# Sum of integers in string, separated by non-numericals such as 'a' and 'Y'

Create a program which sums all integers found in a string which is set as a variable in the program (thus, the program doesn't have to handle any input). The integer numbers are separated by non-numericals (anything but 0, 1, 2, 3, ..., 9).

Examples:

• e7rde f ,fe 43 jfj 54f4sD = 7+43+54+4=108
• 5 = 5
• 64 545,5445-32JIFk0ddk = 64+545+5445+32+0=6086
• 0ab0 = 0+0 = 0

Extra notes:

• Unicode support is not necessary, but allowed
• -n (where n is an integer) is not counted as a negative n, but as a hyphen followed by n.

The answer may be printed on the screen (but not required).

• Should we print the result too? (You mention no I/O). Feb 8, 2011 at 19:43
• @Dogbert - I didn't think about that. Sorry, yes. I will update the post.
– Anto
Feb 8, 2011 at 19:45
• Changed it as some people already had answers and didn't want to "hurt" them. I guess I should sleep now, so I will think a bit clearer ;)
– Anto
Feb 8, 2011 at 19:49
• Anto: A task where a solution has no observable side-effects isn't very nice, though.
– Joey
Feb 8, 2011 at 21:00
• An interesting test case I just ran into would be 5a-3 (my code would skip - if it follows a number immediately, but not if there was a non-number before it). Sep 1, 2015 at 11:48

## Ruby 1.9, 21 characters

eval a.scan(/\d+/)*?+


To print the solution to stdout, 2 additional characters are required:

p eval a.scan(/\d+/)*?+


And to read from stdin instead of using a predefined variable, another 3 characters have to be used:

p eval gets.scan(/\d+/)*?+


For Ruby 1.8, replace ?+ with "+" to get a working solution in 22 characters.

• The input is supposed to be taken from a variable, not stdin. Also scan is shorter than split. So your solution becomes eval s.scan(/\d+/)*?+ - 21 characters. Feb 8, 2011 at 21:37
• @sepp2k: Yeah, didn't read the description correctly. I'm just used to the other golf-tasks, where you usually have to read from stdin and print to stdout. Good point with scan, thanks! Feb 8, 2011 at 21:48
• +1, great use of eval and * '+' Feb 8, 2011 at 21:55

## Perl, 15

Input in $_, sum in $c:

s/\d+/$c+=$&/ge


# Ruby - 36 34 chars

s.scan(/\d+/).map(&:to_i).reduce:+


36 chars if you want the result printed.

p s.scan(/\d+/).map(&:to_i).reduce:+


Assumes the input is present as a string in s.

## Python (60)

import re;print sum(map(int,filter(len,re.split(r'\D',s))))


# JavaScript (ES6), 30

c=0,s.replace(/\d+/g,d=>c+=+d)


Annotated version:

// Store the sum.
c=0,
// Process every number found in the s.
s.replace(/\d+/g,
// Convert the number into an integer.
// Add it to the sum.
d => c += +d
)


With output.

$x-replace'\D','+0'|iex  In fact, without output is exactly the same, you'd just pipe it somewhere else where it's needed. # JavaScript [31 bytes] eval(s.match(/\d+/g).join('+'))  • I count it 31 bytes? – l4m2 Jun 16, 2020 at 7:30 • @l4m2 Yes, you are right, fixed. Jun 16, 2020 at 11:20 • I know this is really late but you can save 2 bytes by changing .join("+") into .join+ Aug 16, 2021 at 5:47 ## J - 40 38 characters Lazy version. Requires the string library. +/".(,' ',.~a.-.'0123456789')charsub y  • Supports Unicode. Supports encoding, come to think of it! Feb 8, 2011 at 22:03 ## Java out of the contest ;) public static long sum(String s) { long sum = 0; String p = ""; char[] ch = s.toCharArray(); for (int i = 0; i < ch.length; i++) { boolean c = false; if (Character.isDigit(ch[i])) { if (i + 1 < ch.length) { if (Character.isDigit(ch[i + 1])) { p += ch[i]; c = true; } } if (!c) { p += ch[i]; sum += Integer.valueOf(p); p = ""; c = false; } } } return sum; }  # Labyrinth, 29 21 bytes (Disclaimer: Labyrinth is newer than this challenge.) Also, Labyrinth doesn't have variables, so I went with a normal input/output program. )_"+ ( "?" ";;,;;(!@  This was fairly simple because of the way Labyrinth's input commands work. ? tries to read a signed integer from STDIN and stops at the first non-digit. If it can't read an integer (because the next character is a - not followed by a digit, or any other non-digit, or we've reached EOF), it will return 0 instead. , on the other hand reads any subsequent byte and pushes the byte value. If this one is called at EOF it will return -1 instead. So here's some pseudocode for the solution: running total = 0 while(true) while(true) try reading a non-zero integer N with ? if(N < 0) running total -= N else if(N > 0) running total += N else break // We've either read a zero or hit a something that isn't a number try reading a character with , if(that returned -1) break print running total  Dealing with negative numbers correctly complicates this solution quite a lot. If it weren't for those, I'd have this 8-byte solution: ?+ ;,;!@  # gs2, 4 bytes W#Θd  Encoded in CP437; the third byte is E9. W reads all numbers /-?\d+/ from a string, #Θ maps absolute value, d sums. (gs2, too, is newer than this challenge, but its read-nums command is a total coincidence.) # Perl 5-MList::Util=sum -pF/\D+/, 8 bytes $_=sum@F


Try it online!

# 05AB1E (legacy), 5 bytes

.œþOà


Try it online or verify a few more test cases. (Times out for the larger test cases due to the .œ builtin.)

Explanation:

.œ     # Get all partitions of the (implicit) input-string,
# which are all possible ways of dividing the input-strings into substrings
þ    # Only leave the items consisting of digits for each partition
# (in the new version of 05AB1E, an explicit € is required)
O   # Sum each inner list
à  # Pop and push its maximum
# (after which the result is output implicitly)


dc<<<tr -c 0-9 +<<<$sp  Try it online! Input in the variable s, output on stdout. # Pip, 7 bytes $+a@+XD


Try it online!

Regexes all numbers to a list and sums them.

+1 byte fix from DLosc.

• This fails the third test case because XN includes leading minus signs in the number. You can use +XD instead, for +1 byte. Aug 15, 2021 at 21:45
• Are you sure? If the third testcase counted minuses it'd be 6022, not 6086 Aug 16, 2021 at 2:43
• XN gives 6022; +XD correctly gives 6086. Aug 16, 2021 at 3:23
• ... although, the XN approach would work on this very similar challenge. Aug 16, 2021 at 3:26
• ah yes trivial modification time! Aug 16, 2021 at 4:13

# Vyxals, 6 bytes

\Dṡ⌊


Try it Online!

Ugh we really need to make ⌊ play nice with strings not containing numbers in v2.6; this would be 6 bytes if it didn't error. 2.6 done did make the change lol.

## Explained

\Dṡ⌊
\D   # The string "\D" (a regex that looks for non-digits)
ṡ  # Split the input on this regex
⌊ # Convert each to a number
# Output the sum with the s flag


# PHP - 37

Without printing;

<?array_sum(@split("[^0-9]+",cat));


With printing (38):

<?=array_sum(@split("[^0-9]+",cat));


## Perl, 16 chars

s/\d+/$r+=$&/ge;


Takes input in $_, output goes on $r. Last semicolon is superfluous, but it will probably be needed when the program does more things. Add say$r for output. • Oops, didn't see your exact same answer when I posted. Though I counted one character more even without the semicolon. – J B Feb 8, 2011 at 21:30 • @J B: I can't count! :P. Actually, I made the mistake of echo'ing a double quoted string to wc -c. Feb 8, 2011 at 21:33 ## R, 30 sum(scan(t=gsub("\\D"," ",x)))  Here, x is the name of the variable. Example: > x <- "e7rde f ,fe 43 jfj 54f4sD" > sum(scan(t=gsub("\\D"," ",x))) Read 4 items  108  # J - 23 char Not a winner, but we get to see a fairly rare primitive in action. +/".(,_=_"."0 y)}y,:' '  Explained: • _"."0 y - For each character in the input string y, try to read it in as a number. If you can't, use the default value _ (infinity) instead. • ,_= - Check each result for equality to _, and then run the final array of 0s and 1s into a vector. ("."0 always adds one too many dimensions to the result, so we correct for that here.) • y,:' ' - Add a row of spaces beneath the input string. • } - Used as it is here, } is called Item Amend, and it uses the list of 0s and 1s on the left as indices to select the row to draw from in the right argument. So what happens is, for each column in the right side, we take the original character if it could be read in as a number, and otherwise we take the space beneath it. Hence, we cover up any non-numeric characters with spaces. • +/". - Now convert this entire string into an list of numbers, and sum them. # Brachylog, 9 bytes {~cịˢ+}ᶠ⌉  Try it online! It's... inefficient... so I've neglected to test the longer cases. It appears to work, though. Takes input through the input variable, which just so happens to be the usual form of input anyhow.  ⌉ The output is the largest { }ᶠ possible + sum of ịˢ integers the string representations of which are some elements of ~c a partition of the input.  ⌉ ensures that numbers are maximally matched as well as never negative, since splitting a number up, invalidating a number, or turning it negative will never result in a larger sum. # C 96 Thanks to @ceilingcat for some very nice pieces of golfing - now even shorter t;main(i,v)int**v;{for(char*q,*s=v;*s;q>s?t+=abs(i),s=q:s++)i=strtol(s,&q,0);printf("%d",t);}  Try it online! An earlier 85 byte version that is cheating a bit by hardcoding the string inside the program: t=0;main(i){for(char*q,*s;i=strtol(s,&q,0),*s;q>s?t+=abs(i),s=q:s++);printf("%d",t);}  To actually use the 85 byte program you need to assign the variable like so: t=0;main(i){for(char*q,*s="text";i=strtol(s,&q,0),*s;q>s?t+=abs(i),s=q:s++);printf("%d",t);}  # K4, 21 bytes Solution: +/"J"$" "\:.Q.n .Q.n?


Explanation:

Not particularly pleased with this, but it works...

+/"J"$" "\:.Q.n .Q.n? / the solution .Q.n? / lookup right in the string "0123456789" .Q.n / index back into "0123456789" " "\: / split string on whitespace "J"$                / cast to numeric
+/                    / sum up


Extra:

• +/0{.[{(0,. y)+1 10*x};(x;y);(+/x),0]}/ for 39 bytes which is 'nicer' but almost twice the count...

# Stax, 8 bytes

µÖ▀o)♫⌐ƒ


Run and debug it

Finds all substrings matching \d+, then eval and sum.

# Funky, 38 bytes

s=>{n=0forv ins::gmatch"\\d+"n-=-v}


## Explained

s=>{                                 } $# A function taking argument "s" n=0$# Preset n to 0
s::gmatch"\\d+"          $# An iterator function which finds all groups of digits in a string forv in$# Iterate through, setting v to the match
n-=       $# Subtract n by -v$# The negative of the whole match, which implicitely numberfies it.
$# Conveniently implcit return from last statement in the function.  Try it online! # Japt v2.0a0 -x, 3 bytes q\D  Try it # Husk, 4 bytes ṁiġ√  Try it online! # Explanation ṁiġ√ ġ√ group on isalpha ṁ sum i the first integer in each string  # Smalltalk (Smalltalk/X) (51 chars) using the regex pakage: (s regex:'\d+' matchesCollect:[:n|n asNumber])sum  w.o. regex: ((s asCollectionOfSubCollectionsSeparatedByAnyForWhich:[:c|c isDigit not]) map:#asNumber)sum  input in s ## Javascript - 43 chars I know it's long, but there was no JS solution so :) c=0 a=a.split(/[^\d]/g) for(i in a)c+=+a[i]  ais the string. c contains answer. ## Tcl, 30 expr [regsub -all \\D+$a.0 +]


It assumes that the input is in the variable \$a (formally, in a`) and stores the answer in the interpreter result. I/O is left as an exercise.