Ethiopian multiplication in Perl6

Post on 11-Jun-2015

1.033 views 4 download

Tags:

description

Perl6 introduces a variety of tools for functional programming and generating readable code in all cases. Among them parameter declarations and lazy lists. This talk looks at how to get Perl6, where to find examples on RakudoCode, and how to use the tools for converting an algorithm from imperative to functional code using Perl6, and parallel dispatch with the ">>" operator.

Transcript of Ethiopian multiplication in Perl6

Hyper-Multiplying Ethiopians:Lambdas, Lazyness, & Perl6

Steven LembarkWorkhorse Computing

lembark@wrkhors.com

Welcome to Perl6

Rakudo-Star distro.

Rosetta Code.

Ethiopic Multiplication:

What it is.

More than one way to do it.

Starting with imperative approach.

Perl6 for functional coding?

Rakudo Star: Perl6 in a tarball

Pretty much what you are used to:

Snag a tarball.

perl Configure

make

tests

install

Acquiring Rakudo-Star

Call it “Rakudo” from now on.http://rakudo.org/

Bi-monthly updates are in ./downloads/starwget .../rakudo-star-2014.01.tar.gz;

gzip -dc < *.tar.gz | tar xf -;

cd rakudo-star-2014.01;

less README;

Acquiring Rakudo-Star

Call it “Rakudo” from now on.http://rakudo.org/

Bi-monthly updates are in ./downloads/starwget .../rakudo-star-2013.10.tar.gz;

gzip -dc < *.tar.gz | tar xf -;

cd rakudo-star-2013.10;

perl Configure.pl –build-parrot;

gmake;

gmake rakudo-test;

gmake install;

Information on Perl6

rakudo.org has links to everything else Perl6-ish.

Rakudo's own doc's page is at:

<http://rakudo.org/documentation>

Also the Perl Maven: Gabor Szabo

<http://perl6maven.com/>

For example:

./parsing-command-line-arguments-perl6

Rosetta Code

Wonderful site, if you care about languages.

Various algorithms, variety of languages.

Perl6 versions largely coded by Larry Wall.

Ethiopian Multiplicationhttp://rosettacode.org/wiki/Ethiopian_Multiplication

Developed in the time of roman numerals:

MCMXIV * XXXIIV = ??

Using integer math:

Halve the first number to 1, doubling the second one.

Sum doubled numbers if the halved ones are odd.

Maybe an example will help...

17 x 34 write down two numbers...

Maybe an example will help...

17 x 34

8 halve the first number to 1

4 using integer math...

2

1

Maybe an example will help...

17 x 34

8 68

4 136 double the second number...

2 272

1 544

Maybe an example will help...

17 34

8 68

4 136

2 272

1 544 sum the doubles with odd halves...

Maybe an example will help...

17 34

8 68

4 136

2 272

1 544

578 and you have the product.

Plan 1: Imperative programming

Loop over the halved value.

Doubling the other.

Summing as you go.

What it looks like in Perl5

sub halve { int((shift) / 2) }sub double { (shift) * 2 }sub iseven { ((shift) & 1) == 0 }

sub ethiopicmult{

my ($plier, $plicand) = @_;my $r = 0;while ($plier >= 1){

$r += $plicand unless iseven $plier;$plier = halve $plier;$plicand = double $plicand;

}return $r;

}

Perl6

More ways to do it:

Parameter types.

Local functions.

Stand-alone subs

sub halve  (Int $n is rw)    { $n div= 2 } # integer dividesub double (Int $n is rw)    { $n *= 2  }sub even   (Int $n ­­> Bool) { $n %% 2  } # is­divisible

sub ethiopicmult(

Int $a is copy, Int $b is copy ­­> Int){

my $r = 0;while $a{

even   $a or $r += $b;halve  $a;double $b;

}return $r;

}

Embedded subssub ethiopicmult(

Int $a is copy, Int $b is copy ­­> Int){

sub halve  (Int $n is rw)    { $n div= 2 }; # subs are sub double (Int $n is rw)    { $n *= 2  }; # lexicallysub even   (Int $n ­­> Bool) { $n %% 2  }; # scopedmy $r = 0;

while $a{

even   $a or $r += $b;halve  $a;double $b;

}return $r;

}

Self-contained with lexical subssub ethiopicmult(

Int $a is copy, Int $b is copy ­­> Int){

my &halve = * div= 2; # placeholders savemy &double = * *=   2; # syntax.my &even = * %%   2;my  $r = 0;

loop # a.k.a. for(;;){

even   $a or $r += $b;halve  $a or return $r;double $b;

}}

Self-contained with lexical subssub ethiopicmult(

Int $a is copy, Int $b is copy ­­> Int){

state &halve = * div= 2; # “state” avoids state &double = * *=   2; # re­compiling.state &even = * %%   2;my $r = 0;

loop{

even   $a or $r += $b;halve  $a or return $r;double $b;

}}

Even more than...

Functional Programming:

Yet Another Way to Avoid Spaghetti Code.

Neither “structured programming” nor “objects”.

Manage issues with state and side effects.

The Evil State

State is hard to maintain.

Coding errors: loosing, failed, multiple updates...

Avoiding it reduces errors.

Unintended Consequences

One bad paste:    loop

    {

      double $b;

      even   $a or $r += $b;

      halve  $a or return $r;

      double $b;

    }

“Functional” programming

e.g., Haskell, Scheme, Scala, Clojure

Declarative: describe the answer more so than steps.

(e.g., order of execution in Haskell is derived).

Ideal: pure functional code

Fully deterministic function calls.

No surprises.

Easy to test, with full validation.

Catch: it doesn't work

Examples?

Current time.

Database query.

Random number.

User input.

Result: Be realistic in applying the theory.

Isolate state, don't eliminate it.

Looking at it in Perl6

Replace steps with declarations.

Don't say “how” say “what”.

Declarative definition

sum a list ...

Declarative definition

sum a list ...

selecting the doubled values whose halved entry is odd ...

Declarative definition

sum a list ...

selecting the doubled values whose halved entry is odd ...

from a list of pairs ...

Declarative definition

sum a list ...

selecting the doubled values whose halved entry is odd ...

from a list of pairs ...

halving the first value to one,

doubling the second one each time.

Iterators & maps & zips, oh my!sub ethiopicmult( Int $a is copy, Int $b is copy ­­> Int ){

state &halve = * div= 2;state &double = * *=   2;state &odd = * %    2;

[+] # iterate '+'.map # map pulls two values each time.{

$^col_2 if odd $^col_1 # parameters extract in lexical order.},zip # new list from the list arguments(

$a, &halve  ... 1 ; # semi­colon separates lists$b, &double ... * # indefinate lazy list

);}

Inline list generators replace subs# what you see is all you get:

sub ethiopicmult( Int $a is copy, Int $b is copy ­­> Int ){

[+]map ­> $half, $dbl { $dbl if $half % 2 },zip(

$a, * div 2 ... 0 ; # implicit subs$b, * *   2 ... *

)}

Result

Minimal, declarative syntax.

Avoid order-of-execution errors.

Hopefully more descriptive, easier to maintain.

Result

Minimal, declarative syntax.

Avoid order-of-execution errors.

Hopefully more descriptive, easier to maintain.

But wait!

There's More!

What you are used to

Processing a list with map:

@result = map { $_->$method( @argz ) } @inputs;

- Single-threaded.

- Fixed order of input.

- Fixed order of processing.

Perl6 adds “hyperoperators”

The “»” or “>>” executes its arguments in parallel:

@result = @objects».&function( @args );

@objects might be blessed or class, literal or variable.

@result is stored in the same order as @objects.

The order of execution is indeterminate.

Zipping pairs, not flat lists.“zip” generates a flat list, we need single vlaues.

Infix “Z=>” normally used for hashes:

my %hash =

( ( $a, * div 2 … 0 ) Z=> ( $b, * * 2 … * ) );

Result is a list of key:value “pairs”:( ( 17, 34 ) ; ( ( 8, 68 ) ; … ) 

Processing pairs

Pair object has a “key” and “value”:

sub odd_values( $pair )

{

$pair.key % 2  ?? $pair.value !! () }

Which leaves us with the values that have to be summed.

Generating the list of odd values

The result of calling “odd_values” is summed just as before:

[+] 

( ( $a, * div 2 … 0 ) Z=> ( $b, * * 2 … * ) )\ ».odd_values

Selection of odd values is parallel.

Generation of pairs and sum are single-threaded.

Reading on the ceiling

Catch: map-ish code reads “upside down”.

Ever wish you could read it going down the page?

Standing on your own two feet

Wish no more: the “Feed operator” pushes the list downhill:

( ( $a, * div 2 … 0 ) Z=> ( $b, * * 2 … * ) )\ ».odd_values ==> [+]

The feed operator shows the direction of data flow.

Feeds odds_values parallel results to block-iterator on '+'.

The result

sub ethiopic_mult( $a, $b )

{

    sub odd_values( $pair ) 

    { $pair.key % 2 ?? $pair.value :: () }

    (( $a, * div 2 ... 0 ) Z=> ( $b, * * 2 ... ) ))\

    >>.&odd_values ==> [+]

}

Hyper-whybotherators?

“$a % 2” is low overhead.

Say you were comparing DNA or proteins...

… running a complex financial calculation...

… playing conway's game of life?

Wouldn't it be nice to queue the operations without loops, forks, breaks, joins, co‑routines, or queues?

What's this got to do with FP?

Hyperoperators lack repeatable sequence.

No way to know order of side effects or state updates.

Manipulating state or side effects don't work.

Simple fix: Use FP [-ish] functions for hyperoperators.

New things in Perl6

%% is “Divisible By” (vs. % for “modulo”).

my &subname = ... for lexically defined subs.

* for parameters.

$^foo for block parameters.

Integer division with 'div'.

Parameterized lists with $start, OP, $end.

List operators with [ OP ].

Hyperopertors & friends.

References

Monads in Perl5:

<http://web.archive.org/web/20080515195640/http://sleepingsquirrel.org/monads/monads.html>

Introduces the basic ideas of functional programming with monads using Perl5 examples.

References

Readable descriptions of “Monad”:

<https://stackoverflow.com/questions/44965/what-is-a-monad>

<https://stackoverflow.com/questions/2704652/monad-in-plain-english-for-the-oop-programmer-with-no-fp-background>

References

What databases mean to Functional Programming:

<https://stackoverflow.com/questions/8406261/most-common-pattern-for-using-a-database-in-a-functional-language-given-desire>

“The most common pattern for dealing with side-effects and impurity in functional languages is:

- be pragmatic, not a purist ...”

References

Realistic view of FP:

<https://stackoverflow.com/questions/330371/are-databases-and-functional-programming-at-odds>

“Functional languages do not have the goal to remain stateless, they have the goal to make management of state explicit.”

References

Good overview of the vocabulary:

<http://en.wikipedia.org/wiki/Functional_Programming>

References

Academic description of Monads:

<http://web.archive.org/web/20071128090030/http://homepages.inf.ed.ac.uk/wadler/papers/marktoberdorf/baastad.pdf>