Julia: performance of matrix multiplication vs for loop - for-loop

I am puzzled about the superior performance of matrix multiplication versus for loop in my application. Similar questions have been asked a few times on StackOverflow already (e.g., here and here), but I couldn't get to the bottom of it with the provided answers.
Here's a script to reproduce my problem (using Julia 1.7.3)
# Packages and seeds
using Optim, BenchmarkTools, Random
Random.seed!(1704)
# Parameters
N = 10
M = 11
σ = 3.0
ID1 = repeat(1:N, inner = M)
ID2 = repeat(1:M, inner = N)
A = rand(N * M)
w = rand(M)
dv = dummyvar(ID2)
x0 = ones(N*M)
# Function 1
function function_1(x, dv, w, A, σ)
c = dv * w ./ A
s = x.^(1-σ) ./ (dv * (dv' * x.^(1-σ)))
el = σ*(1 .- s) .+ s
p = (c.*el) ./ (el .- 1)
return sum(abs2.(x .- p))
end
# Function 2
function function_2(x, dv, w, A, M, ID2, σ)
f = 0.0
for m = 1:M
c = w[m] ./ A[ID2 .== m]
s = x[ID2 .== m].^(1-σ) ./ sum(x[ID2 .== m].^(1-σ))
el = σ*(1 .- s) .+ s
f += sum(abs2.(x[ID2 .== m] .- (c.*el) ./ (el .- 1)))
end
return f
end
and dummyvar is a function that creates a matrix of indices from a vector (similar to the Matlab dummyvar function)
function dummyvar(input_mat::Vector)
# Initialize
n1 = length(input_mat)
n2 = length(unique(input_mat))
# Preallocate output
mat = zeros(n1, n2)
# Fill in output
for i in 1:length(unique(input_mat))
mat[:,i] = input_mat .== unique(input_mat)[i]
end
# Return
return mat
end
Here are the benchmark results for function_1
and here are those of function_2
function_1 allocates significantly less memory and takes less to execute than function_2. However, I expected to find the opposite pattern since function_1 uses matrix multiplication to compute c and s whereas function_2 uses loops. I wonder if pre-allocating f every time function_2 is called explains the difference.
Can function_2 be rewritten to outperform function_1 in terms of execution time and memory allocation?

take this for example
w[m] ./ A[ID2 .== m]
this allocates the following intermediate arrays
ID2 .== m
A[ID2 .== m]
w[m] ./ A[ID2 .== m]
3 is irreducible because you need this for c, but 1 and 2 are not.
Apply this logic to each item in your for-loop.

Related

Algorithm to precisely compare two exponentiations for very large integers (order of 1 billion)

We want to compare a^b to c^d, and tell if the first is smaller, greater, or equal (where ^ denotes exponentiation).
Obviously, for very large numbers, we cannot explicitely compute these values.
The most common approach in this situation is to apply log on both sides and compare b * log(a) to d * log(c). The issue here is that logs are floating-point operations, and as such we cannot trust our answer with 100% confidence (there might be some values which are incredibly close, and because of floating-point error we get a wrong answer).
Is there an algorithm for solving this problem? I've been scouring the intrernet for this, but I can only find solutions which work for particular cases only (e.g. in which one exponent is a multiple of another), or which use floating point in some way (logarithms, division) etc.
This is sort of two questions in one:
Are they equal?
If not, which one is greater?
As Peter O. observes, it's easiest to build in a language that provides an arbitrary-precision fraction type. I'll use Python 3.
Let's assume without loss of generality that a ≤ c (swap if necessary) and b is relatively prime to d (divide both by the greatest common divisor).
To get at the core of the question, I'm going to assume that a, c > 0 and b, d ≥ 0. Removing this assumption is tedious but not difficult.
Equality test
There are some easy cases where a = 1 or b = 0 or c = 1 or d = 0.
Separately, necessary conditions for a^b = c^d are
i. b ≥ d, since otherwise b < d, which together with a ≤ c implies a^b < c^d;
ii. a is a divisor of c, since we know from (i) that a^b = c^d is a divisor of c^b = c^(b−d) c^d.
When these conditions hold, we can divide through by a^d to reduce the problem to testing whether a^(b−d) = (c/a)^d.
In Python 3:
def equal_powers(a, b, c, d):
while True:
lhs_is_one = a == 1 or b == 0
rhs_is_one = c == 1 or d == 0
if lhs_is_one or rhs_is_one:
return lhs_is_one and rhs_is_one
if a > c:
a, b, c, d = c, d, a, b
if b < d:
return False
q, r = divmod(c, a)
if r != 0:
return False
b -= d
c = q
def test_equal_powers():
for a in range(1, 25):
for b in range(25):
for c in range(1, 25):
for d in range(25):
assert equal_powers(a, b, c, d) == (a ** b == c ** d)
test_equal_powers()
Inequality test
Once we've established that the two quantities are not equal, it's time to figure out which one is greater. (Without the equality test, the code here could run forever.)
If you're doing this for real, you should consult an actual reference on computing elementary functions. I'm just going to try to do the simplest thing that works.
Time for a calculus refresher. We have the Taylor series
−log x = (1−x) + (1−x)^2/2 + (1−x)^3/3 + (1−x)^4/4 + ...
To get a lower bound, truncate the series. To get an upper bound, we can truncate but replace the final term (1−x)^n/n with (1−x)^n/n (1/x), since
(1−x)^n/n (1/x)
= (1−x)^n/n (1 + (1−x) + (1−x)^2 + ...)
= (1−x)^n/n + (1−x)^(n+1)/n + (1−x)^(n+2)/n + ...
> (1−x)^n/n + (1−x)^(n+1)/(n+1) + (1−x)^(n+2)/(n+2) + ...
To get a good convergence rate, we're going to want 0.5 ≤ x < 1, which we can achieve by dividing x by a power of two.
In Python, we'll represent a real number as an infinite generator of shrinking intervals that contain the true value. Once the intervals for b log a and d log c are disjoint, we can determine how they compare.
import fractions
def minus(x, y):
while True:
x_lo, x_hi = next(x)
y_lo, y_hi = next(y)
yield x_lo - y_hi, x_hi - y_lo
def times(b, x):
for lo, hi in x:
yield b * lo, b * hi
def restricted_log(a):
series = 0
n = 0
numerator = 1
while True:
n += 1
numerator *= 1 - a
series += fractions.Fraction(numerator, n)
yield -(series + fractions.Fraction(numerator * (1 - a), (n + 1) * a)), -series
def log(a):
n = 0
while a >= 1:
a = fractions.Fraction(a, 2)
n += 1
return minus(restricted_log(a), times(n, restricted_log(fractions.Fraction(1, 2))))
def less_powers(a, b, c, d):
lhs = times(b, log(a))
rhs = times(d, log(c))
while True:
lhs_lo, lhs_hi = next(lhs)
rhs_lo, rhs_hi = next(rhs)
if lhs_hi < rhs_lo:
return True
if rhs_hi < lhs_lo:
return False
def test_less_powers():
for a in range(1, 10):
for b in range(10):
for c in range(1, 10):
for d in range(10):
if a ** b != c ** d:
assert less_powers(a, b, c, d) == (a ** b < c ** d)
test_less_powers()

Cheking a Zero Polynomial

Can anyone explain me in below algorithm how the "ISZERO" function checking whether the polynomial is zero or not. Here "REM(P,e)" function removes all the values with exponent "e".
what i don't able to understand is the significance of "if COEF(P,e) = - c". And also what is this "SMULT" function.
structure POLYNOMIAL
declare ZERO( ) poly; ISZERO(poly) Boolean
COEF(poly,exp) coef;
ATTACH(poly,coef,exp) poly
REM(poly,exp) poly
SMULT(poly,coef,exp) poly
ADD(poly,poly) poly; MULT(poly,poly) poly;
for all P,Q, poly c,d, coef e,f exp let
REM(ZERO,f) :: = ZERO
REM(ATTACH(P,c,e),f) :: =
if e = f then REM(P,f) else ATTACH(REM(P,f),c,e)
***ISZERO(ZERO) :: = true
ISZERO(ATTACH(P,c,e)):: =
if COEF(P,e) = - c then ISZERO(REM(P,e)) else false***
COEF(ZERO,e) :: = 0
COEF(ATTACH(P,c,e),f) :: =
if e = f then c + COEF(P,f) else COEF(P,f)
SMULT(ZERO,d,f) :: = ZERO
SMULT(ATTACH(P,c,e),d,f) :: =
ATTACH(SMULT(P,d,f),c d,e + f)
ADD(P,ZERO):: = P
ADD(P,ATTACH(Q,d,f)) :: = ATTACH(ADD(P,Q),d,f)
MULT(P,ZERO) :: = ZERO
MULT(P,ATTACH(Q,d,f)) :: =
ADD(MULT(P,Q),SMULT(P,d,f))
end
end POLYNOMIAL
Without knowing what language this is, it looks like this line
ISZERO(ATTACH(P,c,e)):: =
if COEF(P,e) = - c then ISZERO(REM(P,e)) else false
is specifying ISZERO recursively. We are trying to determine whether ATTACH(P, c, e), otherwise known as P(x) + cx^e, is zero. It first checks whether the x^e coefficient of P is -c. If not, then P(x) + cx^e is definitely not zero, and you can return false immediately. Otherwise, P(x) + cx^e = REM(P, e), so you have to check ISZERO(REM(P, e)).
I believe SMULT is multiplication, so SMULT(P, a, b) is equivalent to a * x^b * P(x).

Understanding the runtime of a recursive SML function involving list appending (using #)

I'm new to algorithm analysis and SML and got hung up on the average-case runtime of the following SML function. I would appreciate some feedback on my thinking.
fun app([]) = []
| app(h::t) = [h] # app(t)
So after every recursion we will end up with a bunch of single element lists (and one no-element list).
[1]#[2]#[3]#...#[n]#[]
Where n is the number of elements in the original list and 1, 2, 3, ..., n is just to illustrate what elements in the original list we are talking about. L # R takes time linear in the length of list L. Assuming A is the constant amount of time # takes for every element, I imagine this as if:
[1,2]#[3]#[4]#...#[n]#[] took 1A
[1,2,3]#[4]#...#[n]#[] took 2A
[1,2,3,4]#...#[n]#[] took 3A
...
[1,2,3,4,...,n]#[] took (n-1)A
[1,2,3,4,...,n] took nA
I'm therefore thinking that a recurrence for the time would look something like this:
T(0) = C (if n = 0)
T(n) = T(n-1) + An + B (if n > 0)
Where C is just the final matching of the base case app([]) and B is the constant for h::t. Close the recurrence and we will get this (proof omitted):
T(n) = (n²+n)A/2 + Bn + C = (A/2)n² + (A/2)n + Bn + C = Θ(n²)
This is my own conclusion which differs from the answer that was presented to me, namely:
T(0) = B (if n = 0)
T(n) = T(n-1) + A (if n > 0)
Closed form
T(n) = An + B = Θ(n)
Which is quite different. (Θ(n) vs Θ(n²)!) But isn't this assuming that L # R takes constant time rather than linear? For example, it would be true for addition
fun add([]) = 0
| add(h::t) = h + add(t) (* n + ... + 2 + 1 + 0 *)
or even concatenation
fun con([]) = []
| con(h::t) = h::con(t) (* n :: ... :: 2 :: 1 :: [] *)
Am I misunderstanding the way that L # R exists or is my analysis (at least sort of) correct?
Yes. Running the app [1,2,3] command by hand one function call at a time gives:
app [1,2,3]
[1]#(app [2,3])
[1]#([2]#(app [3]))
[1]#([2]#([3]#(app [])))
[1]#([2]#([3]#([])))
[1]#([2]#[3])
[1]#([2,3])
[1,2,3]
This is a consequence of the function call being on the left-side of the #.
Compare this to a naïve version of rev:
fun rev [] = []
| rev (x::xs) = rev xs # [x]
This one has the running time you expect: Once the recursion has fully expanded into an expression ((([])#[3])#[2])#[1] (taking linear time), it requires n + (n - 1) + (n - 2) + ... + 1, or n(n+1)/2, or O(n^2) steps to complete the computation. A more effective rev could look like this:
local
fun rev' [] ys = ys
| rev' (x::xs) ys = rev' xs (x::ys)
in
fun rev xs = rev' xs []
end

performance issue on collatz chain

I have a working program to compute the longest collatz chain in a given range (project euler n°14). I think it works correctly, but is very slow. I tried to look for a better solution, but I can only reduce slightly the evaluated domain. Am I doing something wrong?
The implementation use memoization to avoid computing the same result twice. Is Data.Map bad for general performances?
import Data.Map ((!), member, insert, singleton, assocs, Map)
insertSolution::Integer->(Map Integer Integer)->(Map Integer Integer)
insertSolution n syracMap
| n `member` syracMap = syracMap
|otherwise = let
next = if n `mod` 2 == 0 then n `div` 2 else 3 * n + 1
newMap = insertSolution next syracMap
solution = newMap ! next + 1
in insert n solution newMap
bound = 1::Integer
lower = 999999::Integer
test::[Integer]
test = [lower,lower+2..bound]
values = takeWhile (\(k, v) -> k < bound) $ assocs $ foldr insertSolution (singleton 1 1) test
result = foldr (\(k, v) (k', v') -> if v > v' then (k, v) else (k', v')) (1, 1) values
main = putStr $ show $ result
edit
updated function to remove bug. It is still pretty slow on my laptop.
FWIW, here's my solution:
module Main
where
import Data.List
import Data.Ord
next_hailstone n | even n = n `div` 2
| otherwise = 3*n+1
gen_next_hailstone n
= if nh == 1
then Nothing
else Just (nh, nh)
where nh = next_hailstone n
hailstone n = unfoldr gen_next_hailstone n
hailstone_seqs = map hailstone [1..1000000]
zip_hailstone = zip [1..1000000] hailstone_seqs
max_hailstone = maximumBy (comparing (length . snd)) zip_hailstone
main = print . fst $ max_hailstone
It's relatively fast. If you want more speed, consult the Haskell wiki (SPOILER ALERT!!!).

Factorial Algorithms in different languages

Locked. This question and its answers are locked because the question is off-topic but has historical significance. It is not currently accepting new answers or interactions.
I want to see all the different ways you can come up with, for a factorial subroutine, or program. The hope is that anyone can come here and see if they might want to learn a new language.
Ideas:
Procedural
Functional
Object Oriented
One liners
Obfuscated
Oddball
Bad Code
Polyglot
Basically I want to see an example, of different ways of writing an algorithm, and what they would look like in different languages.
Please limit it to one example per entry.
I will allow you to have more than one example per answer, if you are trying to highlight a specific style, language, or just a well thought out idea that lends itself to being in one post.
The only real requirement is it must find the factorial of a given argument, in all languages represented.
Be Creative!
Recommended Guideline:
# Language Name: Optional Style type
- Optional bullet points
Code Goes Here
Other informational text goes here
I will ocasionally go along and edit any answer that does not have decent formatting.
Polyglot: 5 languages, all using bignums
So, I wrote a polyglot which works in the three languages I often write in, as well as one from my other answer to this question and one I just learned today. It's a standalone program, which reads a single line containing a nonnegative integer and prints a single line containing its factorial. Bignums are used in all languages, so the maximum computable factorial depends only on your computer's resources.
Perl: uses built-in bignum package. Run with perl FILENAME.
Haskell: uses built-in bignums. Run with runhugs FILENAME or your favorite compiler's equivalent.
C++: requires GMP for bignum support. To compile with g++, use g++ -lgmpxx -lgmp -x c++ FILENAME to link against the right libraries. After compiling, run ./a.out. Or use your favorite compiler's equivalent.
brainf*ck: I wrote some bignum support in this post. Using Muller's classic distribution, compile with bf < FILENAME > EXECUTABLE. Make the output executable and run it. Or use your favorite distribution.
Whitespace: uses built-in bignum support. Run with wspace FILENAME.
Edit: added Whitespace as a fifth language. Incidentally, do not wrap the code with <code> tags; it breaks the Whitespace. Also, the code looks much nicer in fixed-width.
char //# b=0+0{- |0*/; #>>>>,----------[>>>>,--------
#define a/*#--]>>>>++<<<<<<<<[>++++++[<------>-]<-<<<
#Perl ><><><> <> <> <<]>>>>[[>>+<<-]>>[<<+>+>-]<->
#C++ --><><> <><><>< > < > < +<[>>>>+<<<-<[-]]>[-]
#Haskell >>]>[-<<<<<[<<<<]>>>>[[>>+<<-]>>[<<+>+>-]>>]
#Whitespace >>>>[-[>+<-]+>>>>]<<<<[<<<<]<<<<[<<<<
#brainf*ck > < ]>>>>>[>>>[>>>>]>>>>[>>>>]<<<<[[>>>>*/
exp; ;//;#+<<<<-]<<<<]>>>>+<<<<<<<[<<<<][.POLYGLOT^5.
#include <gmpxx.h>//]>>>>-[>>>[>>>>]>>>>[>>>>]<<<<[>>
#define eval int main()//>+<<<-]>>>[<<<+>>+>->
#include <iostream>//<]<-[>>+<<[-]]<<[<<<<]>>>>[>[>>>
#define print std::cout << // > <+<-]>[<<+>+>-]<<[>>>
#define z std::cin>>//<< +<<<-]>>>[<<<+>>+>-]<->+++++
#define c/*++++[-<[-[>>>>+<<<<-]]>>>>[<<<<+>>>>-]<<*/
#define abs int $n //>< <]<[>>+<<<<[-]>>[<<+>>-]]>>]<
#define uc mpz_class fact(int $n){/*<<<[<<<<]<<<[<<
use bignum;sub#<<]>>>>-]>>>>]>>>[>[-]>>>]<<<<[>>+<<-]
z{$_[0+0]=readline(*STDIN);}sub fact{my($n)=shift;#>>
#[<<+>+>-]<->+<[>-<[-]]>[-<<-<<<<[>>+<<-]>>[<<+>+>+*/
uc;if($n==0){return 1;}return $n*fact($n-1); }//;#
eval{abs;z($n);print fact($n);print("\n")/*2;};#-]<->
'+<[>-<[-]]>]<<[<<<<]<<<<-[>>+<<-]>>[<<+>+>-]+<[>-+++
-}-- <[-]]>[-<<++++++++++<<<<-[>>+<<-]>>[<<+>+>-++
fact 0 = 1 -- ><><><>< > <><>< ]+<[>-<[-]]>]<<[<<+ +
fact n=n*fact(n-1){-<<]>>>>[[>>+<<-]>>[<<+>+++>+-}
main=do{n<-readLn;print(fact n)}-- +>-]<->+<[>>>>+<<+
{-x<-<[-]]>[-]>>]>]>>>[>>>>]<<<<[>+++++++[<+++++++>-]
<--.<<<<]+written+by+++A+Rex+++2009+.';#+++x-}--x*/;}
lolcode:
sorry I couldn't resist xD
HAI
CAN HAS STDIO?
I HAS A VAR
I HAS A INT
I HAS A CHEEZBURGER
I HAS A FACTORIALNUM
IM IN YR LOOP
UP VAR!!1
TIEMZD INT!![CHEEZBURGER]
UP FACTORIALNUM!!1
IZ VAR BIGGER THAN FACTORIALNUM? GTFO
IM OUTTA YR LOOP
U SEEZ INT
KTHXBYE
This is one of the faster algorithms, up to 170!. It fails inexplicably beyond 170!, and it's relatively slow for small factorials, but for factorials between 80 and 170 it's blazingly fast compared to many algorithms.
curl http://www.google.com/search?q=170!
There's also an online interface, try it out now!
Let me know if you find a bug, or faster implementation for large factorials.
EDIT:
This algorithm is slightly slower, but gives results beyond 170:
curl http://www58.wolframalpha.com/input/?i=171!
It also simplifies them into various other representations.
C++: Template Metaprogramming
Uses the classic enum hack.
template<unsigned int n>
struct factorial {
enum { result = n * factorial<n - 1>::result };
};
template<>
struct factorial<0> {
enum { result = 1 };
};
Usage.
const unsigned int x = factorial<4>::result;
Factorial is calculated completely at compile time based on the template parameter n. Therefore, factorial<4>::result is a constant once the compiler has done its work.
Whitespace
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
It was hard to get it to show here properly, but now I tried copying it from the preview and it works. You need to input the number and press enter.
I find the following implementations just hilarious:
The Evolution of a Haskell Programmer
Evolution of a Python programmer
Enjoy!
C# Lookup:
Nothing to calculate really, just look it up. To extend it,add another 8 numbers to the table and 64 bit integers are at at their limit. Beyond that, a BigNum class is called for.
public static int Factorial(int f)
{
if (f<0 || f>12)
{
throw new ArgumentException("Out of range for integer factorial");
}
int [] fact={1,1,2,6,24,120,720,5040,40320,362880,3628800,
39916800,479001600};
return fact[f];
}
Lazy K
Your pure functional programming nightmares come true!
The only Esoteric Turing-complete Programming Language that has:
A purely functional foundation, core, and libraries---in fact, here's the complete API: S K I
No lambdas even!
No numbers or lists needed or allowed
No explicit recursion but yet, allows recursion
A simple infinite lazy stream-based I/O mechanism
Here's the Factorial code in all its parenthetical glory:
K(SII(S(K(S(S(KS)(S(K(S(KS)))(S(K(S(KK)))(S(K(S(K(S(K(S(K(S(SI(K(S(K(S(S(KS)K)I))
(S(S(KS)K)(SII(S(S(KS)K)I))))))))K))))))(S(K(S(K(S(SI(K(S(K(S(SI(K(S(K(S(S(KS)K)I))
(S(S(KS)K)(SII(S(S(KS)K)I))(S(S(KS)K))(S(SII)I(S(S(KS)K)I))))))))K)))))))
(S(S(KS)K)(K(S(S(KS)K)))))))))(K(S(K(S(S(KS)K)))K))))(SII))II)
Features:
No subtraction or conditionals
Prints all factorials (if you wait long enough)
Uses a second layer of Church numerals to convert the Nth factorial to N! asterisks followed by a newline
Uses the Y combinator for recursion
In case you are interested in trying to understand it, here is the Scheme source code to run through the Lazier compiler:
(lazy-def '(fac input)
'((Y (lambda (f n a) ((lambda (b) ((cons 10) ((b (cons 42)) (f (1+ n) b))))
(* a n)))) 1 1))
(for suitable definitions of Y, cons, 1, 10, 42, 1+, and *).
EDIT:
Lazy K Factorial in Decimal
(10KB of gibberish or else I would paste it). For example, at the Unix prompt:
$ echo "4" | ./lazy facdec.lazy
24
$ echo "5" | ./lazy facdec.lazy
120
Rather slow for numbers above, say, 5.
The code is sort of bloated because we have to include library code for all of our own primitives (code written in Hazy, a lambda calculus interpreter and LC-to-Lazy K compiler written in Haskell).
XSLT 1.0
The input file, factorial.xml:
<?xml version="1.0"?>
<?xml-stylesheet href="factorial.xsl" type="text/xsl" ?>
<n>
20
</n>
The XSLT file, factorial.xsl:
<?xml version="1.0"?>
<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"
xmlns:msxsl="urn:schemas-microsoft-com:xslt" >
<xsl:output method="text"/>
<!-- 0! = 1 -->
<xsl:template match="text()[. = 0]">
1
</xsl:template>
<!-- n! = (n-1)! * n-->
<xsl:template match="text()[. > 0]">
<xsl:variable name="x">
<xsl:apply-templates select="msxsl:node-set( . - 1 )/text()"/>
</xsl:variable>
<xsl:value-of select="$x * ."/>
</xsl:template>
<!-- Calculate n! -->
<xsl:template match="/n">
<xsl:apply-templates select="text()"/>
</xsl:template>
</xsl:stylesheet>
Save both files in the same directory and open factorial.xml in IE.
Python: Functional, One-liner
factorial = lambda n: reduce(lambda x,y: x*y, range(1, n+1), 1)
NOTE:
It supports big integers. Example:
print factorial(100)
93326215443944152681699238856266700490715968264381621468592963895217599993229915\
608941463976156518286253697920827223758251185210916864000000000000000000000000
It does not work for n < 0.
APL (oddball/one-liner):
×/⍳X
⍳X expands X into an array of the integers 1..X
×/ multiplies every element in the array
Or with the built-in operator:
!X
Source: http://www.webber-labs.com/mpl/lectures/ppt-slides/01.ppt
Perl6
sub factorial ($n) { [*] 1..$n }
I hardly know about Perl6. But I guess this [*] operator is same as Haskell's product.
This code runs on Pugs, and maybe Parrot (I didn't check it.)
Edit
This code also works.
sub postfix:<!> ($n) { [*] 1..$n }
# This function(?) call like below ... It looks like mathematical notation.
say 10!;
x86-64 Assembly: Procedural
You can call this from C (only tested with GCC on linux amd64).
Assembly was assembled with nasm.
section .text
global factorial
; factorial in x86-64 - n is passed in via RDI register
; takes a 64-bit unsigned integer
; returns a 64-bit unsigned integer in RAX register
; C declaration in GCC:
; extern unsigned long long factorial(unsigned long long n);
factorial:
enter 0,0
; n is placed in rdi by caller
mov rax, 1 ; factorial = 1
mov rcx, 2 ; i = 2
loopstart:
cmp rcx, rdi
ja loopend
mul rcx ; factorial *= i
inc rcx
jmp loopstart
loopend:
leave
ret
Recursively in Inform 7
(it reminds you of COBOL because it's for writing text adventures; proportional font is deliberate):
To decide what number is the factorial of (n - a number):
    if n is zero, decide on one;
    otherwise decide on the factorial of (n minus one) times n.
If you want to actually call this function ("phrase") from a game you need to define an action and grammar rule:
"The factorial game" [this must be the first line of the source]
There is a room. [there has to be at least one!]
Factorialing is an action applying to a number.
Understand "factorial [a number]" as factorialing.
Carry out factorialing:
    Let n be the factorial of the number understood;
    Say "It's [n]".
C#: LINQ
public static int factorial(int n)
{
return (Enumerable.Range(1, n).Aggregate(1, (previous, value) => previous * value));
}
Erlang: tail recursive
fac(0) -> 1;
fac(N) when N > 0 -> fac(N, 1).
fac(1, R) -> R;
fac(N, R) -> fac(N - 1, R * N).
Haskell:
ones = 1 : ones
integers = head ones : zipWith (+) integers (tail ones)
factorials = head integers : zipWith (*) factorials (tail integers)
Brainf*ck
+++++
>+<[[->>>>+<<<<]>>>>[-<<<<+>>+>>]<<<<>[->>+<<]<>>>[-<[->>+<<]>>[-<<+<+>>>]<]<[-]><<<-]
Written by Michael Reitzenstein.
BASIC: old school
10 HOME
20 INPUT N
30 LET ANS = 1
40 FOR I = 1 TO N
50 ANS = ANS * I
60 NEXT I
70 PRINT ANS
Batch (NT):
#echo off
set n=%1
set result=1
for /l %%i in (%n%, -1, 1) do (
set /a result=result * %%i
)
echo %result%
Usage:
C:>factorial.bat 15
F#: Functional
Straight forward:
let rec fact x =
if x < 0 then failwith "Invalid value."
elif x = 0 then 1
else x * fact (x - 1)
Getting fancy:
let fact x = [1 .. x] |> List.fold_left ( * ) 1
Recursive Prolog
fac(0,1).
fac(N,X) :- N1 is N -1, fac(N1, T), X is N * T.
Tail Recursive Prolog
fac(0,N,N).
fac(X,N,T) :- A is N * X, X1 is X - 1, fac(X1,A,T).
fac(N,T) :- fac(N,1,T).
ruby recursive
(factorial=Hash.new{|h,k|k*h[k-1]})[1]=1
usage:
factorial[5]
=> 120
Scheme
Here is a simple recursive definition:
(define (factorial x)
(if (= x 0) 1
(* x (factorial (- x 1)))))
In Scheme tail-recursive functions use constant stack space. Here is a version of factorial that is tail-recursive:
(define factorial
(letrec ((fact (lambda (x accum)
(if (= x 0) accum
(fact (- x 1) (* accum x))))))
(lambda (x)
(fact x 1))))
Oddball examples? What about using the gamma function! Since, Gamma n = (n-1)!.
OCaml: Using Gamma
let rec gamma z =
let pi = 4.0 *. atan 1.0 in
if z < 0.5 then
pi /. ((sin (pi*.z)) *. (gamma (1.0 -. z)))
else
let consts = [| 0.99999999999980993; 676.5203681218851; -1259.1392167224028;
771.32342877765313; -176.61502916214059; 12.507343278686905;
-0.13857109526572012; 9.9843695780195716e-6; 1.5056327351493116e-7;
|]
in
let z = z -. 1.0 in
let results = Array.fold_right
(fun x y -> x +. y)
(Array.mapi
(fun i x -> if i = 0 then x else x /. (z+.(float i)))
consts
)
0.0
in
let x = z +. (float (Array.length consts)) -. 1.5 in
let final = (sqrt (2.0*.pi)) *.
(x ** (z+.0.5)) *.
(exp (-.x)) *. result
in
final
let factorial_gamma n = int_of_float (gamma (float (n+1)))
Freshman Haskell programmer
fac n = if n == 0
then 1
else n * fac (n-1)
Sophomore Haskell programmer, at MIT
(studied Scheme as a freshman)
fac = (\(n) ->
(if ((==) n 0)
then 1
else ((*) n (fac ((-) n 1)))))
Junior Haskell programmer
(beginning Peano player)
fac 0 = 1
fac (n+1) = (n+1) * fac n
Another junior Haskell programmer
(read that n+k patterns are “a disgusting part of Haskell” [1]
and joined the “Ban n+k patterns”-movement [2])
fac 0 = 1
fac n = n * fac (n-1)
Senior Haskell programmer
(voted for Nixon Buchanan Bush — “leans right”)
fac n = foldr (*) 1 [1..n]
Another senior Haskell programmer
(voted for McGovern Biafra Nader — “leans left”)
fac n = foldl (*) 1 [1..n]
Yet another senior Haskell programmer
(leaned so far right he came back left again!)
-- using foldr to simulate foldl
fac n = foldr (\x g n -> g (x*n)) id [1..n] 1
Memoizing Haskell programmer
(takes Ginkgo Biloba daily)
facs = scanl (*) 1 [1..]
fac n = facs !! n
Pointless (ahem) “Points-free” Haskell programmer
(studied at Oxford)
fac = foldr (*) 1 . enumFromTo 1
Iterative Haskell programmer
(former Pascal programmer)
fac n = result (for init next done)
where init = (0,1)
next (i,m) = (i+1, m * (i+1))
done (i,_) = i==n
result (_,m) = m
for i n d = until d n i
Iterative one-liner Haskell programmer
(former APL and C programmer)
fac n = snd (until ((>n) . fst) (\(i,m) -> (i+1, i*m)) (1,1))
Accumulating Haskell programmer
(building up to a quick climax)
facAcc a 0 = a
facAcc a n = facAcc (n*a) (n-1)
fac = facAcc 1
Continuation-passing Haskell programmer
(raised RABBITS in early years, then moved to New Jersey)
facCps k 0 = k 1
facCps k n = facCps (k . (n *)) (n-1)
fac = facCps id
Boy Scout Haskell programmer
(likes tying knots; always “reverent,” he
belongs to the Church of the Least Fixed-Point [8])
y f = f (y f)
fac = y (\f n -> if (n==0) then 1 else n * f (n-1))
Combinatory Haskell programmer
(eschews variables, if not obfuscation;
all this currying’s just a phase, though it seldom hinders)
s f g x = f x (g x)
k x y = x
b f g x = f (g x)
c f g x = f x g
y f = f (y f)
cond p f g x = if p x then f x else g x
fac = y (b (cond ((==) 0) (k 1)) (b (s (*)) (c b pred)))
List-encoding Haskell programmer
(prefers to count in unary)
arb = () -- "undefined" is also a good RHS, as is "arb" :)
listenc n = replicate n arb
listprj f = length . f . listenc
listprod xs ys = [ i (x,y) | x<-xs, y<-ys ]
where i _ = arb
facl [] = listenc 1
facl n#(_:pred) = listprod n (facl pred)
fac = listprj facl
Interpretive Haskell programmer
(never “met a language” he didn't like)
-- a dynamically-typed term language
data Term = Occ Var
| Use Prim
| Lit Integer
| App Term Term
| Abs Var Term
| Rec Var Term
type Var = String
type Prim = String
-- a domain of values, including functions
data Value = Num Integer
| Bool Bool
| Fun (Value -> Value)
instance Show Value where
show (Num n) = show n
show (Bool b) = show b
show (Fun _) = ""
prjFun (Fun f) = f
prjFun _ = error "bad function value"
prjNum (Num n) = n
prjNum _ = error "bad numeric value"
prjBool (Bool b) = b
prjBool _ = error "bad boolean value"
binOp inj f = Fun (\i -> (Fun (\j -> inj (f (prjNum i) (prjNum j)))))
-- environments mapping variables to values
type Env = [(Var, Value)]
getval x env = case lookup x env of
Just v -> v
Nothing -> error ("no value for " ++ x)
-- an environment-based evaluation function
eval env (Occ x) = getval x env
eval env (Use c) = getval c prims
eval env (Lit k) = Num k
eval env (App m n) = prjFun (eval env m) (eval env n)
eval env (Abs x m) = Fun (\v -> eval ((x,v) : env) m)
eval env (Rec x m) = f where f = eval ((x,f) : env) m
-- a (fixed) "environment" of language primitives
times = binOp Num (*)
minus = binOp Num (-)
equal = binOp Bool (==)
cond = Fun (\b -> Fun (\x -> Fun (\y -> if (prjBool b) then x else y)))
prims = [ ("*", times), ("-", minus), ("==", equal), ("if", cond) ]
-- a term representing factorial and a "wrapper" for evaluation
facTerm = Rec "f" (Abs "n"
(App (App (App (Use "if")
(App (App (Use "==") (Occ "n")) (Lit 0))) (Lit 1))
(App (App (Use "*") (Occ "n"))
(App (Occ "f")
(App (App (Use "-") (Occ "n")) (Lit 1))))))
fac n = prjNum (eval [] (App facTerm (Lit n)))
Static Haskell programmer
(he does it with class, he’s got that fundep Jones!
After Thomas Hallgren’s “Fun with Functional Dependencies” [7])
-- static Peano constructors and numerals
data Zero
data Succ n
type One = Succ Zero
type Two = Succ One
type Three = Succ Two
type Four = Succ Three
-- dynamic representatives for static Peanos
zero = undefined :: Zero
one = undefined :: One
two = undefined :: Two
three = undefined :: Three
four = undefined :: Four
-- addition, a la Prolog
class Add a b c | a b -> c where
add :: a -> b -> c
instance Add Zero b b
instance Add a b c => Add (Succ a) b (Succ c)
-- multiplication, a la Prolog
class Mul a b c | a b -> c where
mul :: a -> b -> c
instance Mul Zero b Zero
instance (Mul a b c, Add b c d) => Mul (Succ a) b d
-- factorial, a la Prolog
class Fac a b | a -> b where
fac :: a -> b
instance Fac Zero One
instance (Fac n k, Mul (Succ n) k m) => Fac (Succ n) m
-- try, for "instance" (sorry):
--
-- :t fac four
Beginning graduate Haskell programmer
(graduate education tends to liberate one from petty concerns
about, e.g., the efficiency of hardware-based integers)
-- the natural numbers, a la Peano
data Nat = Zero | Succ Nat
-- iteration and some applications
iter z s Zero = z
iter z s (Succ n) = s (iter z s n)
plus n = iter n Succ
mult n = iter Zero (plus n)
-- primitive recursion
primrec z s Zero = z
primrec z s (Succ n) = s n (primrec z s n)
-- two versions of factorial
fac = snd . iter (one, one) (\(a,b) -> (Succ a, mult a b))
fac' = primrec one (mult . Succ)
-- for convenience and testing (try e.g. "fac five")
int = iter 0 (1+)
instance Show Nat where
show = show . int
(zero : one : two : three : four : five : _) = iterate Succ Zero
Origamist Haskell programmer
(always starts out with the “basic Bird fold”)
-- (curried, list) fold and an application
fold c n [] = n
fold c n (x:xs) = c x (fold c n xs)
prod = fold (*) 1
-- (curried, boolean-based, list) unfold and an application
unfold p f g x =
if p x
then []
else f x : unfold p f g (g x)
downfrom = unfold (==0) id pred
-- hylomorphisms, as-is or "unfolded" (ouch! sorry ...)
refold c n p f g = fold c n . unfold p f g
refold' c n p f g x =
if p x
then n
else c (f x) (refold' c n p f g (g x))
-- several versions of factorial, all (extensionally) equivalent
fac = prod . downfrom
fac' = refold (*) 1 (==0) id pred
fac'' = refold' (*) 1 (==0) id pred
Cartesianally-inclined Haskell programmer
(prefers Greek food, avoids the spicy Indian stuff;
inspired by Lex Augusteijn’s “Sorting Morphisms” [3])
-- (product-based, list) catamorphisms and an application
cata (n,c) [] = n
cata (n,c) (x:xs) = c (x, cata (n,c) xs)
mult = uncurry (*)
prod = cata (1, mult)
-- (co-product-based, list) anamorphisms and an application
ana f = either (const []) (cons . pair (id, ana f)) . f
cons = uncurry (:)
downfrom = ana uncount
uncount 0 = Left ()
uncount n = Right (n, n-1)
-- two variations on list hylomorphisms
hylo f g = cata g . ana f
hylo' f (n,c) = either (const n) (c . pair (id, hylo' f (c,n))) . f
pair (f,g) (x,y) = (f x, g y)
-- several versions of factorial, all (extensionally) equivalent
fac = prod . downfrom
fac' = hylo uncount (1, mult)
fac'' = hylo' uncount (1, mult)
Ph.D. Haskell programmer
(ate so many bananas that his eyes bugged out, now he needs new lenses!)
-- explicit type recursion based on functors
newtype Mu f = Mu (f (Mu f)) deriving Show
in x = Mu x
out (Mu x) = x
-- cata- and ana-morphisms, now for *arbitrary* (regular) base functors
cata phi = phi . fmap (cata phi) . out
ana psi = in . fmap (ana psi) . psi
-- base functor and data type for natural numbers,
-- using a curried elimination operator
data N b = Zero | Succ b deriving Show
instance Functor N where
fmap f = nelim Zero (Succ . f)
nelim z s Zero = z
nelim z s (Succ n) = s n
type Nat = Mu N
-- conversion to internal numbers, conveniences and applications
int = cata (nelim 0 (1+))
instance Show Nat where
show = show . int
zero = in Zero
suck = in . Succ -- pardon my "French" (Prelude conflict)
plus n = cata (nelim n suck )
mult n = cata (nelim zero (plus n))
-- base functor and data type for lists
data L a b = Nil | Cons a b deriving Show
instance Functor (L a) where
fmap f = lelim Nil (\a b -> Cons a (f b))
lelim n c Nil = n
lelim n c (Cons a b) = c a b
type List a = Mu (L a)
-- conversion to internal lists, conveniences and applications
list = cata (lelim [] (:))
instance Show a => Show (List a) where
show = show . list
prod = cata (lelim (suck zero) mult)
upto = ana (nelim Nil (diag (Cons . suck)) . out)
diag f x = f x x
fac = prod . upto
Post-doc Haskell programmer
(from Uustalu, Vene and Pardo’s “Recursion Schemes from Comonads” [4])
-- explicit type recursion with functors and catamorphisms
newtype Mu f = In (f (Mu f))
unIn (In x) = x
cata phi = phi . fmap (cata phi) . unIn
-- base functor and data type for natural numbers,
-- using locally-defined "eliminators"
data N c = Z | S c
instance Functor N where
fmap g Z = Z
fmap g (S x) = S (g x)
type Nat = Mu N
zero = In Z
suck n = In (S n)
add m = cata phi where
phi Z = m
phi (S f) = suck f
mult m = cata phi where
phi Z = zero
phi (S f) = add m f
-- explicit products and their functorial action
data Prod e c = Pair c e
outl (Pair x y) = x
outr (Pair x y) = y
fork f g x = Pair (f x) (g x)
instance Functor (Prod e) where
fmap g = fork (g . outl) outr
-- comonads, the categorical "opposite" of monads
class Functor n => Comonad n where
extr :: n a -> a
dupl :: n a -> n (n a)
instance Comonad (Prod e) where
extr = outl
dupl = fork id outr
-- generalized catamorphisms, zygomorphisms and paramorphisms
gcata :: (Functor f, Comonad n) =>
(forall a. f (n a) -> n (f a))
-> (f (n c) -> c) -> Mu f -> c
gcata dist phi = extr . cata (fmap phi . dist . fmap dupl)
zygo chi = gcata (fork (fmap outl) (chi . fmap outr))
para :: Functor f => (f (Prod (Mu f) c) -> c) -> Mu f -> c
para = zygo In
-- factorial, the *hard* way!
fac = para phi where
phi Z = suck zero
phi (S (Pair f n)) = mult f (suck n)
-- for convenience and testing
int = cata phi where
phi Z = 0
phi (S f) = 1 + f
instance Show (Mu N) where
show = show . int
Tenured professor
(teaching Haskell to freshmen)
fac n = product [1..n]
D Templates: Functional
template factorial(int n : 1)
{
const factorial = 1;
}
template factorial(int n)
{
const factorial =
n * factorial!(n-1);
}
or
template factorial(int n)
{
static if(n == 1)
const factorial = 1;
else
const factorial =
n * factorial!(n-1);
}
Used like this:
factorial!(5)
Java 1.6: recursive, memoized (for subsequent calls)
private static Map<BigInteger, BigInteger> _results = new HashMap()
public static BigInteger factorial(BigInteger n){
if (0 >= n.compareTo(BigInteger.ONE))
return BigInteger.ONE.max(n);
if (_results.containsKey(n))
return _results.get(n);
BigInteger result = factorial(n.subtract(BigInteger.ONE)).multiply(n);
_results.put(n, result);
return result;
}
PowerShell
function factorial( [int] $n )
{
$result = 1;
if ( $n -gt 1 )
{
$result = $n * ( factorial ( $n - 1 ) )
}
$result
}
Here's a one-liner:
$n..1 | % {$result = 1}{$result *= $_}{$result}
Bash: Recursive
In bash and recursive, but with the added advantage that it deals with each iteration in a new process. The max it can calculate is !20 before overflowing, but you can still run it for big numbers if you don't care about the answer and want your system to fall over ;)
#!/bin/bash
echo $(($1 * `( [[ $1 -gt 1 ]] && ./$0 $(($1 - 1)) ) || echo 1`));

Resources