Prime reciprocal sum: Difference between revisions
Content deleted Content added
m moved |
Thundergnat (talk | contribs) m Promote to full task. Over a year as a draft, multiple implementations, little controversy. |
||
(26 intermediate revisions by 11 users not shown) | |||
Line 1:
{{
Generate the sequence of primes where each term is the smallest prime whose reciprocal can be added to the cumulative sum and remain smaller than '''1'''.
Line 15:
;Task
* Find and display the first '''10''' terms of the sequence. (Or as many as reasonably supported by your language if it is less.) For any values with more than 40 digits, show the first and last 20 digits and the overall digit count.
If any of the tests for primality used in
your program are probabilistic please so indicate.
Line 28 ⟶ 30:
=={{header|ALGOL 68}}==
{{works with|ALGOL 68G|Any - tested with release 2.8.3.win32}}
Uses Algol 68G's LONG LONG INT which has programmer defined precision.<br>
Re-uses code from the [[Arithmetic/Rational]] task, modified to use LONG LONG INT.<br/>
Also uses Miller-Rabin (probabilistic) primality testing.
{{libheader|ALGOL 68-primes}}
<syntaxhighlight lang="algol68">
BEGIN # find a sequence of primes whose members are the smallest prime whose #
# reciprocal can be added to the sum or the reciprocals of the #
# previous primes and the sum remain below 1 #
PR precision 5000 PR # set the precision of LONG LONG INT #
PR read "primes.incl.a68" PR # include prime utilities #
# iterative Greatest Common Divisor routine, returns the gcd of m and n #
PROC gcd = ( LONG LONG INT m, n )LONG LONG INT:
BEGIN
LONG LONG INT a := ABS m, b := ABS n;
WHILE b /= 0 DO
LONG LONG INT new a = b;
b := a MOD b;
a := new a
OD;
a
END # gcd # ;
# code from the Arithmetic/Rational task modified to use LONG LONG INT #
MODE FRAC = STRUCT( LONG LONG INT num #erator#, den #ominator# );
PROC lcm = ( LONG LONG INT a, b )LONG LONG INT: # least common multiple #
a OVER gcd(a, b) * b;
PRIO // = 9; # higher then the ** operator #
OP // = ( LONG LONG INT num, den )FRAC: ( # initialise and normalise #
LONG LONG INT common = gcd( num, den );
IF den < 0 THEN
( -num OVER common, -den OVER common )
ELSE
( num OVER common, den OVER common )
FI
);
OP + = (FRAC a, b)FRAC: (
LONG LONG INT common = lcm( den OF a, den OF b );
FRAC result := ( common OVER den OF a * num OF a + common OVER den OF b * num OF b, common );
num OF result//den OF result
);
OP +:= = (REF FRAC a, FRAC b)REF FRAC: ( a := a + b );
# end code from the Arithmetic/Rational task modified to use LONG LONG INT #
# the sequence starts with the reciprocal of the first prime > 0, i.e. 2 #
LONG LONG INT one = 1;
FRAC sum := one // LONG LONG 2;
print( ( " 1: 2", newline ) );
# each succeeding prime is the next prime > the denominator of the sum #
# divided by the difference of the denominator and the numerator #
FOR i FROM 2 TO 15 DO
LONG LONG INT next := IF num OF sum + 1 = den OF sum
THEN den OF sum + 1
ELSE ( den OF sum OVER ( den OF sum - num OF sum ) ) + 1
FI;
IF NOT ODD next THEN next +:= 1 FI;
WHILE NOT is probably prime( next ) DO next +:= 2 OD;
print( ( whole( i, -2 ), ": " ) );
STRING prime text = whole( next, 0 );
IF INT digits = ( UPB prime text - LWB prime text ) + 1;
digits <= 40
THEN
print( ( prime text ) )
ELSE
print( ( prime text[ : LWB prime text + 19 ], "..."
, prime text[ UPB prime text - 19 : ], " "
, whole( digits, -6 ), " digits"
)
)
FI;
print( ( newline ) );
sum +:= one // next
OD
END
</syntaxhighlight>
{{out}}
<pre>
1: 2
2: 3
3: 7
4: 43
5: 1811
6: 654149
7: 27082315109
8: 153694141992520880899
9: 337110658273917297268061074384231117039
10: 84241975970641143191...13803869133407474043 76 digits
11: 20300753813848234767...91313959045797597991 150 digits
12: 20323705381471272842...21649394434192763213 297 digits
13: 12748246592672078196...20708715953110886963 592 digits
14: 46749025165138838243...65355869250350888941 1180 digits
15: 11390125639471674628...31060548964273180103 2358 digits
</pre>
=={{header|C}}==
Line 138 ⟶ 244:
11390125639471674628..31060548964273180103 (2358 digits)
</syntaxhighlight>
Here, <code>+/@:%</code> is the sum of reciprocals, so <code>1%1-+/@:%</code> is the reciprocal of the amount remaining, and <code>4 p:1%1-+/@:%</code> is the smallest prime which is larger than that value.
Tested in J9.4
=={{header|jq}}==
'''Works with jq, the C implementation of jq''' (*)
'''Works with gojq, the Go implementation of jq''' (**)
(*) Using the two programs presented here, the C implementation
cannot get beyond the 7th prime in the series because it lacks
support for both "big integers" and "big floats".
(**) The `nextPrime` algorithm effectively prevents the Go
implementation from generating more than eight primes in the sequence.
gojq does support infinite-precision integer arithmetic, which in
principle should allow the first program, which uses rational number
arithmetic, to proceed indefinitely but for the slowness of
`nextPrime`.
==={{header|Using rational.jq}}===
The "rational" module referenced here is available at
[[:Category:jq/rational.jq]].
<syntaxhighlight lang="jq">
include "rational" {search:"."}; # see above
# itrunc/0 attempts to compute the "trunc" of the input number in such
# a way that gojq will recognize the result as an integer, while
# leveraging the support that both the C and Go implementations have
# for integer literals.
# It is mainly intended for numbers for which the tostring value does NOT contain an exponent.
# The details are complicated because the goal is to provide portability between jq implementations.
#
# Input: a number or a string matching ^[0-9]+([.][0-9]*)$
# Output: as specified by the implementation.
def itrunc:
if type == "number" and . < 0 then - ((-.) | itrunc)
else . as $in
| tostring as $s
| ($s | test("[Ee]")) as $exp
| ($s | test("[.]")) as $decimal
| if ($exp|not) and ($decimal|not) then $s | tonumber # do not simply return $in
elif ($exp|not) and $decimal then $s | sub("[.].*";"") | tonumber
else trunc
| tostring
| if test("[Ee]")
then if $exp then "itrunc cannot be used on \($in)" | error end
else tonumber
end
end
end;
# rtrunc is like trunc but for Rational numbers, though the input may be
# either a number or a Rational.
def rtrunc:
if type == "number" then r(itrunc;1)
elif 0 == .n or (0 < .n and .n < .d) then r(0;1)
elif 0 < .n or (.n % .d == 0) then .d as $d | r(.n | idivide($d); 1)
else rminus( r( - .n; .d) | rtrunc | rminus; 1)
end;
def is_prime:
. as $n
| if ($n < 2) then false
elif ($n % 2 == 0) then $n == 2
elif ($n % 3 == 0) then $n == 3
elif ($n % 5 == 0) then $n == 5
elif ($n % 7 == 0) then $n == 7
elif ($n % 11 == 0) then $n == 11
elif ($n % 13 == 0) then $n == 13
elif ($n % 17 == 0) then $n == 17
elif ($n % 19 == 0) then $n == 19
else sqrt as $s
| 23
| until( . > $s or ($n % . == 0); . + 2)
| . > $s
end;
# Input: any number
# Output: the next prime
def nextPrime:
if . < 2 then 2
else itrunc # for gojq (to ensure result is a Go integer)
| (if .%2 == 1 then .+2 else .+1 end) as $n
| first( range($n; infinite; 2) | select(is_prime))
end;
# Input: a Rational
# Output: an integer
def rnextPrime:
if rlessthan(r(2;1)) then 2
else rtrunc
| .n
| nextPrime
end ;
def synopsis:
tostring as $i
| ($i|length) as $n
| if $n > 40
then "\($i[:20]) ... \($i[20:]) (\($n))"
else $i
end;
def prs_sequence:
r(1;1) as $one
| {p: 0, s: r(0;1)}
| recurse(
.emit = null
| .count += 1
| .p = (rminus($one; .s) | rinv | rnextPrime)
| .s = radd(.s; r(1; .p))
| .emit = .p )
| select(.emit).emit ;
16 as $n
| "First \($n) elements of the sequence:",
limit($n; prs_sequence)
</syntaxhighlight>
{{output}}
The following output was generated using gojq.
<pre>
First 16 elements of the sequence:
2
3
7
43
1811
654149
27082315109
153694141992520880899
(execution terminated)
</pre>
==={{header|Using floats}}===
The program is essentially as above, except that the directive for
including the "rational" module can be omitted, and the `prs_sequence`
function is replaced by the following definition:
<syntaxhighlight lang="jq">
def prs_sequence:
{p: 0, s: 0}
| recurse(
.emit = null
| .count += 1
| .p = ((1 / (1 - .s)) | nextPrime)
| .s += (1 / .p)
| if .s >= 1 then "Insufficient precision to proceed"|error end
| .emit = .p )
| select(.emit).emit;
</syntaxhighlight>
<pre>
"First 16 elements of the sequence:"
2
3
7
43
1811
654149
27082315109
(program halted)
</pre>
=={{header|Julia}}==
<syntaxhighlight lang="julia">""" rosettacode.org/wiki/Prime_reciprocal_sum """
Line 187 ⟶ 462:
16: 36961763505630520555..02467094377885929191 (4711 digits)
17: 21043364724798439508..14594683820359204509 (9418 digits)
</pre>
=={{header|Nim}}==
{{libheader|bignum}}
<syntaxhighlight lang="Nim">import std/strformat
import bignum
iterator a075442(): Int =
let One = newRat(1)
var sum = newRat(0)
var p = newInt(0)
while true:
let q = reciprocal(One - sum)
p = nextPrime(if q.isInt: q.num else: q.toInt + 1)
yield p
sum += newRat(1, p)
func compressed(str: string; size: int): string =
## Return a compressed value for long strings of digits.
if str.len <= 2 * size: str
else: &"{str[0..<size]}...{str[^size..^1]} ({str.len} digits)"
var count = 0
for p in a075442():
inc count
echo &"{count:2}: {compressed($p, 20)}"
if count == 15: break
</syntaxhighlight>
{{out}}
<pre> 1: 2
2: 3
3: 7
4: 43
5: 1811
6: 654149
7: 27082315109
8: 153694141992520880899
9: 337110658273917297268061074384231117039
10: 84241975970641143191...13803869133407474043 (76 digits)
11: 20300753813848234767...91313959045797597991 (150 digits)
12: 20323705381471272842...21649394434192763213 (297 digits)
13: 12748246592672078196...20708715953110886963 (592 digits)
14: 46749025165138838243...65355869250350888941 (1180 digits)
15: 11390125639471674628...31060548964273180103 (2358 digits)
</pre>
=={{header|Pascal}}==
==={{header|Free Pascal}}===
Most time consuming is finding the next prime.<br>
Now pre-sieving with the primes til 65535, to reduce tests.<br>
That is the same as checking all numbers in sieve as divisible by the small primes.Since the prime are in big distance in that region, that's an improvement.
<syntaxhighlight lang=pascal>program primeRezSum;
{$IFDEF FPC} {$MODE DELPHI}{$Optimization ON,ALL} {$ENDIF}
{$IFDEF WINDOWS}{$APPTYPE CONSOLE}{$ENDIF}
Line 199 ⟶ 520:
sysutils,
gmp;
const
PrimeCount = 6542;
PRIMEMAXVAL = 65535;
SIEVESIZE = 65536;
type
Tsieveprime = record
prime,
offset : Uint16;
end;
tSievePrimes = array[0..PrimeCount-1] of Tsieveprime;
tSieve = array[0..SIEVESIZE-1] of byte;
var
s : AnsiString;
MyPrimes : tSievePrimes;
sieve : tSieve;
procedure OutStr(const s: AnsiString);
var
Line 215 ⟶ 551:
writeln(pChar(@myString[1]),'...',pChar(@myString[l-19]),' (',l:6,' digits)');
end;
end;
function InitPrimes:Uint32;
var
f : extended;
idx,p,pr_cnt : Uint32;
Begin
fillchar(sieve,Sizeof(sieve),#0);
pr_cnt := 0;
p := 2;
f := 1.0;
repeat
while Sieve[p]<> 0 do
inc(p);
MyPrimes[pr_cnt].prime := p;
f := f*(p-1)/p;
inc(pr_cnt);
idx := p*p;
if idx > PRIMEMAXVAL then
Break;
repeat
sieve[idx] := 1;
inc(idx,p);
until idx > high(sieve);
inc(p);
until sqr(p)>PRIMEMAXVAL;
while (pr_cnt<= High(MyPrimes)) AND (p<PRIMEMAXVAL) do
begin
while Sieve[p]<> 0 do
inc(p);
MyPrimes[pr_cnt].prime := p;
f := f*(p-1)/p;
inc(p);
inc(pr_cnt);
end;
Writeln ('reducing factor ',f:10:8);
result := pr_cnt-1;
end;
procedure DoSieveOffsetInit(var tmp:mpz_t);
var
dummy :mpz_t;
i,j,p : Uint32;
Begin
mpz_init(dummy);
for i := 0 to High(MyPrimes) do
with MyPrimes[i] do
Begin
if prime = 0 then Begin writeln(i);halt;end;
offset := prime-mpz_tdiv_q_ui(dummy,tmp,prime);
end;
mpz_set(dummy,tmp);
repeat
//one sieve
fillchar(sieve,Sizeof(sieve),#0);
//sieve
For i := 0 to High(MyPrimes) do
begin
with MyPrimes[i] do
begin
p := prime;
j := offset;
end;
repeat
sieve[j] := 1;
j += p;
until j >= SIEVESIZE;
MyPrimes[i].offset := j-SIEVESIZE;
end;
j := 0;
For i := 0 to High(sieve) do
begin
// function mpz_probab_prime_p(var n: mpz_t; reps: longint): longint;1 = prime
if (sieve[i]= 0) then
begin
mpz_add_ui(dummy,dummy,i-j);
j := i;
if (mpz_probab_prime_p(dummy,1) >0) then
begin
mpz_set(tmp,dummy);
mpz_clear(dummy);
EXIT;
end;
end;
end;
until false;
end;
var
nominator,denominator,tmp,tmpDemP,p : mpz_t;
T1,T0:Int64;
cnt : NativeUint;
begin
InitPrimes;
setlength(s,100000);
cnt := 1;
mpz_init(nominator);
Line 228 ⟶ 653:
mpz_init(tmpDemP);
mpz_init_set_ui(denominator,1);
mpz_init_set_ui(p,
repeat
T0 := GetTickCount64;
DoSieveOffsetInit(p)
else
mpz_nextprime(p,p);
T1 := GetTickCount64;
write(cnt:3,' ',T1-T0,' ms ,delta ');
mpz_sub(tmpDemP,p,tmpDemP);
mpz_get_str(pChar(@s[1]),10, tmpDemP); OutStr(s);
mpz_get_str(pChar(@s[1]),10, p); OutStr(s);
if cnt >=15 then
Break;
repeat
mpz_mul(tmp,nominator,p);
Line 239 ⟶ 675:
if mpz_cmp(tmp,tmpDemP)< 0 then
BREAK;
until false;
mpz_set(nominator,tmp);
mpz_mul(denominator,denominator,p);
Line 248 ⟶ 682:
//next smallest possible number denominator/delta
mpz_sub(tmp,denominator,nominator);
mpz_fdiv_q(p,denominator,tmp);
inc(cnt);
until cnt>
end.</syntaxhighlight>
{{out|@TIO.RUN}}
<pre>
reducing factor 0.05041709
2
3
7
4 0 ms ,delta 1
43
5 0 ms ,delta 5
1811
6 0 ms ,delta 16
654149
7 0 ms ,delta 5
27082315109
8 0 ms ,delta 71
153694141992520880899
9 1 ms ,delta 14
337110658273917297268061074384231117039
10 1 ms ,delta 350
8424197597064114319...13803869133407474043 ( 76 digits)
11 1 ms ,delta 203
2030075381384823476...91313959045797597991 ( 150 digits)
12 3 ms ,delta 33
2032370538147127284...21649394434192763213 ( 297 digits)
13 69 ms ,delta 348
1274824659267207819...20708715953110886963 ( 592 digits)
14 234 ms ,delta 192
4674902516513883824...65355869250350888941 ( 1180 digits)
15 20391 ms ,delta 3510
1139012563947167462...31060548964273180103 ( 2358 digits)
Real time: 21.024 s User time: 20.768 s Sys. time: 0.067 s CPU share: 99.10 %
@home (4.4Ghz 5600G ) modified with primes up to 1E6 ( Uint16-> Uint32 )
78498 999979 reducing factor 0.04059798
...
15 10015 ms ,delta 3510 // first guess than searching for next prime
1139012563947167462...31060548964273180103 ( 2358 digits)
1731172 999979
16 110030 ms ,delta 6493
3696176350563052055...02467094377885929191 ( 4711 digits)
17 5098268 ms ,delta 55552 // first guess than searching for next prime
2104336472479843950...14594683820359204509 ( 9418 digits)
</pre>
=={{header|Perl}}==
<syntaxhighlight lang="perl" line>
use strict; use warnings; use feature 'state';
use Math::AnyNum <next_prime ceil>;
sub abbr { my $d=shift; my $l = length $d; $l < 41 ? $d : substr($d,0,20) . '..' . substr($d,-20) . " ($l digits)" }
sub succ_prime {
state $sum = 0;
my $next = next_prime ceil( 1 / (1-$sum) );
$sum += 1/$next;
$next
}
printf "%2d: %s\n", $_, abbr succ_prime for 1..14;
</syntaxhighlight>
{{out}}
<pre>
1: 2
2: 3
3: 7
4: 43
5: 1811
6: 654149
7: 27082315109
8: 153694141992520880899
9: 337110658273917297268061074384231117039
10: 84241975970641143191..13803869133407474043 (76 digits)
11: 20300753813848234767..91313959045797597991 (150 digits)
12: 20323705381471272842..21649394434192763213 (297 digits)
13: 12748246592672078196..20708715953110886963 (592 digits)
14: 46749025165138838243..65355869250350888941 (1180 digits)
</pre>
=={{header|Phix}}==
The Julia entry took over 4 hours to get the 17th on this box, so I just went with "keep it all under 10s"...<br>
<small>(As others have alluded, it is all about getting the next prime. Even should you land on one straightaway, it still takes quite some time to prove an n-thousand digit number is ''probably'' prime - the standard prime tests are only deterministic to 3,317,044,064,679,887,385,961,981 but at least the 9th and later results agree with others on this page.</small>
<!--(phixonline)-->
<syntaxhighlight lang="phix">
with javascript_semantics
constant limit = iff(platform()=JS?12:14)
include mpfr.e
mpq {q, r, s, u} = mpq_inits(4,{0,0,0,1})
mpz {p, t} = mpz_inits(2)
printf(1,"First %d elements of the sequence:\n", limit);
for count=1 to limit do
mpq_sub(q, u, s)
mpq_inv(q, q)
mpq_get_den(t, q)
mpz_set_q(p, q)
if mpz_cmp_si(t, 1) then mpz_add_si(p, p, 1) end if
mpz_nextprime(p, p)
printf(1,"%2d: %s\n", {count, mpz_get_short_str(p)})
mpq_set_z(r, p)
mpq_inv(r, r)
mpq_add(s, s, r)
end for
</syntaxhighlight>
{{out}}
<pre>
First 14 elements of the sequence:
1: 2
2: 3
3: 7
4: 43
5: 1811
6: 654149
7: 27082315109
8: 153694141992520880899
9: 337110658273917297268061074384231117039
10: 84241975970641143191...13803869133407474043 (76 digits)
11: 20300753813848234767...91313959045797597991 (150 digits)
12: 20323705381471272842...21649394434192763213 (297 digits)
13: 12748246592672078196...20708715953110886963 (592 digits)
14: 46749025165138838243...65355869250350888941 (1,180 digits)
</pre>
If you've really got nothing better to do, you can also get, in 10 mins on 64-bit, thrice that on 32-bit:
<pre>
15: 11390125639471674628...31060548964273180103 (digits: 2358)
16: 36961763505630520555...02467094377885929191 (digits: 4711)
</pre>
=={{header|Python}}==
<syntaxhighlight lang="python">''' rosettacode.org/wiki/Prime_reciprocal_sum '''
from fractions import Fraction
from sympy import nextprime
def abbreviated(bigstr, label="digits", maxlen=40, j=20):
''' Abbreviate string by showing beginning / end and number of chars '''
wid = len(bigstr)
return bigstr if wid <= maxlen else bigstr[:j] + '..' + bigstr[-j:] + f' ({wid} {label})'
def ceil(rat):
''' ceil function for Fractions '''
return rat.numerator if rat.denominator == 1 else rat.numerator // rat.denominator + 1
psum = Fraction(0, 1)
for i in range(1, 15): # get first 14 in sequence
next_in_seq = nextprime(ceil(Fraction(1, 1 - psum)))
psum += Fraction(1, next_in_seq)
print(f'{i:2}:', abbreviated(str(next_in_seq)))
</syntaxhighlight>{{out}}
<pre>
1: 2
2: 3
3: 7
4: 43
5: 1811
6: 654149
7: 27082315109
8: 153694141992520880899
9: 337110658273917297268061074384231117039
10: 84241975970641143191..13803869133407474043 (76 digits)
11: 20300753813848234767..91313959045797597991 (150 digits)
12: 20323705381471272842..21649394434192763213 (297 digits)
13: 12748246592672078196..20708715953110886963 (592 digits)
14: 46749025165138838243..65355869250350888941 (1180 digits)
</pre>
=={{header|Raku}}==
Line 300 ⟶ 887:
15: 11390125639471674628..31060548964273180103 (2358 digits)
16: 36961763505630520555..02467094377885929191 (4711 digits)</pre>
=={{header|RPL}}==
Limited floating-point precision prevents from finding the correct 7th term. Program starts with a non-empty sequence to avoid the <code>∑LIST</code> calculation bug that occurs when the input list has less than two items.
{{works with|HP|49g}}
≪ {2 3}
1 4 '''START''' 1 OVER INV ∑LIST - INV →NUM IP NEXTPRIME + '''NEXT'''
≫ '<span style="color:blue">∑INVPR</span>' STO
{{out}}
<pre>
1: {2 3 7 43 1811 654149}
</pre>
=={{header|Sidef}}==
<syntaxhighlight lang="ruby">var A075442 = Enumerator({|callback|
var sum = 0
loop {
var p = next_prime(ceil(1/(1-sum)))
sum += 1/p
callback(p)
}
})
A075442.first(15).each_kv {|k,n|
var s = Str(n)
s = "#{s.first(20)}..#{s.last(20)} (#{s.len} digits)" if (s.len > 50)
say "#{'%2d' % k+1}: #{s}"
}</syntaxhighlight>
{{out}}
<pre>
1: 2
2: 3
3: 7
4: 43
5: 1811
6: 654149
7: 27082315109
8: 153694141992520880899
9: 337110658273917297268061074384231117039
10: 84241975970641143191..13803869133407474043 (76 digits)
11: 20300753813848234767..91313959045797597991 (150 digits)
12: 20323705381471272842..21649394434192763213 (297 digits)
13: 12748246592672078196..20708715953110886963 (592 digits)
14: 46749025165138838243..65355869250350888941 (1180 digits)
15: 11390125639471674628..31060548964273180103 (2358 digits)
</pre>
=={{header|Wren}}==
Line 305 ⟶ 937:
{{libheader|Wren-fmt}}
Even with GMP takes about 4½ minutes to find the first 16.
GMP’s mpz_nextprime function (which Wren is calling under the hood) uses a probabilistic algorithm to identify primes but, in practice, the chance of a composite number passing is extremely small and should be nil for numbers less than 2^64.
<syntaxhighlight lang="wren">import "./gmp" for Mpz, Mpq
import "./fmt" for Fmt
|