope, then do a find-and-replace to change it to local for dynamic scope. But even if you use, say, JavaScript for lexical scope and Bash for dynamic scope, it should be quick to test both.)
Dynamic scope with deep binding is much trickier, since few widely-deployed languages support it. If you use Perl, you can implement it manually by using a hash (an associative array) that maps from variable-names to scalar-refs, and passing this hash from function to function. Everywhere that the pseudocode declares a local variable, you save the existing scalar-reference in a Perl lexical variable, then put the new mapping in the hash; and at the end of the function, you restore the original scalar-reference. To support the binding, you create a wrapper function that creates a copy of the hash, and passes that to its wrapped function. Here is a dynamically-scoped, deeply-binding implementation of your program in Perl, using that approach:
#!/usr/bin/perl -w
use warnings;
use strict;
# Create a new scalar, initialize it to the specified value,
# and return a reference to it:
sub new_scalar($)
{ return \(shift); }
# Bind the specified procedure to the specified environment:
sub bind_proc(\%$)
{
my $V = { %{+shift} };
my $f = shift;
return sub { $f->($V, #_); };
}
my $V = {};
$V->{u} = new_scalar 42; # int u := 42
$V->{v} = new_scalar 69; # int v := 69
$V->{w} = new_scalar 17; # int w := 17
sub add(\%$)
{
my $V = shift;
my $z = $V->{z}; # save existing z
$V->{z} = new_scalar shift; # create & initialize new z
${$V->{u}} = ${$V->{v}} + ${$V->{u}} + ${$V->{z}};
$V->{z} = $z; # restore old z
}
sub bar(\%$)
{
my $V = shift;
my $fun = shift;
my $u = $V->{u}; # save existing u
$V->{u} = new_scalar ${$V->{w}}; # create & initialize new u
$fun->(${$V->{v}});
$V->{u} = $u; # restore old u
}
sub foo(\%$$)
{
my $V = shift;
my $x = $V->{x}; # save existing x
$V->{x} = new_scalar shift; # create & initialize new x
my $w = $V->{w}; # save existing w
$V->{w} = new_scalar shift; # create & initialize new w
my $v = $V->{v}; # save existing v
$V->{v} = new_scalar ${$V->{x}}; # create & initialize new v
bar %$V, bind_proc %$V, \&add;
$V->{v} = $v; # restore old v
$V->{w} = $w; # restore old w
$V->{x} = $x; # restore old x
}
foo %$V, ${$V->{u}}, 13;
print "${$V->{u}}\n";
__END__
and indeed it prints 126. It's obviously messy and error-prone, but it also really helps you understand what's going on, so for educational purposes I think it's worth it!
Simple and deep binding are Lisp interpreter viewpoints of the pseudocode. Scoping is just pointer arithmetic. Dynamic scope and static scope are the same if there are no free variables.
Static scope relies on a pointer to memory. Empty environments hold no symbol to value associations; denoted by word "End." Each time the interpreter reads an assignment, it makes space for association between a symbol and value.
The environment pointer is updated to point to the last association constructed.
env = End
env = [u,42] -> End
env = [v,69] -> [u,42] -> End
env = [w,17] -> [v,69] -> [u,42] -> End
Let me record this environment memory location as AAA. In my Lisp interpreter, when meeting a procedure, we take the environment pointer and put it our pocket.
env = [add,[closure,(lambda(z)(setq u (+ v u z)),*AAA*]]->[w,17]->[v,69]->[u,42]->End.
That's pretty much all there is until the procedure add is called. Interestingly, if add is never called, you just cost yourself a pointer.
Suppose the program calls add(8). OK, let's roll. The environment AAA is made current. Environment is ->[w,17]->[v,69]->[u,42]->End.
Procedure parameters of add are added to the front of the environment. The environment becomes [z,8]->[w,17]->[v,69]->[u,42]->End.
Now the procedure body of add is executed. Free variable v will have value 69. Free variable u will have value 42. z will have the value 8.
u := v + u + z
u will be assigned the value of 69 + 42 + 8 becomeing 119.
The environment will reflect this: [z,8]->[w,17]->[v,69]->[u,119]->End.
Assume procedure add has completed its task. Now the environment gets restored to its previous value.
env = [add,[closure,(lambda(z)(setq u (+ v u z)),*AAA*]]->[w,17]->[v,69]->[u,119]->End.
Notice how the procedure add has had a side effect of changing the value of free variable u. Awesome!
Regarding dynamic scoping: it just ensures closure leaves out dynamic symbols, thereby avoiding being captured and becoming dynamic.
Then put assignment to dynamic at top of code. If dynamic is same as parameter name, it gets masked by parameter value passed in.
Suppose I had a dynamic variable called z. When I called add(8), z would have been set to 8 regardless of what I wanted. That's probably why dynamic variables have longer names.
Rumour has it that dynamic variables are useful for things like backtracking, using let Lisp constructs.
Static binding, also known as lexical scope, refers to the scoping mechanism found in most modern languages.
In "lexical scope", the final value for u is neither 180 or 119, which are wrong answers.
The correct answer is u=101.
Please see standard Perl code below to understand why.
use strict;
use warnings;
my $u = 42;
my $v = 69;
my $w = 17;
sub add {
my $z = shift;
$u = $v + $u + $z;
}
sub bar {
my $fun = shift;
$u = $w;
$fun->($v);
}
sub foo {
my ($x, $w) = #_;
$v = $x;
bar( \&add );
}
foo($u,13);
print "u: $u\n";
Regarding shallow binding versus deep binding, both mechanisms date from the former LISP era.
Both mechanisms are meant to achieve dynamic binding (versus lexical scope binding) and therefore they produce identical results !
The differences between shallow binding and deep binding do not reside in semantics, which are identical, but in the implementation of dynamic binding.
With deep binding, variable bindings are set within a stack as "varname => varvalue" pairs.
The value of a given variable is retrieved from traversing the stack from top to bottom until a binding for the given variable is found.
Updating the variable consists in finding the binding in the stack and updating the associated value.
On entering a subroutine, a new binding for each actual parameter is pushed onto the stack, potentially hiding an older binding which is therefore no longer accessible wrt the retrieving mechanism described above (that stops at the 1st retrieved binding).
On leaving the subroutine, bindings for these parameters are simply popped from the binding stack, thus re-enabling access to the former bindings.
Please see the the code below for a Perl implementation of deep-binding dynamic scope.
use strict;
use warnings;
use utf8;
##
# Dynamic-scope deep-binding implementation
my #stack = ();
sub bindv {
my ($varname, $varval);
unshift #stack, [ $varname => $varval ]
while ($varname, $varval) = splice #_, 0, 2;
return $varval;
}
sub unbindv {
my $n = shift || 1;
shift #stack while $n-- > 0;
}
sub getv {
my $varname = shift;
for (my $i=0; $i < #stack; $i++) {
return $stack[$i][1]
if $varname eq $stack[$i][0];
}
return undef;
}
sub setv {
my ($varname, $varval) = #_;
for (my $i=0; $i < #stack; $i++) {
return $stack[$i][1] = $varval
if $varname eq $stack[$i][0];
}
return bindv($varname, $varval);
}
##
# EXERCICE
bindv( u => 42,
v => 69,
w => 17,
);
sub add {
bindv(z => shift);
setv(u => getv('v')
+ getv('u')
+ getv('z')
);
unbindv();
}
sub bar {
bindv(fun => shift);
setv(u => getv('w'));
getv('fun')->(getv('v'));
unbindv();
}
sub foo {
bindv(x => shift,
w => shift,
);
setv(v => getv('x'));
bar( \&add );
unbindv(2);
}
foo( getv('u'), 13);
print "u: ", getv('u'), "\n";
The result is u=97
Nevertheless, this constant traversal of the binding stack is costly : 0(n) complexity !
Shallow binding brings a wonderful O(1) enhanced performance over the previous implementation !
Shallow binding is improving the former mechanism by assigning each variable its own "cell", storing the value of the variable within the cell.
The value of a given variable is simply retrieved from the variable's
cell (using a hash table on variable names, we achieve a
0(1) complexity for accessing variable's values!)
Updating the variable's value is simply storing the value into the
variable's cell.
Creating a new binding (entering subs) works by pushing the old value
of the variable (a previous binding) onto the stack, and storing the
new local value in the value cell.
Eliminating a binding (leaving subs) works by popping the old value
off the stack into the variable's value cell.
Please see the the code below for a trivial Perl implementation of shallow-binding dynamic scope.
use strict;
use warnings;
our $u = 42;
our $v = 69;
our $w = 17;
our $z;
our $fun;
our $x;
sub add {
local $z = shift;
$u = $v + $u + $z;
}
sub bar {
local $fun = shift;
$u = $w;
$fun->($v);
}
sub foo {
local $x = shift;
local $w = shift;
$v = $x;
bar( \&add );
}
foo($u,13);
print "u: $u\n";
As you shall see, the result is still u=97
As a conclusion, remember two things :
shallow binding produces the same results as deep binding, but runs faster, since there is never a need to search for a binding.
The problem is not shallow binding versus deep binding versus
static binding BUT lexical scope versus dynamic scope (implemented either with deep or shallow binding).

How do I write a generic memoize function?

I'm writing a function to find triangle numbers and the natural way to write it is recursively:
function triangle (x)
if x == 0 then return 0 end
return x+triangle(x-1)
end
But attempting to calculate the first 100,000 triangle numbers fails with a stack overflow after a while. This is an ideal function to memoize, but I want a solution that will memoize any function I pass to it.
Mathematica has a particularly slick way to do memoization, relying on the fact that hashes and function calls use the same syntax:
triangle[0] = 0;
triangle[x_] := triangle[x] = x + triangle[x-1]
That's it. It works because the rules for pattern-matching function calls are such that it always uses a more specific definition before a more general definition.
Of course, as has been pointed out, this example has a closed-form solution: triangle[x_] := x*(x+1)/2. Fibonacci numbers are the classic example of how adding memoization gives a drastic speedup:
fib[0] = 1;
fib[1] = 1;
fib[n_] := fib[n] = fib[n-1] + fib[n-2]
Although that too has a closed-form equivalent, albeit messier: http://mathworld.wolfram.com/FibonacciNumber.html
I disagree with the person who suggested this was inappropriate for memoization because you could "just use a loop". The point of memoization is that any repeat function calls are O(1) time. That's a lot better than O(n). In fact, you could even concoct a scenario where the memoized implementation has better performance than the closed-form implementation!
You're also asking the wrong question for your original problem ;)
This is a better way for that case:
triangle(n) = n * (n - 1) / 2
Furthermore, supposing the formula didn't have such a neat solution, memoisation would still be a poor approach here. You'd be better off just writing a simple loop in this case. See this answer for a fuller discussion.
I bet something like this should work with variable argument lists in Lua:
local function varg_tostring(...)
local s = select(1, ...)
for n = 2, select('#', ...) do
s = s..","..select(n,...)
end
return s
end
local function memoize(f)
local cache = {}
return function (...)
local al = varg_tostring(...)
if cache[al] then
return cache[al]
else
local y = f(...)
cache[al] = y
return y
end
end
end
You could probably also do something clever with a metatables with __tostring so that the argument list could just be converted with a tostring(). Oh the possibilities.
In C# 3.0 - for recursive functions, you can do something like:
public static class Helpers
{
public static Func<A, R> Memoize<A, R>(this Func<A, Func<A,R>, R> f)
{
var map = new Dictionary<A, R>();
Func<A, R> self = null;
self = (a) =>
{
R value;
if (map.TryGetValue(a, out value))
return value;
value = f(a, self);
map.Add(a, value);
return value;
};
return self;
}
}
Then you can create a memoized Fibonacci function like this:
var memoized_fib = Helpers.Memoize<int, int>((n,fib) => n > 1 ? fib(n - 1) + fib(n - 2) : n);
Console.WriteLine(memoized_fib(40));
In Scala (untested):
def memoize[A, B](f: (A)=>B) = {
var cache = Map[A, B]()
{ x: A =>
if (cache contains x) cache(x) else {
val back = f(x)
cache += (x -> back)
back
}
}
}
Note that this only works for functions of arity 1, but with currying you could make it work. The more subtle problem is that memoize(f) != memoize(f) for any function f. One very sneaky way to fix this would be something like the following:
val correctMem = memoize(memoize _)
I don't think that this will compile, but it does illustrate the idea.
Update: Commenters have pointed out that memoization is a good way to optimize recursion. Admittedly, I hadn't considered this before, since I generally work in a language (C#) where generalized memoization isn't so trivial to build. Take the post below with that grain of salt in mind.
I think Luke likely has the most appropriate solution to this problem, but memoization is not generally the solution to any issue of stack overflow.
Stack overflow usually is caused by recursion going deeper than the platform can handle. Languages sometimes support "tail recursion", which re-uses the context of the current call, rather than creating a new context for the recursive call. But a lot of mainstream languages/platforms don't support this. C# has no inherent support for tail-recursion, for example. The 64-bit version of the .NET JITter can apply it as an optimization at the IL level, which is all but useless if you need to support 32-bit platforms.
If your language doesn't support tail recursion, your best option for avoiding stack overflows is either to convert to an explicit loop (much less elegant, but sometimes necessary), or find a non-iterative algorithm such as Luke provided for this problem.
function memoize (f)
local cache = {}
return function (x)
if cache[x] then
return cache[x]
else
local y = f(x)
cache[x] = y
return y
end
end
end
triangle = memoize(triangle);
Note that to avoid a stack overflow, triangle would still need to be seeded.
Here's something that works without converting the arguments to strings.
The only caveat is that it can't handle a nil argument. But the accepted solution can't distinguish the value nil from the string "nil", so that's probably OK.
local function m(f)
local t = { }
local function mf(x, ...) -- memoized f
assert(x ~= nil, 'nil passed to memoized function')
if select('#', ...) > 0 then
t[x] = t[x] or m(function(...) return f(x, ...) end)
return t[x](...)
else
t[x] = t[x] or f(x)
assert(t[x] ~= nil, 'memoized function returns nil')
return t[x]
end
end
return mf
end
I've been inspired by this question to implement (yet another) flexible memoize function in Lua.
https://github.com/kikito/memoize.lua
Main advantages:
Accepts a variable number of arguments
Doesn't use tostring; instead, it organizes the cache in a tree structure, using the parameters to traverse it.
Works just fine with functions that return multiple values.
Pasting the code here as reference:
local globalCache = {}
local function getFromCache(cache, args)
local node = cache
for i=1, #args do
if not node.children then return {} end
node = node.children[args[i]]
if not node then return {} end
end
return node.results
end
local function insertInCache(cache, args, results)
local arg
local node = cache
for i=1, #args do
arg = args[i]
node.children = node.children or {}
node.children[arg] = node.children[arg] or {}
node = node.children[arg]
end
node.results = results
end
-- public function
local function memoize(f)
globalCache[f] = { results = {} }
return function (...)
local results = getFromCache( globalCache[f], {...} )
if #results == 0 then
results = { f(...) }
insertInCache(globalCache[f], {...}, results)
end
return unpack(results)
end
end
return memoize
Here is a generic C# 3.0 implementation, if it could help :
public static class Memoization
{
public static Func<T, TResult> Memoize<T, TResult>(this Func<T, TResult> function)
{
var cache = new Dictionary<T, TResult>();
var nullCache = default(TResult);
var isNullCacheSet = false;
return parameter =>
{
TResult value;
if (parameter == null && isNullCacheSet)
{
return nullCache;
}
if (parameter == null)
{
nullCache = function(parameter);
isNullCacheSet = true;
return nullCache;
}
if (cache.TryGetValue(parameter, out value))
{
return value;
}
value = function(parameter);
cache.Add(parameter, value);
return value;
};
}
}
(Quoted from a french blog article)
In the vein of posting memoization in different languages, i'd like to respond to #onebyone.livejournal.com with a non-language-changing C++ example.
First, a memoizer for single arg functions:
template <class Result, class Arg, class ResultStore = std::map<Arg, Result> >
class memoizer1{
public:
template <class F>
const Result& operator()(F f, const Arg& a){
typename ResultStore::const_iterator it = memo_.find(a);
if(it == memo_.end()) {
it = memo_.insert(make_pair(a, f(a))).first;
}
return it->second;
}
private:
ResultStore memo_;
};
Just create an instance of the memoizer, feed it your function and argument. Just make sure not to share the same memo between two different functions (but you can share it between different implementations of the same function).
Next, a driver functon, and an implementation. only the driver function need be public
int fib(int); // driver
int fib_(int); // implementation
Implemented:
int fib_(int n){
++total_ops;
if(n == 0 || n == 1)
return 1;
else
return fib(n-1) + fib(n-2);
}
And the driver, to memoize
int fib(int n) {
static memoizer1<int,int> memo;
return memo(fib_, n);
}
Permalink showing output on codepad.org. Number of calls is measured to verify correctness. (insert unit test here...)
This only memoizes one input functions. Generalizing for multiple args or varying arguments left as an exercise for the reader.
In Perl generic memoization is easy to get. The Memoize module is part of the perl core and is highly reliable, flexible, and easy-to-use.
The example from it's manpage:
# This is the documentation for Memoize 1.01
use Memoize;
memoize('slow_function');
slow_function(arguments); # Is faster than it was before
You can add, remove, and customize memoization of functions at run time! You can provide callbacks for custom memento computation.
Memoize.pm even has facilities for making the memento cache persistent, so it does not need to be re-filled on each invocation of your program!
Here's the documentation: http://perldoc.perl.org/5.8.8/Memoize.html
Extending the idea, it's also possible to memoize functions with two input parameters:
function memoize2 (f)
local cache = {}
return function (x, y)
if cache[x..','..y] then
return cache[x..','..y]
else
local z = f(x,y)
cache[x..','..y] = z
return z
end
end
end
Notice that parameter order matters in the caching algorithm, so if parameter order doesn't matter in the functions to be memoized the odds of getting a cache hit would be increased by sorting the parameters before checking the cache.
But it's important to note that some functions can't be profitably memoized. I wrote memoize2 to see if the recursive Euclidean algorithm for finding the greatest common divisor could be sped up.
function gcd (a, b)
if b == 0 then return a end
return gcd(b, a%b)
end
As it turns out, gcd doesn't respond well to memoization. The calculation it does is far less expensive than the caching algorithm. Ever for large numbers, it terminates fairly quickly. After a while, the cache grows very large. This algorithm is probably as fast as it can be.
Recursion isn't necessary. The nth triangle number is n(n-1)/2, so...
public int triangle(final int n){
return n * (n - 1) / 2;
}
Please don't recurse this. Either use the x*(x+1)/2 formula or simply iterate the values and memoize as you go.
int[] memo = new int[n+1];
int sum = 0;
for(int i = 0; i <= n; ++i)
{
sum+=i;
memo[i] = sum;
}
return memo[n];