Python 2, 39 bytes

lambda x:x+len(`x`)+eval("+".join(`x`)) 

Test suite

Using the same eval-trick as in my Pyth-answer.

05AB1E, 28 20 18 8 bytes

ÐgsS'+ýO 

Explanation

Ð # triplicate input g # get length of input sS'+ý # split input and merge with '+' as separator O # sum and implicitly display 

Try it online

Saved 10 bytes thanks to @Adnan

Pyth, 11 10 bytes

Thanks to @LeakyNun for a byte!

++vj\+`Ql` 

Test suite

Explanation

 ++vj\+`Ql`QQ # Q = input, last two implicitly added vj\+`Q # Join the input on '+' and eval it l`Q # Length of the input Q # The input itself ++ # Add those three values to obtain the result 

CJam, 18

q_,\~__Ab(@g*\~]:+ 

Try it online

Explanation:

q_ read the input and make a copy ,\ get the string length and swap with the other copy ~__ evaluate the number and make 2 copies Ab convert to base A=10 (array of digits), it uses the absolute value ( take out the first digit @g* get a copy of the number, get its sign and multiply with the digit \~ dump the other digits on the stack ]:+ add everything together 

Brachylog, 35 32 bytes

 lL,?:ef+:?:L+I,(0>?h:2*:Ir-:1+.;I.) lL,(0>?h:1--I;I0),?b:ef+:?:L:I+. 

Explanation

lL, L is the length of the Input ( 0>? Input < 0 h:1--I I is (First digit - 1) * -1 ; Or I0 I is 0 ), ?b:ef+ Sum all digits of the Input :?:L:I+. Output = sum of digits + (Input minus first digit) + L + I 

XSLT 1.0 (without EXSLT), 673 bytes

<transform xmlns="http://www.w3.org/1999/XSL/Transform" version="1.0"><output method="text"/><param name="i"/><template match="/"><variable name="d"><variable name="s">0<if test="0>$i">1</if></variable><variable name="d"><call-template name="d"><with-param name="i" select="substring($i,$s+2)"/></call-template></variable><value-of select="substring($i,1,$s+1)+$d"/></variable><value-of select="$i+string-length($i)+$d"/></template><template name="d"><param name="i"/>0<if test="$i!=''"><variable name="d"><call-template name="d"><with-param name="i" select="substring($i,2)"/></call-template></variable><value-of select="substring($i,1,1)+$d"/></if></template></transform> 

Lightly inflated:

<transform xmlns="http://www.w3.org/1999/XSL/Transform" version="1.0"> <output method="text"/> <param name="i"/> <template match="/"> <variable name="d"> <variable name="s">0<if test="0&gt;$i">1</if></variable> <variable name="d"> <call-template name="d"> <with-param name="i" select="substring($i,$s+2)"/> </call-template> </variable> <value-of select="substring($i,1,$s+1)+$d"/> </variable> <value-of select="$i+string-length($i)+$d"/> </template> <template name="d"> <param name="i"/>0<if test="$i!=''"> <variable name="d"> <call-template name="d"> <with-param name="i" select="substring($i,2)"/> </call-template> </variable> <value-of select="substring($i,1,1)+$d"/> </if> </template> </transform> 

Run using xsltproc:

xsltproc --param i -87901 ild.xsl ild.xsl 

Yes, ild.xsl is passed twice: Once as the XSLT document and then as the XML document to be transformed. An input document must be present because an XSLT processor generally requires one to begin running. (XSLT is designed to define a transformation from an input document to an output document; running a transform solely with command-line parameters as I've done here is atypical.) For this program, any well-formed XML document will suffice as input, and, XSLT being an application of XML, any well-formed XSLT transform is by definition a well-formed XML document.

MATL, 20 bytes

tVtnw48-PZ}t0<?x_]vs 

Try it Online

All test cases

Explanation

 % Implicitly grab the input tV % Duplicate the input and convert to a string tn % Duplicate and find the length of this string w % Flip the top two stack elements to get us the string again 48- % Subtract 48 (ASCII 'O'). Yields a negative number for a negative sign % and digits otherwise P % Flip the resulting array Z} % Break the array up so each element is pushed to the stack t0<? % If the first character was a negative sign x_ % Pop the negative sign off the stack and negate the first digit ] % End of if vs % Vertically concatenate and sum all stack contents % Implicitly display the result 

Clojure, 102 bytes

(fn[n](load-string(str"(+ "n" "(count(str n))" "(apply str(map #(if(= % \-)%(str %" "))(str n)))")"))) 

Anonymous function which constructs a string that looks like (+ -123 4 -1 2 3 ) and evals it. Everything pretty verbose as it is, construct string from number, its length, and then map each symbol of string representation of the number except minus to itself plus space and minus remains the same

You can see it running here: https://ideone.com/FG4lsB

Dyalog APL, 19 17 16 bytes

≢+#⍎'\d'⎕R'&+',⊢ 

Takes string and returns

 ≢ length
+ plus
 # in root namespace
⍎ evaluation of
 '\d'⎕R'&+' regex append digits with a plus
, followed by
 ⊢ the unmodified string

–3 thanks to ngn

Ruby, 46 bytes

->{eval [n=gets.chop,n.size,*n.chars].join ?+} 

Edit: If not using lambda, 42 bytes

eval [n=gets.chop,n.size,*n.chars].join ?+ 

Edit2: Here's the tio for the case -123

Matlab, 76 67 bytes

n=input('');t=num2str(n)-48;if(n<0)t(1)=0;t(2)=-t(2);end n+sum(t+1) 

9 bytes saved thanks to @Luis Mendo

Explanation:

n=input(''); -- takes input t=num2str(n)-48; -- makes it a string and then array of digits with "-" becoming -3 (48 is code for 0) if(n<0) t(1)=0; -- set first element (-3) to 0 t(2)=-t(2); -- the second element is the most significant digit, so we have to negate it end n+sum(t+1) -- take sum of n, sum of all digits and length of t (guaranteed by +1 of every element) 

dc, 57 bytes

dc -e"0 1?rdsc*d[1r]s+d0>+dZr[+la10~lc*rdsaZ1<A]sAdsaZ1<Ala+++f" 

Explained:

0 1 # Push 0, then 1 on the stack ? # Wait for input from stdin # If input is negative, the leading minus will subtract 1 from 0 r # Swap (rotate) top two items on stack. # Stack status if input (`$') was... # positive negative # TOP 1 <- coefficient -> -1 # $ $ # 0 dsc # Store a copy of coefficient in `c' * # Multiply input by coefficient: # If input was positive, it stays positive. # If input was negative, it's actually interpreted as positive. # In this case, multiply by -1 to make it negative. d # Duplicate signed input [1r]s+ # Define a function `+': Push 1 and rotate d 0>+ # If input is negative, push 1 underneath the top of the stack # This 1 represents the length of the `-` in the input # Note that the stack now has 3 items on it, regardless of input sign dZ # Push the length of the input (not including leading minus) r # Rotate, moving a copy of the input to the top [ # Begin function definition + # Add top two items of stack la # Load value from `a' (which holds nothing at time of function definition) 10~ # Slice the last digit off `a' (spoiler: `a' is going to hold the input while # we gather its digits) lc* # Multiply digit by coefficient # Since the input is signed, the input modulo 10 will have the same sign. # We want all digits to be positive, except the leftmost digit, which should # have the sign of the input. # This ensures that each digit is positive. r # Rotate: move remaining digits to top of stack dsa # Store a copy of the remaining digits in `a' Z 1<A # Count the number of digits left; if more than 1, execute A ]sA # Store the function as `A' d sa # Store a copy of the input in `a' # Props to you if you're still reading this Z 1<A # Count the number of digits left; if more than 1, execute A la # Load leftmost digit of input (still signed appropriately) +++ # Add the top four items on the stack f # Dump stack 

This was far more complicated than I had expected! Good challenge :)

Bash + coreutils, 36 bytes

bc<<<$1+${#1}+$(sed s:\\B:+:g<<<0$1) 

Explanation:

 $1+ # the input number (+) ${#1}+ # the length of the number, the '-' sign included (+) $(sed s:\\B:+:g<<<0$1) # insert '+' between two consecutive word characters #A word character is any letter, digit or underscore. bc<<< # calculate the sum 

In sed, \B also matches between two consecutive non-word characters, so for a negative number it matches between '^' and '-'. Note the 0$1 trick needed for \B to give 0-1+2+3, for example.

Run example: 'input.txt' contains all the test cases in the question's statement

while read N;do echo "$N -> "$(./ILD_sum.sh "$N");done < input.txt 

Output:

87901 -> 87931 123 -> 132 99 -> 119 5 -> 11 1 -> 3 0 -> 1 -3 -> -4 -99 -> -96 -123 -> -115 -900 -> -905 -87901 -> -87886 

PowerShell v4, 48 bytes

param($n)$n,"$n".length+[char[]]"$n"-join'+'|iex 

_{This should work in v2+, but I only tested in v4.}

Takes input $n. Creates a new array with the , operator consisting of $n and the .length when $n is converted to a string. Concatenates with that the string $n cast as a char-array. Then, that whole array is -joined together with + before being piped to iex (similar to eval). The result is left on the pipeline and output is implicit.

For example, for input -123, the array would look like (-123, 4, -, 1, 2, 3), and the string after the -join would look like -123+4+-+1+2+3. Then the Invoke-Expression happens, and the result is -115 as expected.

Factor with load-all, 175 bytes

Well, this is not very short. The special handling of unary minus is really annoying; I guess I could do it better and I will, maybe.

[ dup [ 10 >base length ] [ [ 10 >base >array [ 48 - ] V{ } map-as ] [ 0 < ] bi [ reverse dup pop* dup pop swap [ neg ] dip dup [ push ] dip ] [ ] if 0 [ + ] reduce ] bi + + ] 

Using this substitution regex:

s/(-?[\d]+)\s*->\s*(-?[\d]+)/{ $2 } [ $1 calculate-ild ] unit-test/g 

We can turn the OP's test cases into a Factor test suite.

USING: arrays kernel math math.parser sequences ; IN: sum-ild : sum-digits ( n -- x ) [ number>string >array [ 48 - ] V{ } map-as ] [ 0 < ] bi [ reverse dup pop* dup pop swap [ neg ] dip dup [ push ] dip ] [ ] if 0 [ + ] reduce ; : calculate-ild ( n -- x ) dup [ number>string length ] [ sum-digits ] bi + + ; USING: tools.test sum-ild ; IN: sum-ild.tests { 87931 } [ 87901 calculate-ild ] unit-test { 132 } [ 123 calculate-ild ] unit-test { 119 } [ 99 calculate-ild ] unit-test { 11 } [ 5 calculate-ild ] unit-test { 3 } [ 1 calculate-ild ] unit-test { 1 } [ 0 calculate-ild ] unit-test { -4 } [ -3 calculate-ild ] unit-test { -115 } [ -123 calculate-ild ] unit-test { -905 } [ -900 calculate-ild ] unit-test { -87886 } [ -87901 calculate-ild ] unit-test 

C#, 118 bytes

int k(int a){var s=a.ToString();for(int i=0;i<s.Length;a+=s[i]<46?-(s[++i]-48)+ ++i-i:(s[i++]-48));return a+s.Length;} 

SpecBAS - 147 bytes

1 INPUT a$: l=LEN a$: b$="text "+a$+"+"+STR$ l+"+": FOR i=1 TO l: b$=b$+a$(i)+("+" AND i<l): NEXT i: EXECUTE b$ 

Builds up a string which then gets run. Unfortunately EXECUTE doesn't work with the ? shorthand for PRINT, but TEXT saved 1 character.

C#, 106 bytes

I beat java my a byte, my life is complete

int r(int n){var s=n+"";return n+s.Length+s.Select((k,j)=>int.Parse(s[k==45?1:j]+"")*(k==45?-2:1)).Sum();} 

Ungolfed (kinda)

 public static int r(int n) { var s = n + ""; return n + s.Length + s.Select((k, j) =>int.Parse(s[k==45?1:j]+"")*(k==45?-2:1)).Sum(); } 

Perl 5, 22 + 1 (-p) = 23 bytes

$_+=eval s//+/gr.y///c 

Try it online!

Java 8, 174 136 122 107 105 93 78 bytes

i->{int f=0;for(int j:(i+"").getBytes())i+=j<48?f++:f-->0?50-j:j-47;return i;} 

-14 bytes thanks to @LeakyNun.
-15 bytes thanks to @cliffroot.

Explanation:

Try it online.

i->{ // Method with integer as both parameter and return-type int f=0; // Integer-flag, starting at 0 for(int j:(i+"").getBytes()) // Loop over the digits as bytes i+= // Increase the input with: j<48? // If the current byte is '-': f++ // Increase the input with the flag-integer `f` (which is 0), // and increase the flag-integer `f` by 1 afterwards : // Else: f-->0? // If the flag-integer `f` is 1, // and decrease the flag-integer `f` back to 0 afterwards 50-j // Increase it with 50 minus the current byte : // Else j-47; // Increase it with the byte as digit // + 1 to cover for the length part in ILD return i;} // Return the modified input as result 

Perl 6 - 30 bytes

As literal as it gets

{$^a+$^a.chars+[+]($^a.comb)} 

Use it as an anonymous function

> {$^a+$^a.chars+[+]($^a.comb)}(99) 119 

JavaScript (ES6), 38 bytes

n=>eval([n+=``,n.length,...n].join`+`) 

Uses the old join-and-eval trick. Save 4 bytes if I can insist on string input:

f= n=>eval([n,n.length,...n].join`+`) ;

<input type=number oninput=o.value=f(this.value)><input id=o readonly>

C++, 255 Bytes

#include <iostream> #include <string> #include <stdlib.h> using namespace std; int main(){ string input; cin >> input; int sum = atoi(input.c_str()) + input.length(); for(unsigned i = 0; i < input.length(); ++i) sum += input.at(i) - 48; return 0; } 

Perl 5 - 37 Bytes

warn eval(join'+',/./g)+($_+=()=/./g) 

Input is in $_

Javascript (using external library) (45 bytes)

Using a library I wrote to bring LINQ to Javascript, I was able to write the following:

(i)=>i+(i+"").length+_.From(i+"").Sum(x=>x|0) 

C, 132 116 113 80

t,c;f(char*v){for(c=atoi(v+=t=*v==45);*v;t=0,++v)c+=t?50-*v-2*c:*v-47;return c;} 

Function f() takes the input as a string and returns the result as an integer. Full program version (113 bytes):

t;main(int c,char**v){char*p=v[1];c=atoi(p+=t=*p==45);for(c=t?-c:c;*p;++p,t=0)c+=t?50-*p:*p-47;printf("%d\n",c);} 

Requires one argument.

Perl, 27 bytes

22 bytes code + 5 for -paF.

$"="+";$_+=@F+eval"@F" 

Explanation

Uses the -a autosplit option with an empty delimiter (-F) creating an array of the digits passed in. Uses the magic variable $" which controls which char is used to join an array when it's interpolated into a string (we use "+" here) and the fact that a list used in scalar context will return the length of the list (the number of digits).

Usage

echo -n 99 | perl -paF -e'$"="+";$_+=@F+eval"@F"' 119 

Perl, 27 bytes

22 bytes code + 5 for -paF.

Alternative solution, that's a lot more readable for no more bytes. I prefer the other as it looks more cryptic!

$_+=@F+eval join"+",@F 

dc, 56 bytes

?dZrdd1sa[1+r0r-_1sa]sb0>b[A~rd0<x]dsxxrla*[+z1<y]dsyxp 

No shorter than Joe's above, but a somewhat different implementation (and one that takes negative numbers as input vs. a subtraction command). Can probably be golfed more, but lunch only lasts so long.

? #input dZrdd #find no. of digits, rotate to bottom of stack, dup input twice 1sa #coefficient for first digit stored in register 'a' [1+r0r-_1sa]sb #macro 'b' executes on negative numbers. add one (for the neg. sign) #rotate this value out of the way, leave a positive copy on top 0>b #run the above macro if negative [A~rd0<x]dsxx #create and run macro 'x'; mod 10 to grab least significant digit #keep doing it if quotient is greater than zero rla* #a zero remains in the way of our most significant digit, rotate it down #and multiply said digit by our coefficient 'a' from earlier [+z1<y]dsyx #add two top stack values (we left that zero there to ensure this always #works), check stack depth and keep doing it while there's stack p #print! 

R, 108 bytes

A bit late to the party again but here it goes:

s=strsplit(paste(n<-scan()),"")[[1]];n+nchar(n)+sum(as.integer(if(n<0)c(paste0(s[1],s[2]),s[1:2*-1])else s)) 

In order to generally split the digits of any number (e.g. to sum them), R requires us to first convert to a string and subsequently split the string into a string vector. To sum up the elements, the string vector has to be converted to numeric or integer. This together with the exception with the sum of a the digits of a negative number eats up a lot of bytes.

The exception can be golfed a bit (to 96 bytes) if warning messages are allowed.

s=as.integer(strsplit(paste(n<-scan()),"")[[1]]);if(n<0){s[2]=s[2]*-1;s=s[-1]};n+nchar(n)+sum(s) 

In this case the string vector is converted to integer directly using as.integer. However, for negative numbers the first element in the vector will be a minus sign: "-". This causes some trouble, e.g.: as.numeric(c("-",1,2,3)) will return NA 1 2 3 and a warning message. To circumvent this, remove the NA and then multiply the first element with -1 before taking the sum.